mcp-ffmpeg
<div align="center">
<img src="assets/banner.png" alt="MCP-FFmpeg Banner" width="600" />
# MCP-FFmpeg
**FFmpeg video & audio editing tools via Model Context Protocol (MCP)**
[](https://python.org)
[](LICENSE)
[](https://modelcontextprotocol.io)
[](https://ffmpeg.org)
*Give your AI assistant the power to edit video and audio — trim, transcode, overlay, compose, and more.*
[Quick Start](#quick-start) · [Tools](#available-tools-30) · [Configuration](#mcp-client-configuration) · [Architecture](#architecture)
</div>
---
## Why MCP-FFmpeg?
AI assistants are great at understanding what you want to do with media files, but they can't actually *do* it — until now. MCP-FFmpeg bridges the gap by exposing **30+ FFmpeg operations** as MCP tools that any compatible AI assistant can call directly.
```
You: "Trim this video from 00:30 to 02:00, add a fade-in, and convert to 720p"
AI: Calls trim_video → add_basic_transitions → set_video_resolution → Done!
```
## Architecture
<div align="center">
<img src="assets/architecture.png" alt="MCP-FFmpeg Architecture" width="600" />
</div>
```
┌─────────────────────┐ MCP Protocol ┌──────────────────────────┐
│ AI Assistants │◄───────────────────► │ mcp-ffmpeg Server │
│ │ JSON-RPC/stdio │ │
│ · Claude Code │ │ ┌────────┐ ┌────────┐ │
│ · Claude Desktop │ │ │ Video │ │ Audio │ │
│ · Cursor │ │ │ Tools │ │ Tools │ │
│ · Any MCP Client │ │ └────────┘ └────────┘ │
└─────────────────────┘ │ ┌────────┐ ┌────────┐ │
│ │Overlay │ │Compose │ │
│ │ Tools │ │ Tools │ │
│ └────────┘ └────────┘ │
└──────────┬─────────────┘
│
▼
┌──────────────────────┐
│ FFmpeg Engine │
└──────────────────────┘
```
## Quick Start
### Prerequisites
| Requirement | Version | Purpose |
|-------------|---------|---------|
| **Python** | 3.10+ | Runtime |
| **FFmpeg** | Any recent | Media processing engine |
| **uv** | Latest | Python package manager (recommended) |
### Install & Run
```bash
git clone https://github.com/kevinten-ai/mcp-ffmpeg.git
cd mcp-ffmpeg
uv sync
uv run python main.py
```
## MCP Client Configuration
### Claude Code
Add to your project's `.mcp.json`:
```json
{
"mcpServers": {
"ffmpeg-tools": {
"command": "uv",
"args": ["--directory", "/path/to/mcp-ffmpeg", "run", "python", "main.py"]
}
}
}
```
### Claude Desktop
Add to `claude_desktop_config.json`:
```json
{
"mcpServers": {
"ffmpeg-tools": {
"command": "uv",
"args": ["--directory", "/path/to/mcp-ffmpeg", "run", "python", "main.py"]
}
}
}
```
### Cursor / Other MCP Clients
Use the same server configuration — any client that supports MCP's stdio transport will work.
