mirror of
https://github.com/modelscope/FunASR.git
synced 2026-10-11 22:39:57 +00:00
28 lines
6.0 KiB
HTML
28 lines
6.0 KiB
HTML
<!DOCTYPE html>
|
|
<html lang="en"><head><meta charset="UTF-8"><meta name="viewport" content="width=device-width,initial-scale=1.0">
|
|
<meta name="description" content="Deploy the third-party OpenMOSS MOSS-Transcribe-Diarize model through FunASR, vLLM, SGLang Omni, LocalAI, Docker, Kubernetes, and FunClip.">
|
|
<meta property="og:title" content="MOSS-Transcribe-Diarize deployment guide">
|
|
<meta property="og:description" content="Long-form transcription, timestamps, and anonymous speaker labels through the FunASR ecosystem.">
|
|
<meta property="og:type" content="website"><meta property="og:url" content="https://modelscope.github.io/FunASR/moss-transcribe-diarize.html">
|
|
<link rel="canonical" href="https://modelscope.github.io/FunASR/moss-transcribe-diarize.html"><meta property="og:site_name" content="FunASR">
|
|
<title>MOSS-Transcribe-Diarize deployment guide</title>
|
|
<link href="https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700;800&family=JetBrains+Mono:wght@400;500&display=swap" rel="stylesheet">
|
|
<link rel="stylesheet" href="style.css"></head><body>
|
|
<nav class="nav"><div class="container"><a href="index.html" class="nav-logo">FunASR</a><div class="nav-links"><a href="index.html">Home</a><a href="deployment-matrix.html" class="active">Deployment</a><a href="api.html">API</a><a href="agent.html">Agent</a></div><div class="lang-dropdown"><button class="lang-btn">English</button><div class="lang-menu"><a href="moss-transcribe-diarize.html" class="current">English</a><a href="zh/moss-transcribe-diarize.html">中文</a></div></div><a href="https://github.com/modelscope/FunASR" class="nav-github">GitHub</a></div></nav>
|
|
<div class="content"><div class="container narrow">
|
|
<h1>MOSS-Transcribe-Diarize</h1>
|
|
<p>FunASR integrates the third-party Apache-2.0 model <a href="https://huggingface.co/OpenMOSS-Team/MOSS-Transcribe-Diarize"><code>OpenMOSS-Team/MOSS-Transcribe-Diarize</code></a> published by OpenMOSS. It jointly generates long-form transcription, timestamps, and anonymous labels such as <code>S01</code> and <code>S02</code>, without an external VAD or speaker pipeline.</p>
|
|
<div class="toc-grid"><a href="#contract">Contract</a><a href="#server">FunASR server</a><a href="#runtimes">Runtime matrix</a><a href="#evidence">Evidence</a><a href="#boundaries">Boundaries</a></div>
|
|
<section id="contract"><h2>Output contract</h2><table><tr><th>Field</th><th>Meaning</th></tr><tr><td><code>text</code></td><td>Readable transcript without MOSS control tags.</td></tr><tr><td><code>timestamp</code></td><td>Segment-level start and end times in milliseconds.</td></tr><tr><td><code>sentence_info</code></td><td>Normalized segments with <code>start</code>, <code>end</code>, <code>text</code>, and anonymous <code>spk</code>.</td></tr><tr><td><code>raw_text</code></td><td>Exact tagged model generation retained for auditing.</td></tr></table></section>
|
|
<section id="server"><h2>FunASR OpenAI-compatible service</h2><pre><code>python -m pip install "transformers>=5.6,<6" fastapi uvicorn python-multipart
|
|
funasr-server --model moss-transcribe-diarize --device cuda:0 --port 8000
|
|
|
|
curl -fsS http://127.0.0.1:8000/v1/audio/transcriptions \
|
|
-F [email protected] \
|
|
-F model=moss-transcribe-diarize \
|
|
-F response_format=verbose_json</code></pre><p>The response preserves native anonymous speaker segments. A generic <code>spk=true</code> field does not start a second diarization model for MOSS.</p></section>
|
|
<section id="runtimes"><h2>Deployment paths</h2><table><tr><th>Path</th><th>Use when</th><th>Contract</th></tr><tr><td>FunASR server</td><td>You need a self-hosted HF GPU endpoint, Docker Compose, or Kubernetes.</td><td><code>verbose_json</code> with speaker segments.</td></tr><tr><td>vLLM</td><td>You operate its scheduler and need the native audio endpoint.</td><td><code>diarized_json</code>; FunASR can normalize it through <code>AutoModel</code>.</td></tr><tr><td>SGLang Omni</td><td>You use its native MOSS pipeline.</td><td><code>verbose_json</code>; FunASR validates speaker prefixes before normalization.</td></tr><tr><td>LocalAI / moss-transcribe.cpp</td><td>You need a third-party GGUF CPU or desktop-edge path.</td><td>Independent C++ implementation; validate its output separately.</td></tr><tr><td>FunClip</td><td>You need speaker-aware SRT and clips.</td><td>Consumes the normalized <code>sentence_info</code> contract.</td></tr></table><p><a href="https://www.funasr.com/en/deploy/moss-transcribe-diarize.html">Open the production deployment hub for complete pinned commands →</a></p></section>
|
|
<section id="evidence"><h2>Verified service smoke</h2><p>On one H100 80GB with Transformers <code>5.16.0.dev0</code>, Torch <code>2.11.0+cu130</code>, and model revision <code>e8681d68</code>, the real FunASR HTTP endpoint processed the bundled 6.0-second sample and returned HTTP 200, <code>duration=6.0</code>, and one non-empty monotonic segment labelled <code>S01</code>. This proves the service contract only, not accuracy, throughput, concurrency, or capacity.</p></section>
|
|
<section id="boundaries"><h2>Production boundaries</h2><ul><li>MOSS is owned and published by OpenMOSS; FunASR owns this adapter and deployment contract.</li><li>Pin and audit the model revision because the HF path uses <code>trust_remote_code</code>.</li><li>MOSS is an offline long-form model and is not exposed by the realtime WebSocket service.</li><li>Speaker labels are anonymous clustering labels, not known-person identity or voiceprint verification.</li><li>Do not externally VAD-split a meeting when global speaker continuity matters.</li></ul><p><a href="https://github.com/modelscope/FunASR/blob/main/docs/moss_transcribe_diarize.md">Repository guide</a> · <a href="https://github.com/OpenMOSS/MOSS-Transcribe-Diarize">OpenMOSS upstream</a> · <a href="https://www.funasr.com/en/blog/funclip-v2-2-0-moss-speaker-clipping.html">FunClip integration</a></p></section>
|
|
</div></div><footer><p>FunASR · Tongyi Lab, Alibaba Group</p><p><a href="index.html">Home</a> · <a href="deployment-matrix.html">Deployment</a> · <a href="https://github.com/modelscope/FunASR">GitHub</a></p></footer></body></html>
|