From ad28c43699dfee6566c08af63b8b7bc544d32c2b Mon Sep 17 00:00:00 2001 From: "Danilo M." Date: Sat, 26 Sep 2026 07:20:43 +0200 Subject: python3-kokoro-onnx: add version 0.6.1 Kokoro-82M text to speech through onnxruntime. The espeakng-loader dependency is dropped: its bundled libespeak-ng has a hardcoded CI data path and fails at runtime. The tokenizer defaults to the system espeak-ng library and data instead, and the build fails if a future release still references espeakng-loader after the edit. Co-Authored-By: Claude Opus 5.5 --- python3-kokoro-onnx/README | 15 +++++++++++++++ 1 file changed, 15 insertions(+) create mode 100644 python3-kokoro-onnx/README (limited to 'python3-kokoro-onnx/README') diff --git a/python3-kokoro-onnx/README b/python3-kokoro-onnx/README new file mode 100644 index 0000000..30b632e --- /dev/null +++ b/python3-kokoro-onnx/README @@ -0,0 +1,15 @@ +kokoro-onnx is a Python library for text to speech with the Kokoro-82M +model, run through onnxruntime. It supports several languages, among +them English, Italian, French, Spanish, Portuguese, Japanese and +Chinese. + +Upstream depends on espeakng-loader, a wheel with its own copy of +libespeak-ng. This package drops it and uses the system espeak-ng +(/usr/share/espeak-ng-data and libespeak-ng.so.1) by default. An +EspeakConfig passed to Kokoro still overrides both paths. + +The model and voices are not included. Download kokoro-v1.0.onnx and +voices-v1.0.bin from the model-files-v1.0 release on the kokoro-onnx +GitHub page, and pass their paths to Kokoro(). + +Requires Python 3.10 or newer, so it does not build on Slackware 15.0. -- cgit v1.2.3