Dataset Viewer
Auto-converted to Parquet Duplicate
run
stringclasses
61 values
model
stringclasses
26 values
variant
stringclasses
6 values
input
stringclasses
2 values
sentence_id
stringclasses
21 values
text
stringclasses
21 values
text_nikud
stringclasses
21 values
english
stringclasses
21 values
focus
stringclasses
16 values
audio
audioduration (s)
0.81
40.9
hf_id
stringclasses
10 values
voice
stringclasses
24 values
sample_rate
int64
22.1k
48k
hardware
stringclasses
8 values
seconds_to_generate
float64
0.06
129
publisher
stringclasses
17 values
publisher_type
stringclasses
3 values
release_date
stringclasses
17 values
training_origin
stringclasses
11 values
licence
stringclasses
18 values
run_notes
stringclasses
47 values
phonikud-piper
phonikud-piper
main
nikud
s01
ืื ื™ ื”ื™ื™ืชื™ ืขื•ืฉื” ืืช ื–ื” ืฉื•ื‘.
ืึฒื ึดื™ ื”ึธื™ึดื™ืชึดื™ ืขื•ึนืฉึถื‚ื” ืึถืช ื–ึถื” ืฉืื•ึผื‘.
I would do it again.
Tatoeba grade 1 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.08
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s02
ื–ื” ืœื ืงืฉื” ืื ืืชื” ืœื•ืžื“.
ื–ึถื” ืœึนื ืงึธืฉึถืื” ืึดื ืึทืชึธึผื” ืœื•ึนืžึตื“.
It's not difficult if you study.
Tatoeba grade 1 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.07
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s03
ื”ืžืชื ื” ื”ื–ืืช ื”ื™ื ืœื—ื‘ืจ ืฉืœื™.
ื”ึทืžึทึผืชึธึผื ึธื” ื”ึทื–ึนึผืืช ื”ึดื™ื ืœึถื—ึธื‘ึตืจ ืฉึถืืœึดึผื™.
This gift is for my friend.
Tatoeba grade 2 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.07
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s04
ื”ื–ื”ืจ ืคืŸ ืชืฉื‘ื•ืจ ืืช ื”ื‘ื™ืฆื™ื.
ื”ึดื–ึธึผื”ึตืจ ืคึถึผืŸ ืชึดึผืฉึฐืื‘ึผื•ึนืจ ืึถืช ื”ึทื‘ึตึผื™ืฆึดื™ื.
Take care not to break the eggs.
Tatoeba grade 2 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.07
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s05
ื‘ืจื’ืข ืฉืชื”ื™ื” ืœื™ ื”ื–ื“ืžื ื•ืช, ืื‘ื•ื ืœื‘ืงืจ.
ื‘ึธึผืจึถื’ึทืข ืฉึถืืชึดึผื”ึฐื™ึถื” ืœึดื™ ื”ึดื–ึฐื“ึทึผืžึฐึผื ื•ึผืช, ืึธื‘ื•ึนื ืœึฐื‘ึทืงึตึผืจ.
As soon as I can get the chance, I'll come for a visit.
Tatoeba grade 3 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.09
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s06
ืงื™ื‘ืœื ื• ื”ื•ืจืื•ืช ืื™ืš ืœื”ื›ื™ืŸ ืกืœ ืžื‘ืžื‘ื•ืง.
ืงึดื™ื‘ึทึผืœึฐื ื•ึผ ื”ื•ึนืจึธืื•ึนืช ืึตื™ืšึฐ ืœึฐื”ึธื›ึดื™ืŸ ืกึทืœ ืžึดื‘ึทึผืžึฐื‘ึผื•ึผืง.
We received instructions on how to make a bamboo basket.
Tatoeba grade 3 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.09
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s07
ืขื•ืœืžื ื• ื”ื•ื ืจืง ื—ืœืง ืงื˜ืŸ ืžื”ื™ืงื•ื.
ืขื•ึนืœึธืžึตื ื•ึผ ื”ื•ึผื ืจึทืง ื—ึตืœึถืง ืงึธื˜ึธืŸ ืžึตื”ึทื™ึฐึผืงื•ึผื.
Our world is only one small part of the universe.
Tatoeba grade 4 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.09
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s08
ื”ืจ ืคื•ื’'ื™ ื”ื•ื ื”ื”ืจ ื”ื’ื‘ื•ื” ื‘ื™ื•ืชืจ ื‘ื™ืคืŸ.
ื”ึทืจ ืคึผื•ึผื’ึด'ื™ ื”ื•ึผื ื”ึธื”ึธืจ ื”ึทื’ึธึผื‘ื•ึนื”ึทึผ ื‘ึฐึผื™ื•ึนืชึตืจ ื‘ึฐึผื™ึทืคึธึผืŸ.
Mt. Fuji is the highest mountain in Japan.
Tatoeba grade 4 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.11
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s09
ื’ื•ื’ืœ ืฉื™ื—ืงื” ืชืคืงื™ื“ ื—ืฉื•ื‘ ื‘ื”ื ื’ืฉืช ื”ืžื™ื“ืข ื‘ืื™ื ื˜ืจื ื˜.
ื’ึผื•ึผื’ึถืœ ืฉึดื‚ื™ื—ึฒืงึธื” ืชึทึผืคึฐืงึดื™ื“ ื—ึธืฉืื•ึผื‘ ื‘ึฐึผื”ึทื ึฐื’ึธึผืฉึทืืช ื”ึทืžึตึผื™ื“ึธืข ื‘ึธึผืึดื™ื ึฐื˜ึถืจึฐื ึตื˜.
Google has played an important role in making information more accessible over the internet.
Tatoeba grade 5 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.12
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s10
ืงืฉื” ืœื”ืชื ืขืจ ืžื”ืจื’ืœ ืœืื—ืจ ืฉื”ื•ื ืžืชืงื‘ืข.
ืงึธืฉึถืื” ืœึฐื”ึดืชึฐื ึทืขึตืจ ืžึตื”ึถืจึฐื’ึตึผืœ ืœึฐืึทื—ึทืจ ืฉึถืื”ื•ึผื ืžึดืชึฐืงึทื‘ึตึผืขึท.
A habit is very difficult to shake off once it is formed.
Tatoeba grade 5 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.09
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s11
ืื ื™ ืงื•ืจื ืกืคืจ ื‘ืขืจื‘ ืœื™ื“ ื”ื—ืœื•ืŸ.
ืึฒื ึดื™ ืงื•ึนืจึตื ืกึตืคึถืจ ื‘ึธึผืขึถืจึถื‘ ืœึฐื™ึทื“ ื”ึทื—ึทืœึผื•ึนืŸ.
I read a book in the evening by the window.
stress: segolates sefer, erev
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.08
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s12
ื”ืกืคืจ ืžื”ืฉื›ื•ื ื” ืกืคืจ ืืช ื”ื›ืกืฃ ื•ืžื™ื“ ืกื™ืคืจ ืœื™ ืืช ื›ืœ ื”ืกื™ืคื•ืจ.
ื”ึทืกึทึผืคึธึผืจ ืžึตื”ึทืฉึฐึผืื›ื•ึผื ึธื” ืกึธืคึทืจ ืึถืช ื”ึทื›ึถึผืกึถืฃ ื•ึผืžึดื™ึธึผื“ ืกึดื™ืคึตึผืจ ืœึดื™ ืึถืช ื›ื‡ึผืœ ื”ึทืกึดึผื™ืคึผื•ึผืจ.
The barber from the neighbourhood counted the money and right away told me the whole story.
homographs: ืกืคืจ as sapar, safar
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.13
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s13
ื”ื™ืœื“ ื”ืœืš ืœื’ืŸ ื‘ื‘ื•ืงืจ ืขื ื”ื›ืœื‘.
ื”ึทื™ึถึผืœึถื“ ื”ึธืœึทืšึฐ ืœึทื’ึทึผืŸ ื‘ึทึผื‘ึผื•ึนืงึถืจ ืขึดื ื”ึทื›ึถึผืœึถื‘.
The boy went to kindergarten in the morning with the dog.
stress: segolates yeled, boker, kelev
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.08
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s14
ืื ื™ ืจื•ืฆื” ืœืื›ื•ืœ ืœื—ื ืขื ื—ืžืื” ื•ืฉืžืŸ ื–ื™ืช.
ืึฒื ึดื™ ืจื•ึนืฆึธื” ืœึถืึฑื›ื•ึนืœ ืœึถื—ึถื ืขึดื ื—ึถืžึฐืึธื” ื•ึฐืฉึถืืžึถืŸ ื–ึทื™ึดืช.
