# AI use cases for FunASR (modelscope/FunASR)

Repo: [modelscope/FunASR](https://github.com/modelscope/FunASR) | Use cases: 3 | Page: https://dedupedai.com/use-cases/modelscope-funasr/

FunASR is an industrial speech recognition toolkit for offline, streaming and edge deployment. It bundles ASR models such as SenseVoiceSmall, Paraformer and Fun-ASR-Nano with VAD, punctuation, speaker clustering and emotion models, a `funasr` CLI with JSON and SRT output, an OpenAI-compatible `funasr-server`, and llama.cpp GGUF binaries that run on CPU without Python.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Make SRT subtitles from audio files: https://dedupedai.com/use-cases/modelscope-funasr/make-srt-subtitles-from-audio-with-funasr/
- Transcribe a meeting with speaker labels: https://dedupedai.com/use-cases/modelscope-funasr/transcribe-meetings-with-speaker-labels-using-funasr/
- Run a local OpenAI style transcription API: https://dedupedai.com/use-cases/modelscope-funasr/run-a-local-openai-compatible-transcription-server-with-funa/

## How to make SRT subtitles from audio files with FunASR

Repo: [modelscope/FunASR](https://github.com/modelscope/FunASR) | Difficulty: Beginner | Time: 20 minutes | Basis: From the README

Use the agent friendly funasr CLI to transcribe one file or a whole batch straight to SRT subtitle files.

**Who it is for:** A video creator or podcaster who needs captions.

**The problem:** Captioning by hand is slow, and many tools handle Chinese and mixed language audio poorly.

**The outcome:** An SRT file per audio file in an output folder.

### Steps

1. Install with `pip install torch torchaudio` and then `pip install funasr`
2. Transcribe one file with `funasr audio.wav` to check the output
3. Write subtitles with `funasr audio.wav --output-format srt --output-dir ./subs`
4. Batch a folder with `funasr *.wav --output-format srt --output-dir ./output`

### Prompt to paste into your agent

```text
Using modelscope/FunASR, install it on CPU and run `funasr *.wav --output-format srt --output-dir ./output` in <my audio folder>, then show me the first subtitle file.
```

### Caveat

The default CLI model is sensevoice; pick paraformer-en or fun-asr-nano with --model for other needs.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/modelscope-funasr/make-srt-subtitles-from-audio-with-funasr/

## How to transcribe a meeting with speaker labels with FunASR

Repo: [modelscope/FunASR](https://github.com/modelscope/FunASR) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Combine ASR with FSMN-VAD and CAM++ speaker clustering to get timestamped segments tagged by anonymous speaker.

**Who it is for:** Someone writing up meeting or interview notes.

**The problem:** Plain transcripts do not say who spoke when.

**The outcome:** JSON or printed segments with start times, a speaker index and text.

### Steps

1. Install with `pip install torch torchaudio` and `pip install funasr`
2. For a quick result run `funasr audio.wav --spk --timestamps -f json`
3. For more control, use the README's Python example: AutoModel with model iic/SenseVoiceSmall, vad_model fsmn-vad and spk_model cam++, then loop over sentence_info
4. Rename the anonymous speaker indices to real names yourself after reviewing

### Prompt to paste into your agent

```text
With modelscope/FunASR, run `funasr audio.wav --spk --timestamps -f json` on <my meeting recording> and turn the JSON into a readable transcript grouped by speaker.
```

### Caveat

The README says speaker indices are local to each recording, not known person identities.

Risk, privacy: Get consent before recording and transcribing other people.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/modelscope-funasr/transcribe-meetings-with-speaker-labels-using-funasr/

## How to run a local OpenAI style transcription API with FunASR

Repo: [modelscope/FunASR](https://github.com/modelscope/FunASR) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Start funasr-server on loopback with SenseVoice on CPU and call it like the OpenAI audio transcriptions endpoint.

**Who it is for:** A developer wiring speech to text into an app or agent without a cloud ASR bill.

**The problem:** Cloud transcription APIs cost money and send audio off your machine.

**The outcome:** A local /v1/audio/transcriptions endpoint returning verbose JSON.

### Steps

1. Create an env with `python3.11 -m venv .venv-funasr-http` and `. .venv-funasr-http/bin/activate`
2. Install with `python -m pip install torch torchaudio` and `python -m pip install funasr fastapi uvicorn python-multipart`
3. Start the server with `funasr-server --host 127.0.0.1 --port 8000 --model sensevoice --device cpu`
4. In a second terminal, POST a wav file to http://127.0.0.1:8000/v1/audio/transcriptions with model=sensevoice and response_format=verbose_json, as in the README curl example

### Prompt to paste into your agent

```text
Using modelscope/FunASR, set up the Python 3.11 venv, start `funasr-server --host 127.0.0.1 --port 8000 --model sensevoice --device cpu`, and send <my wav file> to /v1/audio/transcriptions with curl.
```

### Caveat

Wait for the model download and server startup before sending the first request.

Risk, security: The service is unauthenticated, so keep it on 127.0.0.1 as the README advises.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/modelscope-funasr/run-a-local-openai-compatible-transcription-server-with-funa/
