🎞️ VIDEO · AI Video & Avatars Level: Intermediate ⏱ 10 MIN Free & Open Source

Empower Coding Agents with 42 Local Video Editing Superpowers via Open-Source FFmpeg-Skill MCP

The ultimate local video editing toolkit for Claude Code, Cursor, and Codex! FFmpeg-Skill exposes 42 machine-readable local FFmpeg tools over MCP (Model Context Protocol). Instruct your AI agent in plain English to trim silences, reframe landscape to vertical 9:16, burn styled subtitles, normalize audio to -14 LUFS, and export compressed MP4s with zero cloud API costs and 100% on-premise media privacy.

CORE TOOLS FFmpeg-Skill (Open Source), Model Context Protocol (MCP), FFmpeg, Claude Code / Cursor

BEST FOR Content creators, video editors, indie hackers, marketing managers, podcast producers

REQUIREMENTS PC running Windows, macOS, or Linux with FFmpeg installed, paired with an MCP-compatible AI agent (Claude Code, Cursor, Windsurf).

✦ WHY THIS WORKFLOW

FFmpeg command syntax is notoriously complex, while cloud video editing APIs charge exorbitant fees and risk media leakage. FFmpeg-Skill equips coding agents with a machine-readable local editing interface, making professional media manipulation as simple as talking.

✦ Interactive Workflow Animation Preview

Step-by-Step Action Checklist

0 / 4 Completed

✓ Step 1 [Install FFmpeg locally & verify CLI binaries]: Ensure the latest FFmpeg and FFprobe binaries are installed and accessible from your system PATH.
✓ Step 2 [Mount FFmpeg-Skill MCP in your AI coding agent]: Register the `ffmpeg-skill` server into Claude Code, Cursor, or Windsurf MCP settings to expose the 42 editing endpoints.
✓ Step 3 [Issue plain-English video processing instructions]: Prompt the agent: "Cut silences in `raw.mp4`, reframe to 9:16 centered on speaker, and normalize loudness to -14 LUFS".
✓ Step 4 [Autonomous filtergraph compilation & local export]: The agent composes complex FFmpeg filter graphs and executes locally, outputting polished media files in seconds.

READY-TO-USE PROMPT TEMPLATE

Open Official Tool / Repo ↗

I am automating video post-production using FFmpeg-Skill MCP with Claude Code/Cursor.
Please provide:
1) [Production MCP Config]: Exact JSON snippet registering `ffmpeg-skill` into `claude_desktop_config.json` and Cursor settings;
2) [3 High-Frequency Media Automation Prompts]: Prompts for auto silence stripping, 16:9 to 9:16 pillarbox blur reframing, and styled bilingual ASS subtitle baking;
3) [Hardware Acceleration Guidance]: Directives ensuring the agent uses NVENC or Apple VideoToolbox hardware encoders to slash file size by 70% while preserving visually lossless fidelity.

Use Cases & Tags:

Autonomous silent pause and filler-word detection and seamless jump-cutting across 2-hour raw footageSmart auto-reframing from 16:9 landscape to 9:16 vertical shorts keeping subjects centeredAutomating -14 LUFS broadcast audio normalization and burning styled bilingual subtitlesBatch compressing video portfolios according to exact platform codec specifications#ffmpeg-skill#mcp#claude-code#cursor

Related AI Workflows

← Describe your idea on Home

VIDEO · AI Video & Avatars

15 MIN

Deploy Concat: The Open-Source Local-First CapCut Alternative Scriptable via JSON-RPC & MCP

A viral local-first, open-source CapCut replacement with 11k+ GitHub stars! Concat eliminates subscription lock-in and cloud privacy risks. Native on Windows, macOS, and Linux, it features multi-track timeline editing, bezier keyframes, LUT grading, and built-in subtitle generation. Crucially, Concat exposes native JSON-RPC and MCP APIs, enabling AI coding agents (Claude Code, Cursor) to automate clip cutting, transitions, voice-track alignment, and batch exports via code or natural language.