I want to eat bread with butter and olive oil.
stress: segolates lechem, shemen
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.1
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s15
ื”ืคื’ื™ืฉื” ื ืงื‘ืขื” ืœื™ื•ื ืฉืœื™ืฉื™ ื‘ืฉืขื” ืฉืชื™ื™ื ื•ื—ืฆื™.
ื”ึทืคึฐึผื’ึดื™ืฉึธืื” ื ึดืงึฐื‘ึฐึผืขึธื” ืœึฐื™ื•ึนื ืฉึฐืืœึดื™ืฉึดืื™ ื‘ึฐึผืฉึธืืขึธื” ืฉึฐืืชึทึผื™ึดื™ื ื•ึธื—ึตืฆึดื™.
The meeting was set for Tuesday at half past two.
numbers and times in words
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.11
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s16
ืฉืžื™ ื“ื ื™ืืœ ืจื•ื–ื”ื™ืœ ื•ืื ื™ ื’ืจ ื‘ื™ืจื•ืฉืœื™ื.
ืฉึฐืืžึดื™ ื“ึธึผื ึดื™ึตึผืึตืœ ืจื•ึนื–ึฐื”ึดื™ืœ ื•ึทืึฒื ึดื™ ื’ึธึผืจ ื‘ึดึผื™ืจื•ึผืฉึธืืœึทื™ึดื.
My name is Daniel Rosehill and I live in Jerusalem.
names: a foreign surname in Hebrew script
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.09
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s17
ืื ื™ ืžืืžืŸ ืžื•ื“ืœ ื–ื™ื”ื•ื™ ื“ื™ื‘ื•ืจ ืขืœ ื”ืงืœื˜ื•ืช ืฉืœ ืขืฆืžื™.
ืึฒื ึดื™ ืžึฐืึทืžึตึผืŸ ืžื•ึนื“ึถืœ ื–ึดื™ื”ื•ึผื™ ื“ึดึผื™ื‘ึผื•ึผืจ ืขึทืœ ื”ึทืงึฐืœึธื˜ื•ึนืช ืฉึถืืœ ืขึทืฆึฐืžึดื™.
I'm training a speech recognition model on my own recordings.
register: technical talk
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.11
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s18
ืชืฉื™ื ืืช ื”ืืจื’ื– ืขืœ ื”ืžืฉื˜ื— ื•ืชืขื‘ื™ืจ ืื•ืชื• ืขื ื”ืžืœื’ื–ื” ืœืžื—ืกืŸ.
ืชึธึผืฉึดื‚ื™ื ืึถืช ื”ึธืึทืจึฐื’ึธึผื– ืขึทืœ ื”ึทืžึดึผืฉึฐืื˜ึธื— ื•ึฐืชึทืขึฒื‘ึดื™ืจ ืื•ึนืชื•ึน ืขึดื ื”ึทืžึทึผืœึฐื’ึตึผื–ึธื” ืœึทืžึทึผื—ึฐืกึธืŸ.
Put the crate on the pallet and take it to the warehouse with the forklift.
domain: material handling
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.14
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s19
ืฉืœื—ืชื™ ืœืš ืžื™ื™ืœ ืขื ื”ืœื™ื ืง ืœืงื•ื‘ืฅ ื‘ื“ืจื™ื™ื‘.
ืฉึธืืœึทื—ึฐืชึดึผื™ ืœึฐืšึธ ืžึตื™ื™ืœ ืขึดื ื”ึทืœึดึผื™ื ึฐืง ืœึทืงึผื•ึนื‘ึถืฅ ื‘ึทึผื“ึฐึผืจึทื™ื™ื‘.
I sent you an email with the link to the file on the Drive.
loanwords: English words in Hebrew
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.1
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper
phonikud-piper
main
nikud
s20
ืื ืœื ื™ื™ืจื“ ื’ืฉื ืžื—ืจ, ื ื™ืกืข ืœื™ื, ื ืื›ืœ ืืจื•ื—ืช ืฆื”ืจื™ื™ื ื•ื ื—ื–ื•ืจ ืœืคื ื™ ื”ืฉืงื™ืขื”.
ืึดื ืœึนื ื™ึตื™ืจึตื“ ื’ึถึผืฉึถืื ืžึธื—ึธืจ, ื ึดื™ืกึทึผืข ืœึทื™ึธึผื, ื ึนืื›ึทืœ ืึฒืจื•ึผื—ึทืช ืฆื‡ื”ึณืจึทื™ึดื™ื ื•ึฐื ึทื—ึฐื–ื•ึนืจ ืœึดืคึฐื ึตื™ ื”ึทืฉึฐึผืืงึดื™ืขึธื”.
If it doesn't rain tomorrow, we'll go to the beach, have lunch and come back before sunset.
prosody: a long sentence with commas
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.17
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
versions {'phonikud': '0.4.1', 'phonikud-onnx': '1.0.6', 'piper-onnx': '1.0.6', 'onnxruntime': '1.32.0'}; diacritizer Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx. Main wavs: text_nikud (Dicta vowels kept) plus hataama U+05AB and vocal shva U+05BD copied from the Phonikud diacr...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s01
ืื ื™ ื”ื™ื™ืชื™ ืขื•ืฉื” ืืช ื–ื” ืฉื•ื‘.
ืึฒื ึดื™ ื”ึธื™ึดื™ืชึดื™ ืขื•ึนืฉึถื‚ื” ืึถืช ื–ึถื” ืฉืื•ึผื‘.
I would do it again.
Tatoeba grade 1 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.18
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s02
ื–ื” ืœื ืงืฉื” ืื ืืชื” ืœื•ืžื“.
ื–ึถื” ืœึนื ืงึธืฉึถืื” ืึดื ืึทืชึธึผื” ืœื•ึนืžึตื“.
It's not difficult if you study.
Tatoeba grade 1 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.15
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s03
ื”ืžืชื ื” ื”ื–ืืช ื”ื™ื ืœื—ื‘ืจ ืฉืœื™.
ื”ึทืžึทึผืชึธึผื ึธื” ื”ึทื–ึนึผืืช ื”ึดื™ื ืœึถื—ึธื‘ึตืจ ืฉึถืืœึดึผื™.
This gift is for my friend.
Tatoeba grade 2 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.15
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s04
ื”ื–ื”ืจ ืคืŸ ืชืฉื‘ื•ืจ ืืช ื”ื‘ื™ืฆื™ื.
ื”ึดื–ึธึผื”ึตืจ ืคึถึผืŸ ืชึดึผืฉึฐืื‘ึผื•ึนืจ ืึถืช ื”ึทื‘ึตึผื™ืฆึดื™ื.
Take care not to break the eggs.
Tatoeba grade 2 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.15
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s05
ื‘ืจื’ืข ืฉืชื”ื™ื” ืœื™ ื”ื–ื“ืžื ื•ืช, ืื‘ื•ื ืœื‘ืงืจ.
ื‘ึธึผืจึถื’ึทืข ืฉึถืืชึดึผื”ึฐื™ึถื” ืœึดื™ ื”ึดื–ึฐื“ึทึผืžึฐึผื ื•ึผืช, ืึธื‘ื•ึนื ืœึฐื‘ึทืงึตึผืจ.
As soon as I can get the chance, I'll come for a visit.
Tatoeba grade 3 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.2
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s06
ืงื™ื‘ืœื ื• ื”ื•ืจืื•ืช ืื™ืš ืœื”ื›ื™ืŸ ืกืœ ืžื‘ืžื‘ื•ืง.
ืงึดื™ื‘ึทึผืœึฐื ื•ึผ ื”ื•ึนืจึธืื•ึนืช ืึตื™ืšึฐ ืœึฐื”ึธื›ึดื™ืŸ ืกึทืœ ืžึดื‘ึทึผืžึฐื‘ึผื•ึผืง.
We received instructions on how to make a bamboo basket.
Tatoeba grade 3 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.18
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s07
ืขื•ืœืžื ื• ื”ื•ื ืจืง ื—ืœืง ืงื˜ืŸ ืžื”ื™ืงื•ื.
ืขื•ึนืœึธืžึตื ื•ึผ ื”ื•ึผื ืจึทืง ื—ึตืœึถืง ืงึธื˜ึธืŸ ืžึตื”ึทื™ึฐึผืงื•ึผื.
Our world is only one small part of the universe.
Tatoeba grade 4 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.18
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s08
ื”ืจ ืคื•ื’'ื™ ื”ื•ื ื”ื”ืจ ื”ื’ื‘ื•ื” ื‘ื™ื•ืชืจ ื‘ื™ืคืŸ.
ื”ึทืจ ืคึผื•ึผื’ึด'ื™ ื”ื•ึผื ื”ึธื”ึธืจ ื”ึทื’ึธึผื‘ื•ึนื”ึทึผ ื‘ึฐึผื™ื•ึนืชึตืจ ื‘ึฐึผื™ึทืคึธึผืŸ.
