6999bcdb57
Close TCP connection after synthesis so HA receives FIN and unblocks
...
Build and Push Docker Image / build (push) Successful in 2m10s
disconnect() was a no-op in the base class; writer.close() was never
called, leaving HA waiting for a TCP FIN that never arrived.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 19:58:54 -04:00
1cda188c98
Close connection after synthesis so HA knows response is complete
...
Build and Push Docker Image / build (push) Successful in 2m12s
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 19:51:57 -04:00
83aa2ec08c
Split text into sentences to stream audio chunk-by-chunk
...
Build and Push Docker Image / build (push) Successful in 2m15s
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 19:41:19 -04:00
067a853d3b
Add debug logging to synthesis pipeline to trace hang
...
Build and Push Docker Image / build (push) Successful in 2m11s
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 19:34:18 -04:00
8272b6a8c9
Switch hf_cache from Docker volume to host bind mount
...
Build and Push Docker Image / build (push) Successful in 19s
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 19:14:44 -04:00
f2513c12a9
Convert audio tensor to numpy before PCM conversion
...
Build and Push Docker Image / build (push) Successful in 2m11s
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 19:11:30 -04:00
9e907b7573
Enable HSA_OVERRIDE_GFX_VERSION for RX 6700 XT (gfx1031) ROCm compatibility
...
Build and Push Docker Image / build (push) Successful in 19s
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 19:10:03 -04:00
3340e2e8a5
Fix wyoming import: AsyncEventHandler is in wyoming.server not wyoming.handler
...
Build and Push Docker Image / build (push) Successful in 2m9s
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 19:05:33 -04:00
3f279e53ee
Pull image from Gitea registry instead of building locally
...
Build and Push Docker Image / build (push) Has been cancelled
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 19:03:19 -04:00
985f30e3ae
Mount config.yaml into container for live editing
...
Build and Push Docker Image / build (push) Successful in 19s
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 19:01:56 -04:00
e8a3844994
Change Wyoming port from 10200 to 10300
...
Build and Push Docker Image / build (push) Successful in 2m20s
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 18:57:02 -04:00
ad58061b6f
Update .gitea/workflows/docker-build.yml
Build and Push Docker Image / build (push) Failing after 12m22s
2026-04-08 17:54:20 -04:00
0614418dd4
Update README.md
Build and Push Docker Image / build (push) Failing after 11m10s
2026-04-08 17:39:04 -04:00
38c1e5e096
Fix registry secret name in build workflow
...
Build and Push Docker Image / build (push) Failing after 3h58m51s
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 13:32:45 -04:00
5e8e7ad6d4
Replace upstream library with ROCm/Wyoming deployment project
...
Build and Push Docker Image / build (push) Failing after 47s
Remove original Kokoro library source, demo, examples, tests, JS port,
and GitHub config. Add Dockerfile (ROCm 6.1 / PyTorch 2.5.1), Wyoming
TCP server, docker-compose with GPU passthrough, config, entrypoint,
and Gitea Actions build workflow.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com >
2026-04-08 13:30:54 -04:00
Daniel Tang
dfb907a02b
Enable Python 3.13 ( #244 )
...
Tested working with `kokoro -t Hello -o test.wav`. This is needed for
Ubuntu 25.04.
2025-08-06 15:28:53 -07:00
ArEnSc
d4ef0569c7
Bugfix/crash loading voice ( #248 )
...
* add voice data url getter and setter
* Ok Voices.js removing types
* Match the function style.
* Actually use the url for the voice file.
---------
Co-authored-by: CrossPr0duct <alicextensor@gmail.com >
2025-08-06 15:28:29 -07:00
styagi130
f1d129d835
Feat: batch support for onnx and triton compatibility ( #239 )
...
* Added batch support
Signed-off-by: Siddharth Tyagi <siddhartht@nvidia.com >
* file rename
Signed-off-by: Siddharth Tyagi <siddhartht@nvidia.com >
---------
Signed-off-by: Siddharth Tyagi <siddhartht@nvidia.com >
Co-authored-by: Siddharth Tyagi <siddhartht@nvidia.com >
2025-07-25 17:49:52 -07:00
ArEnSc
d7654ba4e8
Feature/get set voice url ( #236 )
...
* add voice data url getter and setter
* Ok Voices.js removing types
* Match the function style.
2025-07-09 14:56:34 -07:00
Kartikeya Mishra
2760831139
feat: set custom cache directory ( #227 )
2025-06-30 13:52:07 -07:00
Eric Izoita
4b647d371e
Fix a couple minor warnings ( #210 )
...
* use non-deprecated import path of weight_norm
* don't pass dropout to single layer LSTM
per the docs, dropout is not applied to the last LSTM layer, so there is
no effect to passing it when `num_layers=1`.
2025-06-30 13:50:50 -07:00
Joshua Lochner
2668b2e279
[kokoro-js] Expose wasmPaths environment variables to users ( #190 )
...
* Bump dependency versions
* Expose wasmPaths from transformers.js
* Update syntax to be bundler friendly
* Bump transformers.js version
* Bump demo versions
* Update vite config
* [version] Update to 1.2.1
2025-05-03 09:31:18 -07:00
fondoger
6d87f4ae7a
Enable MPS GPU Accerlation on MacOS ( #164 )
...
* Enable MPS GPU Accerlation on MacOS
* Fix
2025-04-10 11:40:25 -07:00
hexgrad
1c7bdd971d
Bump to 0.9.4 ( #169 )
2025-04-05 15:00:22 -07:00
Omar Irfan Khan
4f5106e327
adding instructions for setting up espeak-ng on windows ( #143 )
2025-04-01 10:54:49 -07:00
Ash
e44c9b4add
Update README.md ( #154 )
...
