STT/ASR - onnx
Collection
OVOS STT/ASR models for onnx-asr (ONNX runtime). Most ship fp32 + int8 (set quantization: int8 for faster/smaller CPU inference). • 233 items • Updated • 2
distil-large-v3.5 — ONNX
ONNX export of distil-whisper/distil-large-v3.5
(English-only knowledge-distilled Whisper large-v3) for onnx-asr
(standard whisper model type — works with stock onnx-asr, no patches needed).
fp32 and int8 variants included; int8 decoder produced by quantizing the pre-merge
decoders and re-merging (the merged graph's If-subgraphs defeat direct quantization).
Note: this model is English-only (source repo language: [en]). Do not use for
other languages.
Usage
import onnx_asr
model = onnx_asr.load_model("whisper", "path/to/this/repo") # or quantization="int8"
print(model.recognize("audio_16khz.wav", language="en"))
Verified on a FLEURS English clip: fp32 and int8 outputs match each other and the
native transformers pipeline output exactly. RTF on an AMD Ryzen 5 7600 (6C/12T CPU,
shared/loaded host): ~0.9-2.0 fp32, ~0.6-1.0 int8 (wide range reflects host contention
during measurement, not model instability).
Model tree for OpenVoiceOS/distil-large-v3.5-onnx
Base model
distil-whisper/distil-large-v3.5