Mt. Fuji is the highest mountain in Japan.
Tatoeba grade 4 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.2
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s09
ื’ื•ื’ืœ ืฉื™ื—ืงื” ืชืคืงื™ื“ ื—ืฉื•ื‘ ื‘ื”ื ื’ืฉืช ื”ืžื™ื“ืข ื‘ืื™ื ื˜ืจื ื˜.
ื’ึผื•ึผื’ึถืœ ืฉึดื‚ื™ื—ึฒืงึธื” ืชึทึผืคึฐืงึดื™ื“ ื—ึธืฉืื•ึผื‘ ื‘ึฐึผื”ึทื ึฐื’ึธึผืฉึทืืช ื”ึทืžึตึผื™ื“ึธืข ื‘ึธึผืึดื™ื ึฐื˜ึถืจึฐื ึตื˜.
Google has played an important role in making information more accessible over the internet.
Tatoeba grade 5 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.24
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s10
ืงืฉื” ืœื”ืชื ืขืจ ืžื”ืจื’ืœ ืœืื—ืจ ืฉื”ื•ื ืžืชืงื‘ืข.
ืงึธืฉึถืื” ืœึฐื”ึดืชึฐื ึทืขึตืจ ืžึตื”ึถืจึฐื’ึตึผืœ ืœึฐืึทื—ึทืจ ืฉึถืื”ื•ึผื ืžึดืชึฐืงึทื‘ึตึผืขึท.
A habit is very difficult to shake off once it is formed.
Tatoeba grade 5 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.2
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s11
ืื ื™ ืงื•ืจื ืกืคืจ ื‘ืขืจื‘ ืœื™ื“ ื”ื—ืœื•ืŸ.
ืึฒื ึดื™ ืงื•ึนืจึตื ืกึตืคึถืจ ื‘ึธึผืขึถืจึถื‘ ืœึฐื™ึทื“ ื”ึทื—ึทืœึผื•ึนืŸ.
I read a book in the evening by the window.
stress: segolates sefer, erev
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.16
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s12
ื”ืกืคืจ ืžื”ืฉื›ื•ื ื” ืกืคืจ ืืช ื”ื›ืกืฃ ื•ืžื™ื“ ืกื™ืคืจ ืœื™ ืืช ื›ืœ ื”ืกื™ืคื•ืจ.
ื”ึทืกึทึผืคึธึผืจ ืžึตื”ึทืฉึฐึผืื›ื•ึผื ึธื” ืกึธืคึทืจ ืึถืช ื”ึทื›ึถึผืกึถืฃ ื•ึผืžึดื™ึธึผื“ ืกึดื™ืคึตึผืจ ืœึดื™ ืึถืช ื›ื‡ึผืœ ื”ึทืกึดึผื™ืคึผื•ึผืจ.
The barber from the neighbourhood counted the money and right away told me the whole story.
homographs: ืกืคืจ as sapar, safar
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.29
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s13
ื”ื™ืœื“ ื”ืœืš ืœื’ืŸ ื‘ื‘ื•ืงืจ ืขื ื”ื›ืœื‘.
ื”ึทื™ึถึผืœึถื“ ื”ึธืœึทืšึฐ ืœึทื’ึทึผืŸ ื‘ึทึผื‘ึผื•ึนืงึถืจ ืขึดื ื”ึทื›ึถึผืœึถื‘.
The boy went to kindergarten in the morning with the dog.
stress: segolates yeled, boker, kelev
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.16
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s14
ืื ื™ ืจื•ืฆื” ืœืื›ื•ืœ ืœื—ื ืขื ื—ืžืื” ื•ืฉืžืŸ ื–ื™ืช.
ืึฒื ึดื™ ืจื•ึนืฆึธื” ืœึถืึฑื›ื•ึนืœ ืœึถื—ึถื ืขึดื ื—ึถืžึฐืึธื” ื•ึฐืฉึถืืžึถืŸ ื–ึทื™ึดืช.
I want to eat bread with butter and olive oil.
stress: segolates lechem, shemen
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.21
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s15
ื”ืคื’ื™ืฉื” ื ืงื‘ืขื” ืœื™ื•ื ืฉืœื™ืฉื™ ื‘ืฉืขื” ืฉืชื™ื™ื ื•ื—ืฆื™.
ื”ึทืคึฐึผื’ึดื™ืฉึธืื” ื ึดืงึฐื‘ึฐึผืขึธื” ืœึฐื™ื•ึนื ืฉึฐืืœึดื™ืฉึดืื™ ื‘ึฐึผืฉึธืืขึธื” ืฉึฐืืชึทึผื™ึดื™ื ื•ึธื—ึตืฆึดื™.
The meeting was set for Tuesday at half past two.
numbers and times in words
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.25
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s16
ืฉืžื™ ื“ื ื™ืืœ ืจื•ื–ื”ื™ืœ ื•ืื ื™ ื’ืจ ื‘ื™ืจื•ืฉืœื™ื.
ืฉึฐืืžึดื™ ื“ึธึผื ึดื™ึตึผืึตืœ ืจื•ึนื–ึฐื”ึดื™ืœ ื•ึทืึฒื ึดื™ ื’ึธึผืจ ื‘ึดึผื™ืจื•ึผืฉึธืืœึทื™ึดื.
My name is Daniel Rosehill and I live in Jerusalem.
names: a foreign surname in Hebrew script
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.2
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s17
ืื ื™ ืžืืžืŸ ืžื•ื“ืœ ื–ื™ื”ื•ื™ ื“ื™ื‘ื•ืจ ืขืœ ื”ืงืœื˜ื•ืช ืฉืœ ืขืฆืžื™.
ืึฒื ึดื™ ืžึฐืึทืžึตึผืŸ ืžื•ึนื“ึถืœ ื–ึดื™ื”ื•ึผื™ ื“ึดึผื™ื‘ึผื•ึผืจ ืขึทืœ ื”ึทืงึฐืœึธื˜ื•ึนืช ืฉึถืืœ ืขึทืฆึฐืžึดื™.
I'm training a speech recognition model on my own recordings.
register: technical talk
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.24
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s18
ืชืฉื™ื ืืช ื”ืืจื’ื– ืขืœ ื”ืžืฉื˜ื— ื•ืชืขื‘ื™ืจ ืื•ืชื• ืขื ื”ืžืœื’ื–ื” ืœืžื—ืกืŸ.
ืชึธึผืฉึดื‚ื™ื ืึถืช ื”ึธืึทืจึฐื’ึธึผื– ืขึทืœ ื”ึทืžึดึผืฉึฐืื˜ึธื— ื•ึฐืชึทืขึฒื‘ึดื™ืจ ืื•ึนืชื•ึน ืขึดื ื”ึทืžึทึผืœึฐื’ึตึผื–ึธื” ืœึทืžึทึผื—ึฐืกึธืŸ.
Put the crate on the pallet and take it to the warehouse with the forklift.
domain: material handling
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.26
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s19
ืฉืœื—ืชื™ ืœืš ืžื™ื™ืœ ืขื ื”ืœื™ื ืง ืœืงื•ื‘ืฅ ื‘ื“ืจื™ื™ื‘.
ืฉึธืืœึทื—ึฐืชึดึผื™ ืœึฐืšึธ ืžึตื™ื™ืœ ืขึดื ื”ึทืœึดึผื™ื ึฐืง ืœึทืงึผื•ึนื‘ึถืฅ ื‘ึทึผื“ึฐึผืจึทื™ื™ื‘.
I sent you an email with the link to the file on the Drive.
loanwords: English words in Hebrew
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.22
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-auto
phonikud-piper
its own nikud
plain
s20
ืื ืœื ื™ื™ืจื“ ื’ืฉื ืžื—ืจ, ื ื™ืกืข ืœื™ื, ื ืื›ืœ ืืจื•ื—ืช ืฆื”ืจื™ื™ื ื•ื ื—ื–ื•ืจ ืœืคื ื™ ื”ืฉืงื™ืขื”.
ืึดื ืœึนื ื™ึตื™ืจึตื“ ื’ึถึผืฉึถืื ืžึธื—ึธืจ, ื ึดื™ืกึทึผืข ืœึทื™ึธึผื, ื ึนืื›ึทืœ ืึฒืจื•ึผื—ึทืช ืฆื‡ื”ึณืจึทื™ึดื™ื ื•ึฐื ึทื—ึฐื–ื•ึนืจ ืœึดืคึฐื ึตื™ ื”ึทืฉึฐึผืืงึดื™ืขึธื”.