Sorted lang_code tickers in alphabetical order above the input. Before they were oddly stretched out over different steps rather than being sorted and directly above the input.
2025-04-01 10:54:14 -07:00
Michael Currin
26039de2dc
docs: Add missing setup step in README.md ( #145 )
...
* docs: Add missing setup step in README.md
* docs: fix README.md
2025-03-25 11:00:52 -07:00
hexgrad
e43d62643e
Remove scipy ( #139 )
...
* Remove scipy
* No longer need to replace T
* Update README.md
* Remove numpy version lock
* Update README.md
* Update uv.lock
2025-03-18 11:16:34 -07:00
hexgrad
3f9dd88d6f
Bump to 0.8.4 ( #120 )
...
* Bump to 0.8.4
* Update README.md
2025-02-28 18:49:08 -08:00
hexgrad
790ecc9c83
Bump to 0.8.3 ( #119 )
2025-02-28 18:05:27 -08:00
szsteven008
c87df60d4c
add onnx export.py ( #112 )
...
* Add files via upload
onnx export
* Add files via upload
KModelForONNX
* Add files via upload
* Delete export.py
* Add files via upload
* Add files via upload
修正中文的错误
* Add files via upload
增加duration的输出
2025-02-28 11:01:34 -08:00
hexgrad
b15ef354b2
Bump to 0.8.2 ( #117 )
2025-02-27 07:22:30 -08:00
hexgrad
ece280bdcd
Bump to 0.8.1
2025-02-26 17:57:41 -08:00
hexgrad
3a721cce9f
Critical fix to 0.8.0
2025-02-26 17:55:43 -08:00
hexgrad
efa91a8a3f
Match misaki==0.8.0 dev branch ( #114 )
...
* Match misaki==0.8.0 dev branch
* en_callable, speed callable
2025-02-26 17:30:50 -08:00
Kirill R.
52f7eb740b
Add Result.text_index to be able to map segments to paragraphs ( #111 )
...
* Add Result.text_index to be able to map segments to paragraphs
* Fix speed re: #105
2025-02-23 08:30:25 -08:00
Alessandro Saccoia
2dd9df6779
Fix: add text chunking for non-English language pipeline ( #105 )
...
Co-authored-by: Your Name <your.email@example.com >
2025-02-19 18:15:21 -08:00
Adrian Lyjak
e648c0605a
Add additional onnx compat for https://github.com/pytorch/pytorch/issues/92977 ( #104 )
2025-02-18 11:06:48 -08:00
etrotta
cd7afb5c12
Add a CLI interface ( #102 )
...
* Add a CLI interface and update packaging configuration
* Support multiple lines in stdin
---------
Co-authored-by: Eric Trotta <eric.oliveira@magva.com.br >
2025-02-17 21:07:21 -08:00
Joshua Lochner
5229a254b7
Kokoro.js v1.2.0: Streaming support ( #92 )
...
* Set up JS project
* Finalise JS library
* Update README
* Fix package.json repository url
* Rename package -> `kokoro-js`
* Fix samples in README
* Cleanup README
* Bump `phonemizer` version
* Create web demo
* Run prettier
* Link to model used in demo
* Enable multithreading in HF space demo (~40% faster)
* Add link to demo in README
* Bump to v1.0.1
* Update voices
* Update versions
* Update phonemize JSDoc
* Use updated voice pack
* Update versions
* Update demo (v1.0 & WebGPU support)
* Update README
* Enforce maximum number of tokens
* Update README
* [version] Update to 1.1.1
* Create simple sentence splitter
* Update `npm run test`
* Update API to use sync and async iterators
* Add support for streamed generation in kokoro.js
* Always split on newlines
* Remove debug line
* Improvements
* Add more matching puntuation marks
* Update comments
* nits
* Export TextSplitterStream too
* Update splitter.js
* Update README
* [version] Update to 1.2.0
2025-02-15 11:06:33 -08:00
Adrian Lyjak
93abff8795
Modify model for ONNX compatibility ( #87 )
2025-02-15 11:05:57 -08:00
hexgrad
ce71a10c57
Bump to 0.7.16 ( #94 )
2025-02-14 22:49:19 -08:00
Thien Tran
84d64f02d3
replace np.prod() with math.prod() to make Kokoro torch.compile-able ( #91 )
...
* replace np.prod() with math.prod()
* another np.prod()
2025-02-14 22:48:51 -08:00
RobViren
330d110c05
Allow pipeline to take a voice style tensor directly. ( #93 )
2025-02-14 22:48:08 -08:00
hexgrad
1145c0b7f6
Bump to 0.7.15 ( #83 )
2025-02-11 22:54:42 -08:00
hexgrad
bd44d79895
Bump to 0.7.14 ( #82 )
2025-02-11 14:08:57 -08:00
CarelessParsley
f77e52fb4c
Port HuggingFace Space to plain Gradio ( #81 )
...
* Port HuggingFace Space to plain Gradio
* Update app.py
Remove BANNER_TEXT
---------
Co-authored-by: Careless Parsley <carelessparsley@gmail.com >
Co-authored-by: hexgrad <166769057+hexgrad@users.noreply.github.com >
2025-02-11 13:04:59 -08:00
hexgrad
83c8883d32
Delete .DS_Store
2025-02-11 00:24:22 -08:00
hexgrad
fd62d70ec7
Bump to 0.7.13 ( #75 )
2025-02-10 08:54:11 -08:00
hexgrad
15108f11ba
Bump to 0.7.12 ( #67 )
2025-02-07 23:09:34 -08:00
