云来 听间 — 第三方组件与模型
Updated: 2026-11-07

SenseVoiceSmall (FunAudioLLM / Alibaba Group)
Original model: https://huggingface.co/FunAudioLLM/SenseVoiceSmall
INT8 conversion: https://huggingface.co/csukuangfj/sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/tree/2365baeacb507f821a0c8120fcee3d484dba7a07
Distributed as sensevoice_model.onnx and sensevoice_tokens.txt (or transport parts).
License: FunASR-MODEL_LICENSE.txt (v1.1; model weights, not the code license).

pyannote segmentation 3.0 (CNRS)
Original model: https://huggingface.co/pyannote/segmentation-3.0
ONNX conversion: https://huggingface.co/csukuangfj/sherpa-onnx-pyannote-segmentation-3-0/tree/9403a6902bb58e3d5ae8c7e77c3422de279db2e0
Distributed as speaker_segmentation.onnx (or transport parts).
License: Pyannote-LICENSE.txt (MIT, CNRS).
License source: https://github.com/pyannote/pyannote-audio/blob/develop/LICENSE

CAM++ / 3D-Speaker (Alibaba DAMO Academy / ModelScope)
Model: https://modelscope.cn/models/iic/speech_campplus_sv_zh_en_16k-common_advanced
ONNX release: https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-recongition-models
Distributed as speaker_embedding.onnx (or transport parts).
License: CAMplusplus-LICENSE.txt (Apache-2.0).
License source: https://github.com/modelscope/3D-Speaker/blob/main/LICENSE

Silero VAD (Silero Team)
Original project: https://github.com/snakers4/silero-vad
ONNX release: https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models
Distributed as vad.onnx (or transport parts).
License: Silero-VAD-LICENSE.txt (MIT).

sherpa-onnx 2.0.1 (k2-fsa contributors)
Source: https://github.com/k2-fsa/sherpa-onnx/tree/v2.0.1
Distributed as sherpa.js and sherpa.wasm (or transport parts).
License: Sherpa-onnx-LICENSE.txt (Apache-2.0).
Modification notice: sherpa.js adapts the Node package's path helper and
NODERAWFS guard for browser Workers, and bundles the ASR, VAD and diarization
wrappers. The WebAssembly binary is unchanged. The browser adaptation is
documented in tools/prepare_browser_workbench.py in this project.

Mediabunny 1.65.2 (Vanilagy / contributors)
Source: https://github.com/Vanilagy/mediabunny/tree/v1.65.2
Distributed as the unmodified browser bundle vendor/mediabunny.min.mjs.
License: Mediabunny-LICENSE.txt (MPL-2.0, copied from the pinned npm package).

Pages model parts are byte-for-byte slices of the above model/runtime files.
The complete-file SHA256 values remain in workbench-models.json and are
verified by the browser before the files are cached or used.