If it doesn't rain tomorrow, we'll go to the beach, have lunch and come back before sunset.
prosody: a long sentence with commas
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, piper-onnx defaults)
22,050
none (cpu=4)
0.39
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Input is plain text; the nikud is Phonikud's own (Phonikud/phonikud-onnx@b806189fe1fc0085b1012b7560ffb5e8ecfd72a2 phonikud-1.0.int8.onnx add_diacritics, incl. its vowels, hataama and vocal shva), then phonikud.phonemize, then Piper. Our Dicta nikud is not used. nikud.json has Phonikud's diacritized text next to ours pe...
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s01
ืื ื™ ื”ื™ื™ืชื™ ืขื•ืฉื” ืืช ื–ื” ืฉื•ื‘.
ืึฒื ึดื™ ื”ึธื™ึดื™ืชึดื™ ืขื•ึนืฉึถื‚ื” ืึถืช ื–ึถื” ืฉืื•ึผื‘.
I would do it again.
Tatoeba grade 1 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.08
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s02
ื–ื” ืœื ืงืฉื” ืื ืืชื” ืœื•ืžื“.
ื–ึถื” ืœึนื ืงึธืฉึถืื” ืึดื ืึทืชึธึผื” ืœื•ึนืžึตื“.
It's not difficult if you study.
Tatoeba grade 1 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.07
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s03
ื”ืžืชื ื” ื”ื–ืืช ื”ื™ื ืœื—ื‘ืจ ืฉืœื™.
ื”ึทืžึทึผืชึธึผื ึธื” ื”ึทื–ึนึผืืช ื”ึดื™ื ืœึถื—ึธื‘ึตืจ ืฉึถืืœึดึผื™.
This gift is for my friend.
Tatoeba grade 2 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.07
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s04
ื”ื–ื”ืจ ืคืŸ ืชืฉื‘ื•ืจ ืืช ื”ื‘ื™ืฆื™ื.
ื”ึดื–ึธึผื”ึตืจ ืคึถึผืŸ ืชึดึผืฉึฐืื‘ึผื•ึนืจ ืึถืช ื”ึทื‘ึตึผื™ืฆึดื™ื.
Take care not to break the eggs.
Tatoeba grade 2 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.07
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s05
ื‘ืจื’ืข ืฉืชื”ื™ื” ืœื™ ื”ื–ื“ืžื ื•ืช, ืื‘ื•ื ืœื‘ืงืจ.
ื‘ึธึผืจึถื’ึทืข ืฉึถืืชึดึผื”ึฐื™ึถื” ืœึดื™ ื”ึดื–ึฐื“ึทึผืžึฐึผื ื•ึผืช, ืึธื‘ื•ึนื ืœึฐื‘ึทืงึตึผืจ.
As soon as I can get the chance, I'll come for a visit.
Tatoeba grade 3 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.09
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s06
ืงื™ื‘ืœื ื• ื”ื•ืจืื•ืช ืื™ืš ืœื”ื›ื™ืŸ ืกืœ ืžื‘ืžื‘ื•ืง.
ืงึดื™ื‘ึทึผืœึฐื ื•ึผ ื”ื•ึนืจึธืื•ึนืช ืึตื™ืšึฐ ืœึฐื”ึธื›ึดื™ืŸ ืกึทืœ ืžึดื‘ึทึผืžึฐื‘ึผื•ึผืง.
We received instructions on how to make a bamboo basket.
Tatoeba grade 3 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.09
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s07
ืขื•ืœืžื ื• ื”ื•ื ืจืง ื—ืœืง ืงื˜ืŸ ืžื”ื™ืงื•ื.
ืขื•ึนืœึธืžึตื ื•ึผ ื”ื•ึผื ืจึทืง ื—ึตืœึถืง ืงึธื˜ึธืŸ ืžึตื”ึทื™ึฐึผืงื•ึผื.
Our world is only one small part of the universe.
Tatoeba grade 4 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.09
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s08
ื”ืจ ืคื•ื’'ื™ ื”ื•ื ื”ื”ืจ ื”ื’ื‘ื•ื” ื‘ื™ื•ืชืจ ื‘ื™ืคืŸ.
ื”ึทืจ ืคึผื•ึผื’ึด'ื™ ื”ื•ึผื ื”ึธื”ึธืจ ื”ึทื’ึธึผื‘ื•ึนื”ึทึผ ื‘ึฐึผื™ื•ึนืชึตืจ ื‘ึฐึผื™ึทืคึธึผืŸ.
Mt. Fuji is the highest mountain in Japan.
Tatoeba grade 4 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.11
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s09
ื’ื•ื’ืœ ืฉื™ื—ืงื” ืชืคืงื™ื“ ื—ืฉื•ื‘ ื‘ื”ื ื’ืฉืช ื”ืžื™ื“ืข ื‘ืื™ื ื˜ืจื ื˜.
ื’ึผื•ึผื’ึถืœ ืฉึดื‚ื™ื—ึฒืงึธื” ืชึทึผืคึฐืงึดื™ื“ ื—ึธืฉืื•ึผื‘ ื‘ึฐึผื”ึทื ึฐื’ึธึผืฉึทืืช ื”ึทืžึตึผื™ื“ึธืข ื‘ึธึผืึดื™ื ึฐื˜ึถืจึฐื ึตื˜.
Google has played an important role in making information more accessible over the internet.
Tatoeba grade 5 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.12
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s10
ืงืฉื” ืœื”ืชื ืขืจ ืžื”ืจื’ืœ ืœืื—ืจ ืฉื”ื•ื ืžืชืงื‘ืข.
ืงึธืฉึถืื” ืœึฐื”ึดืชึฐื ึทืขึตืจ ืžึตื”ึถืจึฐื’ึตึผืœ ืœึฐืึทื—ึทืจ ืฉึถืื”ื•ึผื ืžึดืชึฐืงึทื‘ึตึผืขึท.
A habit is very difficult to shake off once it is formed.
Tatoeba grade 5 sentence
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.09
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s11
ืื ื™ ืงื•ืจื ืกืคืจ ื‘ืขืจื‘ ืœื™ื“ ื”ื—ืœื•ืŸ.
ืึฒื ึดื™ ืงื•ึนืจึตื ืกึตืคึถืจ ื‘ึธึผืขึถืจึถื‘ ืœึฐื™ึทื“ ื”ึทื—ึทืœึผื•ึนืŸ.
I read a book in the evening by the window.
stress: segolates sefer, erev
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.08
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s12
ื”ืกืคืจ ืžื”ืฉื›ื•ื ื” ืกืคืจ ืืช ื”ื›ืกืฃ ื•ืžื™ื“ ืกื™ืคืจ ืœื™ ืืช ื›ืœ ื”ืกื™ืคื•ืจ.
ื”ึทืกึทึผืคึธึผืจ ืžึตื”ึทืฉึฐึผืื›ื•ึผื ึธื” ืกึธืคึทืจ ืึถืช ื”ึทื›ึถึผืกึถืฃ ื•ึผืžึดื™ึธึผื“ ืกึดื™ืคึตึผืจ ืœึดื™ ืึถืช ื›ื‡ึผืœ ื”ึทืกึดึผื™ืคึผื•ึผืจ.
The barber from the neighbourhood counted the money and right away told me the whole story.
homographs: ืกืคืจ as sapar, safar
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.13
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s13
ื”ื™ืœื“ ื”ืœืš ืœื’ืŸ ื‘ื‘ื•ืงืจ ืขื ื”ื›ืœื‘.
ื”ึทื™ึถึผืœึถื“ ื”ึธืœึทืšึฐ ืœึทื’ึทึผืŸ ื‘ึทึผื‘ึผื•ึนืงึถืจ ืขึดื ื”ึทื›ึถึผืœึถื‘.
The boy went to kindergarten in the morning with the dog.
stress: segolates yeled, boker, kelev
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.08
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s14
ืื ื™ ืจื•ืฆื” ืœืื›ื•ืœ ืœื—ื ืขื ื—ืžืื” ื•ืฉืžืŸ ื–ื™ืช.
ืึฒื ึดื™ ืจื•ึนืฆึธื” ืœึถืึฑื›ื•ึนืœ ืœึถื—ึถื ืขึดื ื—ึถืžึฐืึธื” ื•ึฐืฉึถืืžึถืŸ ื–ึทื™ึดืช.
I want to eat bread with butter and olive oil.
stress: segolates lechem, shemen
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.1
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s15
ื”ืคื’ื™ืฉื” ื ืงื‘ืขื” ืœื™ื•ื ืฉืœื™ืฉื™ ื‘ืฉืขื” ืฉืชื™ื™ื ื•ื—ืฆื™.
ื”ึทืคึฐึผื’ึดื™ืฉึธืื” ื ึดืงึฐื‘ึฐึผืขึธื” ืœึฐื™ื•ึนื ืฉึฐืืœึดื™ืฉึดืื™ ื‘ึฐึผืฉึธืืขึธื” ืฉึฐืืชึทึผื™ึดื™ื ื•ึธื—ึตืฆึดื™.
