# Third-party material

## The voice

`lt_LT-ingute-medium.onnx` is the Ingutė voice from the Piper voice catalogue
(https://huggingface.co/rhasspy/piper-voices, `lt/lt_LT/ingute/medium`) with sherpa-onnx metadata
added (`model_type`, `comment`, `language`, `voice`, `has_espeak`, `n_speakers`, `sample_rate`);
the network weights are unchanged. `tokens.txt` is generated from the voice's
`phoneme_id_map`: keys longer than one code point are left out (sherpa-onnx cannot read them and
eSpeak NG never produces them), and `ʂ` gets the id of `s`.

Licence: NewGenLTU Open RAIL-D License 1.0, full text in [LICENSE-VOICE](LICENSE-VOICE)
(https://sitti.vdu.lt/newgenltu-openrail-d-license/). Voice repository:
https://huggingface.co/RobertasTa/lt_LT-ingute-medium

## eSpeak NG data (`lt_dict`, `espeak-ng-lt/`)

`lt_dict` is compiled with eSpeak NG's dictionary compiler from the files in
`espeak-ng-lt/dictsource/`:

- `lt_rules`, the first part of `lt_list`, `lt_emoji`: eSpeak NG
  (https://github.com/espeak-ng/espeak-ng), GPL-3.0. `lt_rules` and `lt_list` as in eSpeak NG
  master; `lt_emoji` as in the eSpeak NG fork used by sherpa-onnx
  (https://github.com/csukuangfj/espeak-ng), so that the unchanged files compile to the same
  `lt_dict` that SherpaTTS 3.4 ships (md5 `5426503daf0a29836d0f168d38b38581`).
- The added block at the end of `lt_list` (189,209 lines `word $N`): the stressed syllable of
  each word form, taken from `lt_kirciai.tsv` as shipped in OHF-Voice/piper1-gpl
  (`src/piper/lithuanian/`, GPL-3.0), which is built from two works licensed under the
  Creative Commons Attribution 4.0 International licence:
  - LIEPA corpus annotation, Vilnius University:
    https://huggingface.co/datasets/meldynamics/liepa-tts
  - g2p-lt-lexicon, Arūnas Smaliukas: https://huggingface.co/datasets/svogunas/g2p-lt-lexicon

The resulting `lt_dict` is distributed under GPL-3.0 with these sources.

## The ears (`parakeet-lt-kmynas-int8.zip`, release v1.1.0)

A Lithuanian speech recognition model for Parakeetype, licensed under the Creative Commons
Attribution 4.0 International licence (full text and an attribution notice inside the archive as
`LICENSE.txt` and `NOTICE.txt`):

- "parakeet-tdt-0.6b-v3" by NVIDIA: https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3
- fine-tuned for Lithuanian on the LIEPA-3 corpus as "kmynas-parakeet-lt-v4" by Kristijonas
  Jakubsonas: https://huggingface.co/kristijonas/kmynas-parakeet-lt-v4
- `nemo128.onnx` (the audio front end) from the Parakeet TDT 0.6B v3 ONNX export by istupakov:
  https://huggingface.co/istupakov/parakeet-tdt-0.6b-v3-onnx

Our changes: exported to ONNX (encoder, decoder, joiner), the joint network's input projections
moved into the encoder and decoder graphs with the same weights (the output is identical), tensors
renamed to the layout Parakeetype reads, encoder quantised to int8 (MatMul weights only), the
vocabulary written as `vocab.txt`.

`silero_vad_v6.onnx` in the archive is the Silero VAD v6 model by the Silero Team
(https://github.com/snakers4/silero-vad), unmodified, MIT licence (`LICENSE-silero-vad.txt`).

## Not included

SherpaTTS (https://github.com/woheller69/ttsengine, GPL-3.0) is a separate app and is not part
of this package. Install it from F-Droid.

Parakeetype (https://github.com/theScrabi/Parakeetype, GPL-3.0) is a separate app as well. Our
proposed change to it (registering the Lithuanian model) lives in a public branch:
https://github.com/RobertasTa/Parakeetype/tree/add-lithuanian-model