valenti/spark-pocket-tts
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
78 Commits 1 Branch 0 Tags
7db278ee57d02302934d9b6be716fe8e1f96aa3b
Commit Graph
78 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Gabriel de Marmiesse 7db278ee57 Add default voice for each language (#180) 
Fix https://github.com/kyutai-labs/pocket-tts/issues/166
 2026-05-04 15:35:22 +02:00
Václav Volhejn de010ea890 Rephrase default German sentence (#173) 
"von Kyutai" is a bit easier for the TTS than "Kyutais" altough it still
struggles. "du wirst mich mögen" sounded unnatural to me
 2026-04-30 14:45:06 +02:00
Lee Felarca d529606ced Add tts-audiobook-tool project to README (#169) 2026-04-27 10:09:58 +02:00
Gabriel de Marmiesse 13542b4f9d Fix support for devices that are not cpu (#165) 
Fixes https://github.com/kyutai-labs/pocket-tts/issues/164
 2026-04-22 19:18:48 +02:00
Gabriel de Marmiesse 5304da453b Add the demo (#163) 2026-04-20 15:55:33 +02:00
gabrieldemarmiesse c9ebef4d1a Bump version to 2.0.0 2026-04-17 17:56:23 +02:00
Gabriel de Marmiesse c90fc8ced6 Add french, italian, portuguese, spanish, german (#155) 
The french 6 layers variant wasn't included because it's not good enough
yet. Use the `french_24l` variant for now.
 2026-04-17 17:52:16 +02:00
Pieter Venter 9fc89c64f2 Add OpenClaw integrations to README (#160) 2026-04-17 11:57:40 +02:00
Ajaykrishnan R 119ca2e618 Update README to add pocket-tts-csharp to alterative implementations (#159) 2026-04-13 12:24:19 +02:00
markd89 05bb41011f Add FrontPocket to Projects using Pocket-TTS (#157) 2026-04-07 10:10:07 +02:00
John Costa ef69ab8652 Split long sentences on commas to prevent skipped words (#143) 
## Summary

When a single sentence exceeds `max_tokens` (50) and has no
sentence-ending punctuation (`.`, `!`, `?`), the text splitting logic
now falls back to splitting on commas, semicolons, and colons. This
prevents the model from silently skipping parts of long sentences.

Fixes #38
 2026-03-24 11:50:44 +01:00
nabil-tazi 670e777260 Add int8 dynamic quantization support (#147) 
# Add int8 dynamic quantization
 
Closes #7
 
## Summary
 
This PR implements mixed-precision int8 dynamic quantization for
pocket-tts, as requested in issue #7. It adds a `quantize=True` flag to
`TTSModel.load_model()` that reduces runtime memory usage by ~48% and
improves inference speed by ~27% on x86 and up to 23% on ARM, with no
measurable impact on audio quality.
 2026-03-24 11:29:14 +01:00
Fangjun Kuang 0dcf6cb442 Update README to add sherpa-onnx to alternative implementations (#150) 2026-03-24 10:59:17 +01:00
Gabriel de Marmiesse 25e1b673cd Added code to generate the default voices (#144) 2026-03-12 18:12:47 +01:00
VolgaGerm 5df2f4c383 Add PocketTTS.cpp to alternative implementations (#142) 
Adds PocketTTS.cpp to the alternative implementations list — a C++ ONNX
Runtime port with CLI, streaming HTTP server, and a C API for FFI.
 2026-03-12 11:04:21 +01:00
tonelord e5b0555382 Add pocket-tts-deno to README.md (#140) 2026-03-04 10:24:17 +01:00
Thiago Duarte aca5c641fb Add cursed codex to projects using pocket tts (#139) 2026-03-03 10:19:58 +01:00
Alvin Mei 270cc658eb Add discord-tts project link to README (#138) 2026-03-01 21:18:49 +01:00
joshwhiton b942bc423b Raise minimum huggingface_hub to 0.13.0 for consistent offline behavior (#137) 2026-02-27 10:54:08 +01:00
A.I Joe cc1e654931 Update README with new project links (#136) 
Added new resources for ComfyUI-Pocket-TTS and pocket-tts-server.
 2026-02-21 18:30:30 +01:00
darknight054 33a93718f6 Changed the implementation to fuse the transformers (#85) 2026-02-20 14:45:34 +01:00
Gabriel de Marmiesse 09d81f7ad0 Add Laurent's implementation of Pocket TTS 2026-02-17 14:18:58 +01:00
Gabriel de Marmiesse e43b4851d0 Fix export-voice docs (#135) 2026-02-17 11:13:57 +01:00
gabrieldemarmiesse aba1465fe5 Bump version to 1.1.1 2026-02-16 18:06:30 +01:00
Alexandre Défossezandgabrieldemarmiesse a5da339972 making pocket tts compat with cuda generation (#132) 
Co-authored-by: gabrieldemarmiesse <gabrieldemarmiesse@gmail.com>
 2026-02-16 18:05:59 +01:00
Gabriel de Marmiesse 2f6ac5605c Links now point to the docs instead of github (#131) 2026-02-16 16:40:44 +01:00
Gabriel de Marmiesse 9e058f2f44 Fix docs generation in ci (#130) 2026-02-16 16:35:21 +01:00
gabrieldemarmiesse 3f7dcf4de9 Bump version to 1.1.0 2026-02-16 15:46:38 +01:00
thataboy 9878cd0fd2 Add latencyHint when creating AudioContext (test frontend) (#124) 2026-02-16 15:42:00 +01:00
Gabriel de Marmiesse 8443de8e06 Add functions to export and import the model state for a given voice (#129) 2026-02-16 15:37:00 +01:00
Gabriel de Marmiesse aca7dc8db6 Remove cudagraph reference (#123) 2026-02-12 16:11:13 +01:00
Gabriel de Marmiesse 5d8b15c75e Use a symlink for the index page in the docs (#122) 2026-02-12 16:03:28 +01:00
Sachin Sankar 4ae9fb2590 Automated Docs Site (#101) 2026-02-12 14:01:42 +01:00
Gabriel de Marmiesse ef56ab4972 Rework related projects sections 2026-02-12 12:15:49 +01:00
Emir Belkahia a8c735a9a4 Add basic public API tests to test_python_api.py (#114) 2026-02-12 11:35:29 +01:00
Mukul bisht 431878c0db Adding possibility to read the input text from stdin in the CLI (#110) 2026-02-12 11:32:42 +01:00
Gabriel de Marmiesse 70d47c88f8 Add note about other languages 2026-02-10 10:52:37 +01:00
Gabriel de Marmiesse d773bf9c98 Add "alternative implementations" (#100) 2026-02-03 12:43:13 +01:00
darknight054 a47c7e4ba9 Reduce KV cache memory (#51) 
Hi! This PR tackles the memory issues discussed in [Reduce Memory
Usage](https://github.com/kyutai-labs/pocket-tts/issues/14) by removing
the "always allocate 1000 steps" KV-cache behavior.
 2026-02-03 12:32:46 +01:00
Simon HopstätterandCopilot a1341e8b6b Add link to OpenAI-compatible streaming server (#96) 
Added a new project link for OpenAI-compatible streaming server.

