Embedl's build hexgrad/kokoro-82m of the original model hexgrad/Kokoro-82M
huggingface_fp16_tensorrt
- huggingface
- INT8 + FP16/FP32
- Orin NX 16GB
- Labels
- Text
- Speech
- Generation
- Build
- Embedl-optimized
- Precision
- INT8 + FP16/FP32
- Runtime
- TensorRT 10.16.2.10 with Embedl optimizations
- Output
- 24 kHz mono speech, US English
Kokoro-82M turns English text into natural 24 kHz speech in three voices. Choose af_heart, af_bella or am_michael and a speed of 0.8–1.2.
Apache-2.0 model and code; see NOTICE in the build for the frontend and text licences.