The meeting was set for Tuesday at half past two.
numbers and times in words
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.11
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s16
ืฉืžื™ ื“ื ื™ืืœ ืจื•ื–ื”ื™ืœ ื•ืื ื™ ื’ืจ ื‘ื™ืจื•ืฉืœื™ื.
ืฉึฐืืžึดื™ ื“ึธึผื ึดื™ึตึผืึตืœ ืจื•ึนื–ึฐื”ึดื™ืœ ื•ึทืึฒื ึดื™ ื’ึธึผืจ ื‘ึดึผื™ืจื•ึผืฉึธืืœึทื™ึดื.
My name is Daniel Rosehill and I live in Jerusalem.
names: a foreign surname in Hebrew script
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.09
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s17
ืื ื™ ืžืืžืŸ ืžื•ื“ืœ ื–ื™ื”ื•ื™ ื“ื™ื‘ื•ืจ ืขืœ ื”ืงืœื˜ื•ืช ืฉืœ ืขืฆืžื™.
ืึฒื ึดื™ ืžึฐืึทืžึตึผืŸ ืžื•ึนื“ึถืœ ื–ึดื™ื”ื•ึผื™ ื“ึดึผื™ื‘ึผื•ึผืจ ืขึทืœ ื”ึทืงึฐืœึธื˜ื•ึนืช ืฉึถืืœ ืขึทืฆึฐืžึดื™.
I'm training a speech recognition model on my own recordings.
register: technical talk
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.11
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s18
ืชืฉื™ื ืืช ื”ืืจื’ื– ืขืœ ื”ืžืฉื˜ื— ื•ืชืขื‘ื™ืจ ืื•ืชื• ืขื ื”ืžืœื’ื–ื” ืœืžื—ืกืŸ.
ืชึธึผืฉึดื‚ื™ื ืึถืช ื”ึธืึทืจึฐื’ึธึผื– ืขึทืœ ื”ึทืžึดึผืฉึฐืื˜ึธื— ื•ึฐืชึทืขึฒื‘ึดื™ืจ ืื•ึนืชื•ึน ืขึดื ื”ึทืžึทึผืœึฐื’ึตึผื–ึธื” ืœึทืžึทึผื—ึฐืกึธืŸ.
Put the crate on the pallet and take it to the warehouse with the forklift.
domain: material handling
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.14
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s19
ืฉืœื—ืชื™ ืœืš ืžื™ื™ืœ ืขื ื”ืœื™ื ืง ืœืงื•ื‘ืฅ ื‘ื“ืจื™ื™ื‘.
ืฉึธืืœึทื—ึฐืชึดึผื™ ืœึฐืšึธ ืžึตื™ื™ืœ ืขึดื ื”ึทืœึดึผื™ื ึฐืง ืœึทืงึผื•ึนื‘ึถืฅ ื‘ึทึผื“ึฐึผืจึทื™ื™ื‘.
I sent you an email with the link to the file on the Drive.
loanwords: English words in Hebrew
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.1
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
phonikud-piper-plain
phonikud-piper
no stress marks
nikud
s20
ืื ืœื ื™ื™ืจื“ ื’ืฉื ืžื—ืจ, ื ื™ืกืข ืœื™ื, ื ืื›ืœ ืืจื•ื—ืช ืฆื”ืจื™ื™ื ื•ื ื—ื–ื•ืจ ืœืคื ื™ ื”ืฉืงื™ืขื”.
ืึดื ืœึนื ื™ึตื™ืจึตื“ ื’ึถึผืฉึถืื ืžึธื—ึธืจ, ื ึดื™ืกึทึผืข ืœึทื™ึธึผื, ื ึนืื›ึทืœ ืึฒืจื•ึผื—ึทืช ืฆื‡ื”ึณืจึทื™ึดื™ื ื•ึฐื ึทื—ึฐื–ื•ึนืจ ืœึดืคึฐื ึตื™ ื”ึทืฉึฐึผืืงึดื™ืขึธื”.
If it doesn't rain tomorrow, we'll go to the beach, have lunch and come back before sunset.
prosody: a long sentence with commas
Phonikud/phonikud-tts-checkpoints
Phonikud Piper voice model.onnx (single voice, default settings of piper-onnx)
22,050
none (cpu=4)
0.17
Phonikud (HF/GitHub org; code under thewh1teagle/phonikud); paper authors Yakov Kolani, Maxim Melichov, Cobi Calev, Morris Alper
open-source community org
2025-06-02
UNRESOLVED for released voices. Paper (sec 5.2) says Piper/StyleTTS2 were FINE-TUNED from pre-trained English checkpoints on 20 h of synthetic Hebrew audio (Gemini 2.5 Pro) with Phonikud IPA. Model card is empty, so whether Shaul/Michael were instead trained from scratch on SASPEECH is not confirmed. Not a confirmed He...
Non-commercial (CC-NC) per model card; G2P code CC BY 4.0
Same voice and nikud, without the Phonikud stress and vocal-shva marks (plain phonemize()): how the recorder's Listen clips sounded before this bake-off.
piper-he-noa
piper-he-noa
main
plain
s01
ืื ื™ ื”ื™ื™ืชื™ ืขื•ืฉื” ืืช ื–ื” ืฉื•ื‘.
ืึฒื ึดื™ ื”ึธื™ึดื™ืชึดื™ ืขื•ึนืฉึถื‚ื” ืึถืช ื–ึถื” ืฉืื•ึผื‘.
I would do it again.
Tatoeba grade 1 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.12
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s02
ื–ื” ืœื ืงืฉื” ืื ืืชื” ืœื•ืžื“.
ื–ึถื” ืœึนื ืงึธืฉึถืื” ืึดื ืึทืชึธึผื” ืœื•ึนืžึตื“.
It's not difficult if you study.
Tatoeba grade 1 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.08
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s03
ื”ืžืชื ื” ื”ื–ืืช ื”ื™ื ืœื—ื‘ืจ ืฉืœื™.
ื”ึทืžึทึผืชึธึผื ึธื” ื”ึทื–ึนึผืืช ื”ึดื™ื ืœึถื—ึธื‘ึตืจ ืฉึถืืœึดึผื™.
This gift is for my friend.
Tatoeba grade 2 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.1
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s04
ื”ื–ื”ืจ ืคืŸ ืชืฉื‘ื•ืจ ืืช ื”ื‘ื™ืฆื™ื.
ื”ึดื–ึธึผื”ึตืจ ืคึถึผืŸ ืชึดึผืฉึฐืื‘ึผื•ึนืจ ืึถืช ื”ึทื‘ึตึผื™ืฆึดื™ื.
Take care not to break the eggs.
Tatoeba grade 2 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.1
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s05
ื‘ืจื’ืข ืฉืชื”ื™ื” ืœื™ ื”ื–ื“ืžื ื•ืช, ืื‘ื•ื ืœื‘ืงืจ.
ื‘ึธึผืจึถื’ึทืข ืฉึถืืชึดึผื”ึฐื™ึถื” ืœึดื™ ื”ึดื–ึฐื“ึทึผืžึฐึผื ื•ึผืช, ืึธื‘ื•ึนื ืœึฐื‘ึทืงึตึผืจ.
As soon as I can get the chance, I'll come for a visit.
Tatoeba grade 3 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.16
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s06
ืงื™ื‘ืœื ื• ื”ื•ืจืื•ืช ืื™ืš ืœื”ื›ื™ืŸ ืกืœ ืžื‘ืžื‘ื•ืง.
ืงึดื™ื‘ึทึผืœึฐื ื•ึผ ื”ื•ึนืจึธืื•ึนืช ืึตื™ืšึฐ ืœึฐื”ึธื›ึดื™ืŸ ืกึทืœ ืžึดื‘ึทึผืžึฐื‘ึผื•ึผืง.
We received instructions on how to make a bamboo basket.
Tatoeba grade 3 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.12
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s07
ืขื•ืœืžื ื• ื”ื•ื ืจืง ื—ืœืง ืงื˜ืŸ ืžื”ื™ืงื•ื.
ืขื•ึนืœึธืžึตื ื•ึผ ื”ื•ึผื ืจึทืง ื—ึตืœึถืง ืงึธื˜ึธืŸ ืžึตื”ึทื™ึฐึผืงื•ึผื.
Our world is only one small part of the universe.
Tatoeba grade 4 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.11
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s08
ื”ืจ ืคื•ื’'ื™ ื”ื•ื ื”ื”ืจ ื”ื’ื‘ื•ื” ื‘ื™ื•ืชืจ ื‘ื™ืคืŸ.
