简体中文 | English

ASRKit

One interface to run any speech-to-text model — local & cloud, any engine.

ASRKit is Ollama + LiteLLM for speech recognition. Pull an on-device model and run it with one command; call a cloud API by changing one string; snap in a whole new inference engine as a plugin. One contract, everything behind it.

pip install asrkit
asrkit run local/sensevoice audio.wav      # downloads on first run, then transcribes

One interface, swap anything

The model string is the only thing that changes — engine, model, on-device vs cloud, all the same call:

asrkit run local/sensevoice                      audio.wav              # sherpa-onnx (on-device, default)
asrkit run faster-whisper/large-v3               audio.wav              # faster-whisper engine
asrkit run whispercpp/base                       audio.wav              # whisper.cpp engine
asrkit run transformers/openai/whisper-large-v3  audio.wav              # any HuggingFace model
asrkit run siliconflow/sensevoice                audio.wav --api-key …  # cloud, bring your own key

from asrkit import transcribe
print(transcribe("local/sensevoice", "audio.wav").text)

Why ASRKit

• Curated and open. 47 on-device models work out of the box (asrkit list); anything else drops in via a JSON entry or a plugin — no fork required.
• Four engines, more via plugins. sherpa-onnx, faster-whisper, whisper.cpp, and transformers (the entire HuggingFace hub, incl. LLM-architecture SOTA). pip install asrkit-<engine> adds your own.
• Transparent by design. ASRKit doesn't touch your audio or change a model's native behavior by default. Wrong format? An honest error — never silent garbage. Format conversion and long-audio chunking are opt-in flags.
• Private. Your audio and API keys never leave your machine — ASRKit is a library, not a hosted service.
• Pull-and-go. Models download on demand and cache locally, Ollama-style. pip install asrkit ships only the default engine; the rest are optional.

Commands

Engines & models

Extend it

Add any sherpa model — drop an entry in ~/.asrkit/models.json:

[{"id": "local/my-model", "download_url": "https://…/model.tar.bz2", "config_type": "senseVoice", "langs": ["zh"]}]

Add an engine — ship a small package that declares an asrkit.adapters entry point; pip install auto-registers it. No fork, no core changes. Recipe: docs/engines-and-addressing.md.

Docs

Usage · Adapter contract · Engines & addressing · Model management

Apache-2.0. Each model's license is its own — verify before commercial use (asrkit show <model> points to the source). Your audio and keys stay on your machine.