Repolode
Submit
modelscope

modelscope/FunClip

Open-source local video clipping tool that transcribes speech with FunASR and cuts videos by selected text or speaker, with optional LLM-assisted clipping.

View on GitHubWebsite
Stars
6,503
Forks
774
Open issues
17
Last commit
6 days ago

What is it?

FunClip is a locally deployed tool for clipping videos based on their speech. It runs speech recognition on a video using FunASR's Paraformer models, then lets you pick text segments or speakers from the transcript and export the video clips that match them. It also generates SRT subtitles for the full video and for the selected segments.

The main interface is a Gradio web UI that runs on localhost and can be deployed on a server for browser access. The command line offers a two-step workflow: recognize first, then clip by target text with optional start and end offsets. Recognition can use Paraformer, the Fun-ASR-Nano, SenseVoice, or MOSS backends, and Paraformer supports hotword customization and CAM++ speaker identification.

FunClip also includes an LLM-assisted clipping mode. It sends the transcript to a chosen language model, including models reached through OrcaRouter, Cheaper Inference, or TwelveLabs Pegasus, which analyzes the video itself, and returns segments for clipping.

Key features

  • 01Transcribes video speech with FunASR Paraformer and other selectable ASR models.
  • 02Clips video by selecting text segments or speakers from the transcript.
  • 03Exports SRT subtitles for the full video and the clipped segments.
  • 04Provides a Gradio web UI that runs locally or on a server.
  • 05Supports LLM-assisted clipping through configurable model gateways.

Use cases

  • Pulling highlight clips out of long Chinese or multilingual talks and interviews.
  • Generating subtitle files for a video without manual typing.
  • Extracting one speaker's segments from a multi-person recording.
  • Running a private video clipping service on your own machine or server.

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/modelscope/FunClip.git
  2. 2

    Enter the project folder

    cd FunClip
  3. 3

    Create and activate a Python 3.12 virtual environment, then install requirements

    python3.12 -m venv .venv && . .venv/bin/activate && python -m pip install -r requirements.txt
  4. 4

    Start the local Gradio service, then open localhost:7860 in a browser

    python funclip/launch.py

Things to know

  • Requires installing PyTorch/torchaudio manually for your platform before the other requirements.
  • Optional LLM and Pegasus clipping needs a paid or separately obtained API key.

In collections

AI video & audio tools

Topics

#speech-recognition#video-editing#subtitles#asr#transcription#llm#gradio#chinese#ai-tools#ai-video-editing

FAQ

What is FunClip?
Open-source local video clipping tool that transcribes speech with FunASR and cuts videos by selected text or speaker, with optional LLM-assisted clipping. FunClip is a locally deployed tool for clipping videos based on their speech. It runs speech recognition on a video using FunASR's Paraformer models, then lets you pick text segments or speakers from the transcript and export the video clips that match them. It also generates SRT subtitles for the full video and for the selected segments.
Is FunClip free to use? What license does it have?
FunClip is open source under the MIT license. Check the license terms before using it in commercial products.
Is FunClip still maintained?
Actively maintained. The last commit to FunClip was on Nov 1, 2026.
What can I use FunClip for?
Pulling highlight clips out of long Chinese or multilingual talks and interviews. Generating subtitle files for a video without manual typing. Extracting one speaker's segments from a multi-person recording. Running a private video clipping service on your own machine or server.
How do I install FunClip?
Clone the repository: git clone https://github.com/modelscope/FunClip.git Enter the project folder: cd FunClip Create and activate a Python 3.12 virtual environment, then install requirements: python3.12 -m venv .venv && . .venv/bin/activate && python -m pip install -r requirements.txt Start the local Gradio service, then open localhost:7860 in a browser: python funclip/launch.py
Where can I get FunClip?
The source code and installation instructions are on GitHub: https://github.com/modelscope/FunClip

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects