Simplify README: fp16-only variant. Document int8 ConvInteger limitation.

Remove int8 variant: ConvInteger NOT_IMPLEMENTED in ORT minimal/Node builds. Use fp16.

Update README: add int8 variant, voice catalog, conversion details

Remove old hybrid/ folder (renamed to int8/)

Rename hybrid -> int8 (selective quantization: VE int8 + rest fp16, 130MB)

Add hybrid quantization (VE int8 dynamic + rest fp16, 130MB)

docs: drop .fp16. infix from filename examples in README

Remove old fp16/onnx/vocoder.fp16.onnx (renamed without .fp16. infix)

Remove old fp16/onnx/vector_estimator.fp16.onnx (renamed without .fp16. infix)

Remove old fp16/onnx/duration_predictor.fp16.onnx (renamed without .fp16. infix)

Remove old fp16/onnx/text_encoder.fp16.onnx (renamed without .fp16. infix)

Rename ONNX files: drop redundant .fp16. infix (variant lives in folder)

docs: update README to reflect fp16/onnx + voice_styles/ layout

Remove old fp16/voice_styles/ (moved to repo root)

Remove old fp16/unicode_indexer.json (moved to fp16/onnx/)

Remove old fp16/tts.json (moved to fp16/onnx/)

Move voice_styles to repo root (variant-independent)

Move tts.json + unicode_indexer.json into fp16/onnx/

Add fp16 variant

Update README.md

Create LICENSE

Update README.md

Update README.md

Update README.md

Update README.md

initial commit