ื”ึทืจ ืคึผื•ึผื’ึด'ื™ ื”ื•ึผื ื”ึธื”ึธืจ ื”ึทื’ึธึผื‘ื•ึนื”ึทึผ ื‘ึฐึผื™ื•ึนืชึตืจ ื‘ึฐึผื™ึทืคึธึผืŸ.
Mt. Fuji is the highest mountain in Japan.
Tatoeba grade 4 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.12
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s09
ื’ื•ื’ืœ ืฉื™ื—ืงื” ืชืคืงื™ื“ ื—ืฉื•ื‘ ื‘ื”ื ื’ืฉืช ื”ืžื™ื“ืข ื‘ืื™ื ื˜ืจื ื˜.
ื’ึผื•ึผื’ึถืœ ืฉึดื‚ื™ื—ึฒืงึธื” ืชึทึผืคึฐืงึดื™ื“ ื—ึธืฉืื•ึผื‘ ื‘ึฐึผื”ึทื ึฐื’ึธึผืฉึทืืช ื”ึทืžึตึผื™ื“ึธืข ื‘ึธึผืึดื™ื ึฐื˜ึถืจึฐื ึตื˜.
Google has played an important role in making information more accessible over the internet.
Tatoeba grade 5 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.15
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s10
ืงืฉื” ืœื”ืชื ืขืจ ืžื”ืจื’ืœ ืœืื—ืจ ืฉื”ื•ื ืžืชืงื‘ืข.
ืงึธืฉึถืื” ืœึฐื”ึดืชึฐื ึทืขึตืจ ืžึตื”ึถืจึฐื’ึตึผืœ ืœึฐืึทื—ึทืจ ืฉึถืื”ื•ึผื ืžึดืชึฐืงึทื‘ึตึผืขึท.
A habit is very difficult to shake off once it is formed.
Tatoeba grade 5 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.12
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s11
ืื ื™ ืงื•ืจื ืกืคืจ ื‘ืขืจื‘ ืœื™ื“ ื”ื—ืœื•ืŸ.
ืึฒื ึดื™ ืงื•ึนืจึตื ืกึตืคึถืจ ื‘ึธึผืขึถืจึถื‘ ืœึฐื™ึทื“ ื”ึทื—ึทืœึผื•ึนืŸ.
I read a book in the evening by the window.
stress: segolates sefer, erev
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.1
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s12
ื”ืกืคืจ ืžื”ืฉื›ื•ื ื” ืกืคืจ ืืช ื”ื›ืกืฃ ื•ืžื™ื“ ืกื™ืคืจ ืœื™ ืืช ื›ืœ ื”ืกื™ืคื•ืจ.
ื”ึทืกึทึผืคึธึผืจ ืžึตื”ึทืฉึฐึผืื›ื•ึผื ึธื” ืกึธืคึทืจ ืึถืช ื”ึทื›ึถึผืกึถืฃ ื•ึผืžึดื™ึธึผื“ ืกึดื™ืคึตึผืจ ืœึดื™ ืึถืช ื›ื‡ึผืœ ื”ึทืกึดึผื™ืคึผื•ึผืจ.
The barber from the neighbourhood counted the money and right away told me the whole story.
homographs: ืกืคืจ as sapar, safar
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.19
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s13
ื”ื™ืœื“ ื”ืœืš ืœื’ืŸ ื‘ื‘ื•ืงืจ ืขื ื”ื›ืœื‘.
ื”ึทื™ึถึผืœึถื“ ื”ึธืœึทืšึฐ ืœึทื’ึทึผืŸ ื‘ึทึผื‘ึผื•ึนืงึถืจ ืขึดื ื”ึทื›ึถึผืœึถื‘.
The boy went to kindergarten in the morning with the dog.
stress: segolates yeled, boker, kelev
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.1
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s14
ืื ื™ ืจื•ืฆื” ืœืื›ื•ืœ ืœื—ื ืขื ื—ืžืื” ื•ืฉืžืŸ ื–ื™ืช.
ืึฒื ึดื™ ืจื•ึนืฆึธื” ืœึถืึฑื›ื•ึนืœ ืœึถื—ึถื ืขึดื ื—ึถืžึฐืึธื” ื•ึฐืฉึถืืžึถืŸ ื–ึทื™ึดืช.
I want to eat bread with butter and olive oil.
stress: segolates lechem, shemen
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.11
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s15
ื”ืคื’ื™ืฉื” ื ืงื‘ืขื” ืœื™ื•ื ืฉืœื™ืฉื™ ื‘ืฉืขื” ืฉืชื™ื™ื ื•ื—ืฆื™.
ื”ึทืคึฐึผื’ึดื™ืฉึธืื” ื ึดืงึฐื‘ึฐึผืขึธื” ืœึฐื™ื•ึนื ืฉึฐืืœึดื™ืฉึดืื™ ื‘ึฐึผืฉึธืืขึธื” ืฉึฐืืชึทึผื™ึดื™ื ื•ึธื—ึตืฆึดื™.
The meeting was set for Tuesday at half past two.
numbers and times in words
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.16
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s16
ืฉืžื™ ื“ื ื™ืืœ ืจื•ื–ื”ื™ืœ ื•ืื ื™ ื’ืจ ื‘ื™ืจื•ืฉืœื™ื.
ืฉึฐืืžึดื™ ื“ึธึผื ึดื™ึตึผืึตืœ ืจื•ึนื–ึฐื”ึดื™ืœ ื•ึทืึฒื ึดื™ ื’ึธึผืจ ื‘ึดึผื™ืจื•ึผืฉึธืืœึทื™ึดื.
My name is Daniel Rosehill and I live in Jerusalem.
names: a foreign surname in Hebrew script
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.12
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s17
ืื ื™ ืžืืžืŸ ืžื•ื“ืœ ื–ื™ื”ื•ื™ ื“ื™ื‘ื•ืจ ืขืœ ื”ืงืœื˜ื•ืช ืฉืœ ืขืฆืžื™.
ืึฒื ึดื™ ืžึฐืึทืžึตึผืŸ ืžื•ึนื“ึถืœ ื–ึดื™ื”ื•ึผื™ ื“ึดึผื™ื‘ึผื•ึผืจ ืขึทืœ ื”ึทืงึฐืœึธื˜ื•ึนืช ืฉึถืืœ ืขึทืฆึฐืžึดื™.
I'm training a speech recognition model on my own recordings.
register: technical talk
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.13
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s18
ืชืฉื™ื ืืช ื”ืืจื’ื– ืขืœ ื”ืžืฉื˜ื— ื•ืชืขื‘ื™ืจ ืื•ืชื• ืขื ื”ืžืœื’ื–ื” ืœืžื—ืกืŸ.
ืชึธึผืฉึดื‚ื™ื ืึถืช ื”ึธืึทืจึฐื’ึธึผื– ืขึทืœ ื”ึทืžึดึผืฉึฐืื˜ึธื— ื•ึฐืชึทืขึฒื‘ึดื™ืจ ืื•ึนืชื•ึน ืขึดื ื”ึทืžึทึผืœึฐื’ึตึผื–ึธื” ืœึทืžึทึผื—ึฐืกึธืŸ.
Put the crate on the pallet and take it to the warehouse with the forklift.
domain: material handling
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.16
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s19
ืฉืœื—ืชื™ ืœืš ืžื™ื™ืœ ืขื ื”ืœื™ื ืง ืœืงื•ื‘ืฅ ื‘ื“ืจื™ื™ื‘.
ืฉึธืืœึทื—ึฐืชึดึผื™ ืœึฐืšึธ ืžึตื™ื™ืœ ืขึดื ื”ึทืœึดึผื™ื ึฐืง ืœึทืงึผื•ึนื‘ึถืฅ ื‘ึทึผื“ึฐึผืจึทื™ื™ื‘.
I sent you an email with the link to the file on the Drive.
loanwords: English words in Hebrew
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.12
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
s20
ืื ืœื ื™ื™ืจื“ ื’ืฉื ืžื—ืจ, ื ื™ืกืข ืœื™ื, ื ืื›ืœ ืืจื•ื—ืช ืฆื”ืจื™ื™ื ื•ื ื—ื–ื•ืจ ืœืคื ื™ ื”ืฉืงื™ืขื”.
ืึดื ืœึนื ื™ึตื™ืจึตื“ ื’ึถึผืฉึถืื ืžึธื—ึธืจ, ื ึดื™ืกึทึผืข ืœึทื™ึธึผื, ื ึนืื›ึทืœ ืึฒืจื•ึผื—ึทืช ืฆื‡ื”ึณืจึทื™ึดื™ื ื•ึฐื ึทื—ึฐื–ื•ึนืจ ืœึดืคึฐื ึตื™ ื”ึทืฉึฐึผืืงึดื™ืขึธื”.
