Kokoro 82M FP16 ONNX

--- license: apache-2.0 base_model: hexgrad/Kokoro-82M base_model_relation: quantized pipeline_tag: text-to-speech library_name: onnxruntime language: - en tags: - onnx - onnxruntime - kokoro - tts - text-to-speech - fp16 --- # Kokoro-82M — FP16 ONNX ONNX export of [hexgrad/Kokoro-82M](https://huggingface.co/hexgrad/Kokoro-82M), an open-weight English text-to-speech model, stored in FP16 (half-precision) format. It is not a newly trained model. This is an unofficial community conversion; hexgrad is the original model author. ## Source | Field | Value | |---|---| | Upstream model | [onnx-community/Kokoro-82M-v1.0-ONNX](https://huggingface.co/onnx-community/Kokoro-82M-v1.0-ONNX) | | Upstream source revision | `1939ad2a8e416c0acfeecc08a694d14ef25f2231` | | Export tool/script | Mirrored from onnx-community (no re-export) | | Quantization recipe | FP16 ONNX mirrored from `onnx/model_fp16.onnx` | ## Files | File | Size | Description | |---|---|---| | `onnx/model_fp16.onnx` | ~156 MB | FP16 ONNX model (self-contained) | ## Intended Use An English text-to-speech model packaged for local inference via ONNX Runtime. The FP16 variant targets GPU execution for faster inference and higher audio-quality preservation than quantized variants. The Q8/FP16 variant is in [`tonythethompson/kokoro-82m-q8f16-onnx`](https://huggingface.co/tonythethompson/kokoro-82m-q8f16-onnx). ## Runtime Notes - Designed for ONNX Runtime compatible runtimes. - Execution: GPU (CUDA) preferred for FP16; CPU fallback may work but is untested and slower. - Sample rate: 24 kHz (inherited from Kokoro-82M). - Validate on the target execution provider before production use. ## Precision and Packaging Export tooling, precision, and quantization are recorded in the **Source** table above. This packaging mirror does not publish independent parity benchmarks; validate on your target execution provider before production use. ## Limitations - English-only TTS. - No parity evaluation against the original Kokoro-82M PyTorch model is documented here. - Voice quality may differ from the reference PyTorch implementation due to ONNX export approximations. - FP16 requires hardware support for float16; some CPU runtimes may upcast silently or error. ## Safety and Responsible Use Kokoro-82M produces realistic synthetic speech (from preset voice styles, not arbitrary...