Whisper Large v3 Turbo - ONNX

Official OpenAI Whisper large-v3-turbo checkpoint exported as ONNX encoder/decoder graphs for speech-core and sherpa-onnx-compatible runtimes.

Part of the soniqo.audio speech toolkit. This is the ONNX Runtime bundle used by speech-core for server, desktop, and Android-style runtimes; the graphs remain sherpa-onnx compatible. Browse ONNX bundles in the soniqo ONNX collection.

Model


Source	openai/whisper-large-v3-turbo
Export format	ONNX for sherpa-onnx
Variants	FP32, INT8, FP16
Runtime	speech-core `OnnxWhisperStt` / ONNX Runtime; sherpa-onnx compatible
Total artifact size	5619.69 MiB

The FP16 graphs use external *.onnx.data files. Large-style bundles also use external FP32 *.weights files. Keep external-data files beside their matching .onnx files.

Files

File	Size	Description
`turbo-decoder.fp16.onnx`	0.59 MB	Whisper decoder graph
`turbo-decoder.fp16.onnx.data`	302.88 MB	External tensor data for the adjacent ONNX graph
`turbo-decoder.int8.onnx`	344.34 MB	Whisper decoder graph
`turbo-decoder.onnx`	0.97 MB	Whisper decoder graph
`turbo-decoder.weights`	605.76 MB	External tensor data for the adjacent ONNX graph
`turbo-encoder.fp16.onnx`	0.61 MB	Whisper encoder graph
`turbo-encoder.fp16.onnx.data`	1239.93 MB	External tensor data for the adjacent ONNX graph
`turbo-encoder.int8.onnx`	643.37 MB	Whisper encoder graph
`turbo-encoder.onnx`	0.6 MB	Whisper encoder graph
`turbo-encoder.weights`	2479.86 MB	External tensor data for the adjacent ONNX graph
`turbo-tokens.txt`	0.78 MB	Tokenizer tokens for speech-core and sherpa-onnx-compatible runtimes

Usage

Use with speech-core's native ONNX Whisper runtime:

#include <speech_core/models/onnx_whisper_stt.h>

speech_core::OnnxWhisperStt stt(
    "turbo-encoder.int8.onnx",
    "turbo-decoder.int8.onnx",
    "turbo-tokens.txt");

auto result = stt.transcribe(audio, length, 16000);

The same encoder/decoder/token files can also be loaded by sherpa-onnx:

import sherpa_onnx

recognizer = sherpa_onnx.OfflineRecognizer.from_whisper(
    encoder="turbo-encoder.fp16.onnx",
    decoder="turbo-decoder.fp16.onnx",
    tokens="turbo-tokens.txt",
    language="en",
    task="transcribe",
    provider="cpu",
)

Full FLEURS WER / RTF

M5 Pro, local FLEURS test split, sherpa-onnx CPU provider, language hint supplied, temperature fallback disabled.

Dataset	INT8 WER / RTF	FP16 WER / RTF
`fleurs-en_us`	5.32% / 0.239	5.27% / 0.189
`fleurs-fr_fr`	6.84% / 0.188	6.57% / 0.215
`fleurs-ar_eg`	14.53% / 0.201	14.30% / 0.215

Note: the French split contains one 33.54 s clip; sherpa-onnx Whisper processes the first 30 s.

soniqo/Whisper-Large-v3-Turbo-ONNX

Whisper Large v3 Turbo - ONNX

Model

Files

Usage

Full FLEURS WER / RTF

Links

Seed this model