If it doesn't rain tomorrow, we'll go to the beach, have lunch and come back before sunset.
prosody: a long sentence with commas
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.22
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa
piper-he-noa
main
plain
p01
ืืชืžื•ืœ ื‘ื‘ื•ืงืจ ื™ืฆืืชื™ ืžื•ืงื“ื ืžื”ื‘ื™ืช, ื›ื™ ื”ื™ื™ืชื” ืœื™ ืคื’ื™ืฉื” ื—ืฉื•ื‘ื” ื‘ืชืœ ืื‘ื™ื‘. ื”ืจื›ื‘ืช ื”ืชืขื›ื‘ื” ื›ืžืขื˜ ื—ืฆื™ ืฉืขื”, ืื– ื”ืชืงืฉืจืชื™ ืœืžืฉืจื“ ื•ื”ื•ื“ืขืชื™ ืฉืืื—ืจ. ื›ืฉืกื•ืฃ ืกื•ืฃ ื”ื’ืขืชื™, ื›ื•ืœื ื›ื‘ืจ ื™ืฉื‘ื• ืกื‘ื™ื‘ ื”ืฉื•ืœื—ืŸ ื•ื—ื™ื›ื• ืœื™. ืœืžื–ืœื™, ื”ืคื’ื™ืฉื” ื”ืœื›ื” ื˜ื•ื‘, ื•ื‘ื“ืจืš ื”ื‘ื™ืชื” ืงื ื™ืชื™ ืœืขืฆืžื™ ืงืคื” ื•ืžืืคื”.
ืึถืชึฐืžื•ึนืœ ื‘ึทึผื‘ึผื•ึนืงึถืจ ื™ึธืฆึธืืชึดื™ ืžื•ึผืงึฐื“ึธึผื ืžึตื”ึทื‘ึทึผื™ึดืช, ื›ึดึผื™ ื”ึธื™ึฐื™ืชึธื” ืœึดื™ ืคึฐึผื’ึดื™ืฉึธืื” ื—ึฒืฉืื•ึผื‘ึธื” ื‘ึฐึผืชึตืœ ืึธื‘ึดื™ื‘. ื”ึธืจึทื›ึถึผื‘ึถืช ื”ึดืชึฐืขึทื›ึฐึผื‘ึธื” ื›ึดึผืžึฐืขึทื˜ ื—ึฒืฆึดื™ ืฉึธืืขึธื”, ืึธื– ื”ึดืชึฐืงึทืฉึทึผืืจึฐืชึดึผื™ ืœึทืžึดึผืฉึฐื‚ืจึธื“ ื•ึฐื”ื•ึนื“ึทืขึฐืชึดึผื™ ืฉึถืืึฒืึทื—ึตืจ. ื›ึฐึผืฉึถืืกึผื•ึนืฃ ืกื•ึนืฃ ื”ึดื’ึทึผืขึฐืชึดึผื™, ื›ึผื•ึผืœึธึผื ื›ึฐึผื‘ึธืจ ื™ึธืฉึฐืื‘ื•ึผ ืกึฐื‘ึดื™ื‘ ื”ึทืฉึผืื•ึผืœึฐื—ึธืŸ ื•ึฐื—ึดื™ื›ึผื•ึผ ืœึดื™. ืœึฐืžึทื–ึธึผืœึดื™, ื”ึทืคึฐึผื’ึด...
Yesterday morning I left home early, because I had an important meeting in Tel Aviv. The train was delayed almost half an hour, so I called the office and said I would be late. When I finally arrived, everyone was already sitting around the table waiting for me. Luckily the meeting went well, and on the way home I boug...
round 2: a four-sentence paragraph for pacing, pauses and prosody across sentence boundaries; first person kept gender-neutral
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
1.06
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Plain Hebrew -> renikud-onnx 0.1.1 G2P (stressed IPA, as the card says the voice was trained on) -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio with the voice config's default inference settings. piper-tts 1.8.0 phoneme_type hebrew does not use renikud (Nakdimon + own G2P); that native path was the first run...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s01
ืื ื™ ื”ื™ื™ืชื™ ืขื•ืฉื” ืืช ื–ื” ืฉื•ื‘.
ืึฒื ึดื™ ื”ึธื™ึดื™ืชึดื™ ืขื•ึนืฉึถื‚ื” ืึถืช ื–ึถื” ืฉืื•ึผื‘.
I would do it again.
Tatoeba grade 1 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.08
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s02
ื–ื” ืœื ืงืฉื” ืื ืืชื” ืœื•ืžื“.
ื–ึถื” ืœึนื ืงึธืฉึถืื” ืึดื ืึทืชึธึผื” ืœื•ึนืžึตื“.
It's not difficult if you study.
Tatoeba grade 1 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.06
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s03
ื”ืžืชื ื” ื”ื–ืืช ื”ื™ื ืœื—ื‘ืจ ืฉืœื™.
ื”ึทืžึทึผืชึธึผื ึธื” ื”ึทื–ึนึผืืช ื”ึดื™ื ืœึถื—ึธื‘ึตืจ ืฉึถืืœึดึผื™.
This gift is for my friend.
Tatoeba grade 2 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.07
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s04
ื”ื–ื”ืจ ืคืŸ ืชืฉื‘ื•ืจ ืืช ื”ื‘ื™ืฆื™ื.
ื”ึดื–ึธึผื”ึตืจ ืคึถึผืŸ ืชึดึผืฉึฐืื‘ึผื•ึนืจ ืึถืช ื”ึทื‘ึตึผื™ืฆึดื™ื.
Take care not to break the eggs.
Tatoeba grade 2 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.07
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s05
ื‘ืจื’ืข ืฉืชื”ื™ื” ืœื™ ื”ื–ื“ืžื ื•ืช, ืื‘ื•ื ืœื‘ืงืจ.
ื‘ึธึผืจึถื’ึทืข ืฉึถืืชึดึผื”ึฐื™ึถื” ืœึดื™ ื”ึดื–ึฐื“ึทึผืžึฐึผื ื•ึผืช, ืึธื‘ื•ึนื ืœึฐื‘ึทืงึตึผืจ.
As soon as I can get the chance, I'll come for a visit.
Tatoeba grade 3 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.09
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s06
ืงื™ื‘ืœื ื• ื”ื•ืจืื•ืช ืื™ืš ืœื”ื›ื™ืŸ ืกืœ ืžื‘ืžื‘ื•ืง.
ืงึดื™ื‘ึทึผืœึฐื ื•ึผ ื”ื•ึนืจึธืื•ึนืช ืึตื™ืšึฐ ืœึฐื”ึธื›ึดื™ืŸ ืกึทืœ ืžึดื‘ึทึผืžึฐื‘ึผื•ึผืง.
We received instructions on how to make a bamboo basket.
Tatoeba grade 3 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.1
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s07
ืขื•ืœืžื ื• ื”ื•ื ืจืง ื—ืœืง ืงื˜ืŸ ืžื”ื™ืงื•ื.
ืขื•ึนืœึธืžึตื ื•ึผ ื”ื•ึผื ืจึทืง ื—ึตืœึถืง ืงึธื˜ึธืŸ ืžึตื”ึทื™ึฐึผืงื•ึผื.
Our world is only one small part of the universe.
Tatoeba grade 4 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.08
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s08
ื”ืจ ืคื•ื’'ื™ ื”ื•ื ื”ื”ืจ ื”ื’ื‘ื•ื” ื‘ื™ื•ืชืจ ื‘ื™ืคืŸ.
ื”ึทืจ ืคึผื•ึผื’ึด'ื™ ื”ื•ึผื ื”ึธื”ึธืจ ื”ึทื’ึธึผื‘ื•ึนื”ึทึผ ื‘ึฐึผื™ื•ึนืชึตืจ ื‘ึฐึผื™ึทืคึธึผืŸ.
Mt. Fuji is the highest mountain in Japan.
Tatoeba grade 4 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.1
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s09
ื’ื•ื’ืœ ืฉื™ื—ืงื” ืชืคืงื™ื“ ื—ืฉื•ื‘ ื‘ื”ื ื’ืฉืช ื”ืžื™ื“ืข ื‘ืื™ื ื˜ืจื ื˜.
ื’ึผื•ึผื’ึถืœ ืฉึดื‚ื™ื—ึฒืงึธื” ืชึทึผืคึฐืงึดื™ื“ ื—ึธืฉืื•ึผื‘ ื‘ึฐึผื”ึทื ึฐื’ึธึผืฉึทืืช ื”ึทืžึตึผื™ื“ึธืข ื‘ึธึผืึดื™ื ึฐื˜ึถืจึฐื ึตื˜.
