soniqo/Whisper-Medium-ONNX

Whisper Medium - ONNX

Official OpenAI Whisper medium 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-medium
Export format	ONNX for sherpa-onnx
Variants	FP32, INT8, FP16
Runtime	speech-core OnnxWhisperStt / ONNX Runtime; sherpa-onnx compatible
Total artifact size	5277.42 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
medium-decoder.fp16.onnx	1.62 MB	Whisper decoder graph
medium-decoder.fp16.onnx.data	774.93 MB	External tensor data for the adjacent ONNX graph
medium-decoder.int8.onnx	544.56 MB	Whisper decoder graph
medium-decoder.onnx	1551.82 MB	Whisper decoder graph
medium-encoder.fp16.onnx	0.44 MB	Whisper encoder graph
medium-encoder.fp16.onnx.data	682.02 MB	External tensor data for the adjacent ONNX graph
medium-encoder.int8.onnx	356.81 MB	Whisper encoder graph
medium-encoder.onnx	1364.44 MB	Whisper encoder graph
medium-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(
    "medium-encoder.int8.onnx",
    "medium-decoder.int8.onnx",
    "medium-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="medium-encoder.fp16.onnx",
    decoder="medium-decoder.fp16.onnx",
    tokens="medium-tokens.txt",
    language="en",
    task="transcribe",
    provider="cpu",
)

Benchmarks

This repository ships the exported model artifacts. Full published WER/RTF is currently available for the turbo bundle; run the benchmark command above for local numbers on this variant.

Links

• Website: soniqo.audio
• Speech Core: github.com/soniqo/speech-core
• Docs: soniqo.audio/speech-core
• Blog: soniqo.audio/blog