Concat (Open Source)Model Context Protocol (MCP)FFmpegPython 3.10+

Best for: Short-form video creators, media matrix operators, AI automation engineers, post-production editors, privacy-conscious enterprises

Full guide ↗

VIDEO · AI Video & Avatars

15 MIN

Build an Autonomous 24/7 Anime Live2D VTuber Streamer with Open-LLM-VTuber for Bilibili & YouTube

A viral open-source AI VTuber framework with 14k+ stars! Open-LLM-VTuber breathes autonomous life into Live2D and 3D avatars. Connects natively to real-time live chat streams across Bilibili, YouTube, and Twitch; features full-duplex hands-free conversational voice interaction with interruption handling; and hooks into local Ollama or DeepSeek to project vibrant anime personas for 24/7 autonomous virtual streaming.

Open-LLM-VTuber (Open Source)Live2D SDK / VTube StudioPython 3.10+DeepSeek / Ollama

Best for: VTuber creators, anime streamers, indie game devs, animation studios, companion bot builders

Full guide ↗

VIDEO · AI Video & Avatars

20 MIN

Synthesize Expressive Semi-Body Human Avatars with Gestures & Poses via Ant Group EchoMimicV2

Selected for CVPR 2025! EchoMimicV2 by Ant Group is a state-of-the-art semi-body human animation framework. Eliminating the awkward "talking head sticker" limitation, EchoMimicV2 drives coordinated facial expressions, shoulder dynamics, and full arm/hand gesturing directly from audio. Ideal for live commerce streamers pointing at products, accentuating value propositions, and engaging audiences naturally.

EchoMimicV2 (Open Source)PyTorchCUDA 12+Diffusers

Best for: E-commerce video creators, live selling teams, 3D visual animators, PR managers, online educators

Full guide ↗

VIDEO · AI Video & Avatars

15 MIN

Clone Photorealistic Digital Humans and Deploy On-Device Interactive Avatars with Open-Source DUIX

An industry-grade open-source AI avatar toolkit from Duix.com with 16k+ stars. Clones photorealistic digital humans from standard real-person video footage without expensive 3D motion capture rigs. Paired with Duix-Mobile SDK for Android/iOS/embedded screens, it runs completely offline on-device with sub-1.5s latency, eliminating recurring cloud GPU costs for kiosks and mobile apps.

Duix-Avatar (Open Source)Duix-MobileWindows / Android / iOSPyTorch

Best for: Trainers, video production agencies, exhibition kiosk integrators, marketing teams, corporate content creators

Full guide ↗

VIDEO · AI Video & Avatars

20 MIN

Deploy Open-Source LiveTalking for 24/7 Autonomous E-Commerce Live Streaming with RTMP & Bullet-Chat Answering

The leading open-source real-time streaming digital human engine with 10k+ stars! LiveTalking is engineered for 24/7 autonomous e-commerce live streaming on TikTok, Douyin, and Bilibili. Features RTMP/WebRTC and virtual camera feeds directly into OBS; motion orchestration plays subtle idle breathing and gestures during pauses to avoid static-screen penalties; and hooks LLMs to read live bullet-chats with natural voice interruptions.

LiveTalking (Open Source)Python 3.10+OBS StudioDeepSeek / Qwen API

Best for: E-commerce merchants, MCN agency managers, cross-border TikTok streamers, digital human integrators, private domain marketers

Full guide ↗

VIDEO · AI Video & Avatars

15 MIN

Cinematic 15s 2K Video Generation with Native Stereo Audio & Consistent Characters using MiniMax-H3 Director

Conquer mute videos and character inconsistency once and for all! MiniMax-H3 is an open-source full-modal generation system. Its breakthrough Ref2VA multi-modal reference mode accepts up to 9 images, 3 reference videos, and 3 audio stems simultaneously—locking character identity, wardrobe, and gestures. Generates 24FPS video with native 32kHz stereo audio, first-and-last keyframe steering, and 15-second 2K output via ComfyUI Director workflows.

MiniMax-H3 (Open Source)ComfyUI-MiniMaxH3-DirectorPythonMiniMax API / 本地推理引擎

Best for: Film directors, web drama creators, commercial cinematographers, 3D animators, content creators, game cinematic designers

Full guide ↗