DialectLearn

docs: fix model attribution — we use CosyVoice 3 ONNX weights

docs: add comprehensive README and MIT license for open source

test: add test-output7 after prompt text fix

fix: correct prompt texts to match actual reference audio content

test audio 6