AI Frameworks · modelscope

FunClip

FunClip is an open-source video transcription and clipping tool built on Alibaba's FunASR, with a local Gradio UI for subtitle generation and LLM-assisted intelligent clipping. It supports Chinese and English audio, speaker diarization, and integrates with large language models for content-aware segment selection.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/modelscope/FunClip

5.9k

GitHub stars

709

Forks

Python

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	modelscope/FunClip
Owner	modelscope
Primary language	Python
License	MIT — OSI-approved
Stars	5.9k
Forks	709
Open issues	0
Latest release	Unknown
Last updated	2026-07-07
Source	https://github.com/modelscope/FunClip

What FunClip is

Python-based tool leveraging Paraformer ASR models (with optional Fun-ASR-Nano, SenseVoice, Whisper alternatives) for speech-to-text with timestamp prediction, CAM++ for speaker recognition, and optional LLM integration (Qwen, GPT, TwelveLabs Pegasus) for semantic clipping. Deployed via Gradio UI with FFmpeg/ImageMagick for video processing.

Quickstart

Get the FunClip source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/modelscope/FunClip.git2cd FunClip3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Content Creator Workflow Automation

Rapidly generate SRT subtitles and auto-clip highlight segments from long-form video (interviews, streams, tutorials) without manual timeline scrubbing. Speaker diarization enables multi-speaker content segmentation.

Multilingual Video Localization

Process 31+ languages (via Fun-ASR-Nano) or multilingual audio with emotion/event detection (SenseVoice) to create localized subtitle files and clip region-specific segments for international audiences.

LLM-Driven Intelligent Clipping

Use Qwen/GPT or TwelveLabs Pegasus to semantically identify key moments (e.g., 'funniest moments', 'technical explanations') from transcript+audio context, reducing manual curation for social media repurposing.

Implementation considerations

Dependency on external model repos (Modelscope, HuggingFace) for weights; verify download reliability and caching strategy in your deployment region.
LLM integrations (Qwen, GPT, TwelveLabs) require API keys and real-time external calls; budget for API costs and implement retry/fallback logic.
Optional ImageMagick dependency for embedded subtitle rendering requires OS-level package installation and policy file modification (potential security/compatibility risk).
No built-in queue management or concurrent job handling; Gradio UI suitable for single-user or small team workflows; scale horizontally with multiple instances for production.
ASR accuracy varies by language, domain, and audio quality; hotword customization mitigates some domain issues but requires tuning per use case.

When to avoid it — and what to weigh

Strict Real-Time Requirements — FunClip is designed for local batch processing; ASR inference latency (especially for long videos) makes it unsuitable for live streaming or sub-second transcription needs.
Enterprise Compliance/Audit Trails — No mention of logging, audit mechanisms, or compliance frameworks (SOC2, HIPAA); unclear how to integrate with enterprise governance or data handling policies.
Proprietary Codec or DRM Content — Relies on FFmpeg for video processing; support for proprietary formats, DRM, or fragmented media (HLS, DASH) is not documented. May fail silently on unsupported containers.
GPU-Constrained Environments — While Fun-ASR-Nano is lighter, standard Paraformer and Pegasus integration require significant VRAM; no guidance on CPU-only inference or quantized model support documented.

License & commercial use

MIT License: permissive, allows commercial use, modification, and distribution with minimal restriction (attribution required). No copyleft obligations.

MIT permits commercial deployment without licensing restrictions. However, external dependencies (FunASR models from Modelscope, Qwen/GPT APIs, TwelveLabs Pegasus) have separate terms: verify compliance with model provider agreements, API ToS, and any enterprise licensing for large-scale usage. ImageMagick license (various: AGPL, commercial) may apply if embedded subtitle rendering is used; review per your jurisdiction.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Good
Assessment confidence	High

Security considerations

Gradio UI exposes endpoints without documented authentication; public deployment (via `-s True`) without reverse proxy auth is unsafe for untrusted networks. File upload handling not detailed; ensure input validation and storage isolation. API key management for LLM/Pegasus stored client-side (potential exposure in Gradio state). FFmpeg subprocess execution via system binaries—validate input file paths to prevent injection. No mentions of data retention, deletion, or privacy handling for uploaded videos.

Alternatives to consider

OpenAI Whisper + FFmpeg-python

Whisper offers robust multilingual ASR without hotword tuning; clip logic is custom-coded but transparent. Scales better on CPU. No LLM integration; requires separate orchestration.

AssemblyAI API + Zapier/Make

Fully managed, no local infrastructure; real-time transcription and speaker diarization. Higher per-minute cost; less control over models and no offline fallback.

All-in-one video editing, transcription, and AI clipping; polished UI and enterprise support. Closed-source; no local deployment; higher cost for high volume.

Software development agency

Build on FunClip with DEV.co software developers

Deploy FunClip locally for instant transcription and AI-driven clipping. Integrate with your existing infrastructure, customize ASR models, and scale with your team.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

autoclip NarratoAI fg-clip-base

Related on DEV.co

Explore the category and the services that help you build with it.

AI Frameworks AI Application Development API Development Cloud Deployment Talk to DEV.co

FunClip FAQ

Can I use FunClip offline after model download?

Yes, for ASR and clipping. LLM-based clipping (Qwen, GPT, Pegasus) requires online API calls. FunASR models download on first run; ensure stable internet for initial setup.

What video formats are supported?

Any format supported by FFmpeg (MP4, MKV, AVI, MOV, etc.). Codec support depends on your FFmpeg build; verify h.264, h.265, VP9 availability for your deployment.

How do I scale this for a team or production service?

Gradio UI is single-instance; for multi-user: deploy multiple instances behind a load balancer, add a job queue (Celery, RQ), and manage API keys securely. No native auth—add reverse proxy (Nginx with OAuth2) for access control.

What are typical ASR inference times?

Not documented. Depends on video duration, audio quality, model (Paraformer vs. Fun-ASR-Nano vs. SenseVoice), and GPU. Recommend benchmarking on your target hardware.

Custom software development services

DEV.co helps companies turn open-source tools like FunClip into production software. Our software development services cover the full lifecycle — architecture, web development, integration, and maintenance — delivered by software developers and web developers who ship. Engage our software development agency to implement or customize it for your ai frameworks stack.

Automate Your Video Workflow with FunClip

Deploy FunClip locally for instant transcription and AI-driven clipping. Integrate with your existing infrastructure, customize ASR models, and scale with your team.

Talk to DEV.co Schedule a Consultation