Google has played an important role in making information more accessible over the internet.
Tatoeba grade 5 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.12
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s10
ืงืฉื” ืœื”ืชื ืขืจ ืžื”ืจื’ืœ ืœืื—ืจ ืฉื”ื•ื ืžืชืงื‘ืข.
ืงึธืฉึถืื” ืœึฐื”ึดืชึฐื ึทืขึตืจ ืžึตื”ึถืจึฐื’ึตึผืœ ืœึฐืึทื—ึทืจ ืฉึถืื”ื•ึผื ืžึดืชึฐืงึทื‘ึตึผืขึท.
A habit is very difficult to shake off once it is formed.
Tatoeba grade 5 sentence
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.09
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s11
ืื ื™ ืงื•ืจื ืกืคืจ ื‘ืขืจื‘ ืœื™ื“ ื”ื—ืœื•ืŸ.
ืึฒื ึดื™ ืงื•ึนืจึตื ืกึตืคึถืจ ื‘ึธึผืขึถืจึถื‘ ืœึฐื™ึทื“ ื”ึทื—ึทืœึผื•ึนืŸ.
I read a book in the evening by the window.
stress: segolates sefer, erev
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.08
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s12
ื”ืกืคืจ ืžื”ืฉื›ื•ื ื” ืกืคืจ ืืช ื”ื›ืกืฃ ื•ืžื™ื“ ืกื™ืคืจ ืœื™ ืืช ื›ืœ ื”ืกื™ืคื•ืจ.
ื”ึทืกึทึผืคึธึผืจ ืžึตื”ึทืฉึฐึผืื›ื•ึผื ึธื” ืกึธืคึทืจ ืึถืช ื”ึทื›ึถึผืกึถืฃ ื•ึผืžึดื™ึธึผื“ ืกึดื™ืคึตึผืจ ืœึดื™ ืึถืช ื›ื‡ึผืœ ื”ึทืกึดึผื™ืคึผื•ึผืจ.
The barber from the neighbourhood counted the money and right away told me the whole story.
homographs: ืกืคืจ as sapar, safar
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.15
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s13
ื”ื™ืœื“ ื”ืœืš ืœื’ืŸ ื‘ื‘ื•ืงืจ ืขื ื”ื›ืœื‘.
ื”ึทื™ึถึผืœึถื“ ื”ึธืœึทืšึฐ ืœึทื’ึทึผืŸ ื‘ึทึผื‘ึผื•ึนืงึถืจ ืขึดื ื”ึทื›ึถึผืœึถื‘.
The boy went to kindergarten in the morning with the dog.
stress: segolates yeled, boker, kelev
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.08
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s14
ืื ื™ ืจื•ืฆื” ืœืื›ื•ืœ ืœื—ื ืขื ื—ืžืื” ื•ืฉืžืŸ ื–ื™ืช.
ืึฒื ึดื™ ืจื•ึนืฆึธื” ืœึถืึฑื›ื•ึนืœ ืœึถื—ึถื ืขึดื ื—ึถืžึฐืึธื” ื•ึฐืฉึถืืžึถืŸ ื–ึทื™ึดืช.
I want to eat bread with butter and olive oil.
stress: segolates lechem, shemen
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.09
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s15
ื”ืคื’ื™ืฉื” ื ืงื‘ืขื” ืœื™ื•ื ืฉืœื™ืฉื™ ื‘ืฉืขื” ืฉืชื™ื™ื ื•ื—ืฆื™.
ื”ึทืคึฐึผื’ึดื™ืฉึธืื” ื ึดืงึฐื‘ึฐึผืขึธื” ืœึฐื™ื•ึนื ืฉึฐืืœึดื™ืฉึดืื™ ื‘ึฐึผืฉึธืืขึธื” ืฉึฐืืชึทึผื™ึดื™ื ื•ึธื—ึตืฆึดื™.
The meeting was set for Tuesday at half past two.
numbers and times in words
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.1
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s16
ืฉืžื™ ื“ื ื™ืืœ ืจื•ื–ื”ื™ืœ ื•ืื ื™ ื’ืจ ื‘ื™ืจื•ืฉืœื™ื.
ืฉึฐืืžึดื™ ื“ึธึผื ึดื™ึตึผืึตืœ ืจื•ึนื–ึฐื”ึดื™ืœ ื•ึทืึฒื ึดื™ ื’ึธึผืจ ื‘ึดึผื™ืจื•ึผืฉึธืืœึทื™ึดื.
My name is Daniel Rosehill and I live in Jerusalem.
names: a foreign surname in Hebrew script
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.1
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s17
ืื ื™ ืžืืžืŸ ืžื•ื“ืœ ื–ื™ื”ื•ื™ ื“ื™ื‘ื•ืจ ืขืœ ื”ืงืœื˜ื•ืช ืฉืœ ืขืฆืžื™.
ืึฒื ึดื™ ืžึฐืึทืžึตึผืŸ ืžื•ึนื“ึถืœ ื–ึดื™ื”ื•ึผื™ ื“ึดึผื™ื‘ึผื•ึผืจ ืขึทืœ ื”ึทืงึฐืœึธื˜ื•ึนืช ืฉึถืืœ ืขึทืฆึฐืžึดื™.
I'm training a speech recognition model on my own recordings.
register: technical talk
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.11
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s18
ืชืฉื™ื ืืช ื”ืืจื’ื– ืขืœ ื”ืžืฉื˜ื— ื•ืชืขื‘ื™ืจ ืื•ืชื• ืขื ื”ืžืœื’ื–ื” ืœืžื—ืกืŸ.
ืชึธึผืฉึดื‚ื™ื ืึถืช ื”ึธืึทืจึฐื’ึธึผื– ืขึทืœ ื”ึทืžึดึผืฉึฐืื˜ึธื— ื•ึฐืชึทืขึฒื‘ึดื™ืจ ืื•ึนืชื•ึน ืขึดื ื”ึทืžึทึผืœึฐื’ึตึผื–ึธื” ืœึทืžึทึผื—ึฐืกึธืŸ.
Put the crate on the pallet and take it to the warehouse with the forklift.
domain: material handling
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.12
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
piper-he-noa-nikud
piper-he-noa
with our nikud
nikud
s19
ืฉืœื—ืชื™ ืœืš ืžื™ื™ืœ ืขื ื”ืœื™ื ืง ืœืงื•ื‘ืฅ ื‘ื“ืจื™ื™ื‘.
ืฉึธืืœึทื—ึฐืชึดึผื™ ืœึฐืšึธ ืžึตื™ื™ืœ ืขึดื ื”ึทืœึดึผื™ื ึฐืง ืœึทืงึผื•ึนื‘ึถืฅ ื‘ึทึผื“ึฐึผืจึทื™ื™ื‘.
I sent you an email with the link to the file on the Drive.
loanwords: English words in Hebrew
thewh1teagle/piper-he-noa
noa (speaker 0, default; he_IL-noa-medium.onnx)
22,050
none (cpu=4)
0.09
thewh1teagle
individual
2026-10-26
fine-tuned for Hebrew from en_US-ljspeech-medium (which the card says was trained from scratch on LJSpeech). NOT Hebrew-only from scratch; Hebrew voice is trained on synthetic audio
CC BY 4.0
Input nikud: Dicta text_nikud (vowels kept) + hataama U+05AB and vocal shva U+05BD transplanted from Phonikud/phonikud-onnx diacritizer run on plain text (same as phonikud-piper) -> phonikud 0.4.1 phonemize -> piper-tts 1.8.0 phonemes_to_ids / phoneme_ids_to_audio on the noa voice (speaker 0, config defaults). Bypasses...
End of preview. Expand in Data Studio

Hebrew TTS Bake-off 11/26: generated clips

340 clips: 10 open Hebrew TTS models (17 runs) reading the 20 sentences of Hebrish/Hebrew-TTS-Eval-Set, generated 2026-11-18. Part of Hebrish. Listen side by side in the Space Hebrish/Hebrew-TTS-Bakeoff-1126; method, code and provenance: GitHub.

Where a model can take both, it was run on plain Hebrew (it infers the vowels) and on the checked nikud (input nikud, variant "with our nikud"): the pair isolates its nikud inference.

All audio is synthetic. Each clip is under its model's licence (column licence); several are non-commercial.

Column Meaning
run Run slug (<model>, <model>-nikud, <model>-auto, <model>-plain)
model, variant, input Base model, variant label, plain or nikud
sentence_id, text, text_nikud, english, focus The sentence
audio The clip at the model's native sample rate
hf_id, voice, sample_rate, hardware, seconds_to_generate Run settings
publisher, publisher_type, release_date, training_origin, licence Model provenance
run_notes Notes from the run
Downloads last month
72