---------

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
 2026-02-03 10:22:05 +01:00
Jishnu Venugopalandjishnuvenugopal cb555a5fbc Add pocket-tts-mlx to projects list (#97) 
Adds the MLX backend fork to Projects using Pocket TTS.

Co-authored-by: jishnuvenugopal <jishnuvenugopal@users.noreply.github.com>
 2026-02-03 10:21:14 +01:00
Prabhudayal Vaishnav 61d4b1c23b feat: add docker-compose.yaml with model cache persistence (#64) 
Closes #28
 2026-02-02 15:52:11 +01:00
Václav Volhejn 6ee89b12ff Mention slaughters85j/pocket-tts repo 2026-01-30 11:32:21 +01:00
Václav Volhejn 28b8244428 Improve consistency in variant -> config rename (#87) 
Small follow-up on #59
 2026-01-27 08:55:34 +00:00
teddybear082 2b51649a65 Basic changes to allow loading of local model files (#59) 
-modify main,py to not hardcode using DEFAULT_VARIANT

-allow submission of path string to path of variant.yaml file saved in
custom directory in lieu of model variant identifier in CLI to use when
loading model

-addresses #57
 2026-01-27 09:42:35 +01:00
Václav Volhejn 77cc74a443 Clarifications regarding voices + fix missing @classmethod (#86) 2026-01-26 12:46:44 +01:00
thataboyandVáclav Volhejn edec7f8790 Add .safetensors save/load to get_state_for_audio_prompt (#69) 
Quality of life improvement to save user voice embedding as a
.safetensors file that can be loaded for subsequent use, achieving the
same efficiency as the predefined voices, all via the existing
get_state_for_audio_prompt function.

---------

Co-authored-by: Václav Volhejn <8401624+vvolhejn@users.noreply.github.com>
 2026-01-26 10:36:27 +01:00
Václav Volhejn bfbaf80437 Bump version to 1.0.3 2026-01-22 12:02:14 +01:00
Václav Volhejn a3aaa135bd Add links to more community implementations for browsers (#77) 2026-01-22 11:30:36 +01:00
Saleem DjimaandVáclav Volhejn 4ec89acf95 Feat/add chunk size control arg (#56) 
This pull request addresses issue #38 

**Changes included:**

- Added a `max_tokens` parameter to the typer cli for the `generate`
command, allowing user to set a default number of tokens per chunk.
- Introduced a default variable `MAX_TOKEN_PER_CHUNK` to centralize the
default value.
- Fixed a tokenization issue by inserting a comma, allowing the model to
correctly read each chunk as intended rather than treating one chunk as
a single sentence.

**Future improvement suggestion :**

- Consider enhancing the `split_into_best_sentences` function using
regex to split text more accurately without modifying it, which would
improve clarity and maintainability.

---------

Co-authored-by: Václav Volhejn <8401624+vvolhejn@users.noreply.github.com>
 2026-01-21 11:54:53 +01:00