Funclip
Open-source, accurate and easy-to-use video speech recognition & clipping tool. LLM-based AI clipping integrated.
Last verified:
What is Funclip?
FunClip is a fully open-source, locally deployed automated video clipping tool that leverages Alibaba TONGYI speech lab's open-source FunASR Paraformer series models to perform speech recognition on videos. After transcribing the audio, users can freely select text segments or specific speakers from the recognition results and click a clip button to extract the corresponding video portions.
Key features include integration of Alibaba's industrial-grade Paraformer-Large model (one of the best-performing open-source Chinese ASR models with over 13 million Modelscope downloads), SeACo-Paraformer hotword customization for specifying entity words and names during recognition, and CAM++ speaker recognition model for clipping segments from specific speakers by ID. The tool provides a Gradio-based web interface for easy interaction and can be deployed locally or on a server accessible via browser.
FunClip supports multi-segment free clipping, automatically generating full-video SRT subtitles and target-segment SRT subtitles. It accurately predicts timestamps in an integrated manner and allows users to adjust subtitle settings before clipping. The tool is ideal for content creators, video editors, social media managers, and anyone who needs to extract highlights from long videos without manual scrubbing.
The application works through a simple four-step workflow: upload video, copy needed text segments to 'Text to Clip', adjust subtitle settings if needed, then click 'Clip' or 'Clip and Generate Subtitles'. It also supports command-line usage for batch processing and automated workflows.
Funclip pricing
Pricing model: Freemium
FunClip is 100% free and open-source with no paid tiers. It is fully self-hosted on your own hardware, so there are no subscription fees, usage limits, or API costs. All features including Paraformer-Large ASR, hotword customization, speaker diarization, and subtitle generation are included at no cost. Users only need to provide their own hardware (Python 3.8+, GPU 8GB+ recommended) and download the font file for subtitle rendering.
Funclip pros
- Fully open-source with no licensing costs
- Locally deployed for complete data privacy
- Paraformer-Large model with industry-leading Chinese ASR accuracy
- Over 13 million downloads on Modelscope proving reliability
- Hotword customization for names and entity words
- CAM++ speaker diarization for speaker-specific clipping
- Automatic timestamp prediction with high precision
- Gradio web interface is simple and intuitive
- Multi-segment free clipping in one operation
- Auto-generates SRT subtitles for full video and clipped segments
- Supports both GUI and command-line usage
- Can be deployed on server for team access
- No code required for basic video clipping
- Integrated VAD segmentation and punctuation
- Works with example videos for quick testing
Funclip cons
- Primarily optimized for Chinese audio recognition
- Requires本地 GPU with 8GB+ VRAM for best performance
- English support limited, Whisper model not yet available
- ImageMagick installation required for embedded subtitles
- Font file must be manually downloaded for subtitle rendering
- No cloud-based option, must self-host
- LLM integration still in development
- Steep setup for command-line users unfamiliar with Python
Frequently asked questions about Funclip
What is FunClip and what does it do?
FunClip is a fully open-source, locally deployed automated video clipping tool. It uses Alibaba's FunASR Paraformer models to transcribe video audio, then lets users select text segments or speakers from the transcription to automatically extract the corresponding video clips with SRT subtitles.
Is FunClip free to use?
Yes, FunClip is completely free and open-source. There are no subscription fees, usage limits, or paid tiers. You self-host it on your own hardware and get all features including Paraformer-Large ASR, hotword customization, speaker diarization, and automatic subtitle generation at no cost.
How do I install FunClip?
Clone the repository with 'git clone https://github.com/alibaba-damo-academy/FunClip.git', cd into the directory, then run 'pip install -r ./requirements.txt'. To start the local Gradio service, run 'python funclip/launch.py' and visit localhost:7860 in your browser.
What languages does FunClip support?
FunClip primarily supports Chinese audio recognition using the Paraformer-Large model, which is one of the best-performing open-source Chinese ASR models. It supports 7 Chinese dialects and 26 regional accents. English support is limited, and Whisper model integration for English users is announced as 'coming soon'.
Can I clip videos by specific speaker?
Yes, FunClip integrates the CAM++ speaker recognition model. When you enable 'Recognize Speakers' during transcription, each sentence gets a speaker ID (spk0, spk1, etc.). You can then clip segments from specific speakers by entering their ID like 'spk0' or multiple speakers like 'spk0#spk3'.
What are hotwords and how do I use them?
Hotwords are custom entity words or names you specify to improve recognition accuracy for specific terms. FunClip uses SeACo-Paraformer's hotword customization feature. You can specify hotwords during the ASR process to enhance recognition of proper nouns, technical terms, or personalized vocabulary.
Does FunClip generate subtitles?
Yes, FunClip automatically generates SRT subtitle files for both the full original video and your clipped segments. You can choose 'Clip and Generate Subtitles' to get video with embedded subtitles, or just 'Clip' to get the video file alone.
Can I use FunClip on a server?
Yes, FunClip can be deployed on a server and accessed via browser. After running 'python funclip/launch.py' on your server, you can access the Gradio interface from any browser by navigating to the server's IP address and port 7860.
What hardware do I need to run FunClip?
FunClip requires Python 3.8+ and a GPU with 8GB+ VRAM for optimal performance, though CPU is also supported. For embedded subtitle rendering, you need to install ffmpeg and ImageMagick, and manually download a font file to the funclip/font directory.
Can I use FunClip via command line?
Yes, FunClip supports command-line usage for stage 1 recognition and stage 2 clipping. Use 'python funclip/videoclipper.py --stage 1' for recognition, then '--stage 2' with --dest_text parameter to clip specific text segments. This is useful for batch processing and automation.