PocketTTS 100M โ€” LiteRT

Self-contained .litertlm packages of Kyutai PocketTTS (100M-parameter continuous FlowLM + Mimi neural audio codec, 24 kHz mono PCM output) quantized to INT8 for on-device inference with LiteRT.

Each .litertlm bundle packages the INT8 LiteRT subgraphs (text conditioner, stateful autoregressive FlowLM backbone, LSD flow-matching network, and stateful Mimi streaming decoder), the SentencePiece tokenizer, and pre-encoded voice conditioning embeddings in a single file.

Available Files

File Size Languages Voices
pocket_tts_en_int8.litertlm 156 MB English (en) alba, marius, javert, jean, fantine, cosette, eponine, azelma
pocket_tts_int8.litertlm 725 MB 7 languages (en, fr, de, es, it, pt, nl) All 8 English voices + narrator_f and narrator_m for each non-English language (14 voices total)
pocket_tts_fr_int8.litertlm 122 MB French (fr) narrator_f, narrator_m
pocket_tts_de_int8.litertlm 121 MB German (de) narrator_f, narrator_m
pocket_tts_es_int8.litertlm 121 MB Spanish (es) narrator_f, narrator_m
pocket_tts_it_int8.litertlm 120 MB Italian (it) narrator_f, narrator_m
pocket_tts_pt_int8.litertlm 121 MB Portuguese (pt) narrator_f, narrator_m
pocket_tts_nl_int8.litertlm 120 MB Dutch (nl) narrator_f, narrator_m

Download

pip install -U huggingface_hub

# English-only bundle (156 MB, 8 voices)
hf download litert-community/pocket-tts pocket_tts_en_int8.litertlm --local-dir .

# Multilingual 7-language bundle (725 MB, 14 voices)
hf download litert-community/pocket-tts pocket_tts_int8.litertlm --local-dir .

License & Attribution

Downloads last month
-
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for litert-community/pocket-tts

Quantized
(1)
this model