TypeWhisper/typewhisper-win model-fluid-speech-pro-int8-v1

Fluid Speech Pro model files (int8)

on GitHub

Model files the NVIDIA Parakeet plugin downloads for Fluid Speech Pro. This is not an app or plugin release.

FluidVoice-Labs/fluid-speech-pro-onnx at revision fa588e9f87f1b9ad0ca2e2d8f7eaf84dc3803581, FluidVoice's ONNX export of Fluid Speech Pro. TypeWhisper quantized its float files to int8 with eng/quantize_fluid_speech_pro.py (onnx 1.18.0, onnxruntime 1.22.1): dynamic quantization like csukuangfj/sherpa-onnx-nemo-parakeet-tdt-0.6b-v3-int8, with the encoder's convolutions left in float. tokens.txt is FluidVoice's vocab.txt.

License: CC-BY-4.0. Fluid Speech Pro by FluidVoice. Built on NVIDIA Parakeet TDT 0.6B v3 and Moondream Parakeet Ultra.

SHA-256:

• encoder.int8.onnx 44a0312c54cc9b337ec72f78216df26720c107a9f14d41bd6062255fcde35a0a
• decoder.int8.onnx ce5cb0576d3c072d12188944a29737d7b81f329a29d1c9f636669e9b9d4672bf
• joiner.int8.onnx 231442707e1850d53a7d66dea6a4b5264ae9f38d2468980cfe8ee8b7847097ab
• tokens.txt d58544679ea4bc6ac563d1f545eb7d474bd6cfa467f0a6e2c1dc1c7d37e3c35d

(#780)