EDChronicle-models

Backup copies of the third-party model files and voice packages that EDChronicle downloads at install time. They are kept here so a fresh EDChronicle install still works if an original download disappears.

The files are not stored in this repository's code. Each version is a GitHub release with the files attached, and older releases are never deleted, so every EDChronicle version can always fetch the exact files it was tested with.

Releases

Release	Files	Used for	Original source
kokoro-v1.0	kokoro-v1.0.int8.onnx, voices-v1.0.bin	Kokoro offline text-to-speech voices	thewh1teagle/kokoro-onnx (model by hexgrad/Kokoro-82M)
vosk-en-us-0.22-lgraph	vosk-model-en-us-0.22-lgraph.zip	Vosk offline speech recognition (voice commands)	alphacephei.com/vosk/models
packages-2026-10	Wheels (Windows x64, Python 3.12) and source archives for kokoro-onnx 0.6.1, vosk 0.3.45, espeakng-loader 0.2.4, phonemizer 3.4.0, plus espeak-ng 1.52.0 source; SHA256SUMS	Python packages, used by EDChronicle's install.bat (pip --find-links) if PyPI no longer has them	PyPI; source from PyPI sdists or the matching GitHub tag/commit

SHA-256

EDChronicle checks every model download against these values, from either source, and rejects a file that doesn't match. Package hashes are in the SHA256SUMS file of the packages-* release.

6e742170d309016e5891a994e1ce1559c702a2ccd0075e67ef7157974f6406cb  kokoro-v1.0.int8.onnx
bca610b8308e8d99f32e6fe4197e7ec01679264efed0cac9140fe9c29f1fbf7d  voices-v1.0.bin
d9838b4aaa82a75c4a17f5aca300eaca129aaab2a7cbf951bafbb500eb9c4334  vosk-model-en-us-0.22-lgraph.zip

Adding a new model version

1. Test the new files in EDChronicle.
2. Publish them as a new release here, for example kokoro-v1.1. Don't replace files in an existing release.
3. Update the URLs and SHA-256 values in EDChronicle's download_models.py and edc/audio/voice_commands.py.

For new package versions, publish a new packages-* release with the wheels, source archives and SHA256SUMS, then update EDC_PACKAGES in EDChronicle's install.bat and the pins in requirements.txt.

License

The model files are redistributed unmodified under the Apache License 2.0 (see LICENSE). The packages keep their own licenses (MIT, Apache-2.0, GPL-3.0); the GPL-3.0 parts (phonemizer, and the espeak-ng build inside the espeakng-loader wheel) are accompanied by their complete source in the same release. Copyright remains with the original authors. See NOTICE for attribution.