---
## Available Tools (30+)
### Video Processing
| Tool | Description |
|------|-------------|
| `trim_video` | Trim video to specific start/end times |
| `convert_video_format` | Convert between formats (mp4, mov, avi, etc.) |
| `convert_video_properties` | Batch convert format, resolution, codecs, bitrate, frame rate |
| `change_aspect_ratio` | Change aspect ratio with pad or crop mode |
| `set_video_resolution` | Scale to target resolution (e.g., 1920x1080) |
| `set_video_codec` | Change video codec (libx264, libx265, vp9) |
| `set_video_bitrate` | Set video bitrate (e.g., 2500k, 5M) |
| `set_video_frame_rate` | Adjust frame rate (24, 30, 60 fps) |
| `change_video_speed` | Speed up or slow down playback (0.25x to 4x+) |
### Audio Processing
| Tool | Description |
|------|-------------|
| `extract_audio_from_video` | Extract audio track from video file |
| `convert_audio_format` | Convert between audio formats (mp3, wav, aac) |
| `convert_audio_properties` | Batch convert format, bitrate, sample rate, channels |
| `set_audio_bitrate` | Set audio bitrate (128k, 192k, 320k) |
| `set_audio_sample_rate` | Set sample rate (44100, 48000 Hz) |
| `set_audio_channels` | Set mono (1) or stereo (2) |
### Video Audio Track
| Tool | Description |
|------|-------------|
| `set_video_audio_track_codec` | Change audio codec within video (aac, mp3) |
| `set_video_audio_track_bitrate` | Set audio bitrate within video |
| `set_video_audio_track_sample_rate` | Set audio sample rate within video |
| `set_video_audio_track_channels` | Set audio channels within video |
### Overlays & Subtitles
| Tool | Description |
|------|-------------|
| `add_text_overlay` | Add timed text overlays with full styling control |
| `add_image_overlay` | Add image watermarks/logos with position & opacity |
| `add_subtitles` | Burn SRT subtitles with customizable font styling |
### Composition & Editing
| Tool | Description |
|------|-------------|
| `concatenate_videos` | Join multiple videos with optional xfade transitions |
| `add_b_roll` | Insert B-roll clips as overlays at specific timestamps |
| `add_basic_transitions` | Add fade-in / fade-out effects |
| `remove_silence` | Detect and remove silent segments automatically |
### System
| Tool | Description |
|------|-------------|
| `health_check` | Verify server is running and responsive |
---
## Usage Examples
### Basic Video Editing
```
"Trim intro.mp4 from 10s to 2:30 and save as intro_trimmed.mp4"
→ trim_video(video_path="intro.mp4", output_video_path="intro_trimmed.mp4",
start_time="10", end_time="2:30")
```
### Format Conversion
```
"Convert recording.mov to mp4 at 720p with h265 codec"
→ convert_video_properties(input_video_path="recording.mov",
output_video_path="recording.mp4",
target_format="mp4", resolution="720",
video_codec="libx265")
```
### Adding Subtitles
```
"Burn subtitles.srt onto my video with large white text"
→ add_subtitles(video_path="video.mp4", srt_file_path="subtitles.srt",
output_video_path="video_subtitled.mp4",
font_style={"font_size": 28, "font_color": "&HFFFFFF"})
```
### Video Composition
```
"Join clip1.mp4 and clip2.mp4 with a dissolve transition"
→ concatenate_videos(video_paths=["clip1.mp4", "clip2.mp4"],
output_video_path="final.mp4",
transition_effect="dissolve",
transition_duration=1.0)
```
---
## Supported Transitions
The `concatenate_videos` tool supports **30+ xfade transitions** for joining two videos:
| Category | Transitions |
|----------|-------------|
| **Fade** | `fade`, `fadeblack`, `fadewhite`, `fadegrays`, `dissolve`, `distance` |
| **Wipe** | `wipeleft`, `wiperight`, `wipeup`, `wipedown` |
| **Slide** | `slideleft`, `slideright`, `slideup`, `slidedown` |
| **Smooth** | `smoothleft`, `smoothright`, `smoothup`, `smoothdown` |
| **Shape** | `circlecrop`, `rectcrop`, `circleopen`, `circleclose` |
| **Split** | `vertopen`, `vertclose`, `horzopen`, `horzclose` |
| **Diagonal** | `diagtl`, `diagtr`, `diagbl`, `diagbr` |
| **Slice** | `hlslice`, `hrslice`, `vuslice`, `vdslice` |
| **Other** | `pixelize`, `radial`, `hblur` |
## Project Structure
```
mcp-ffmpeg/
├── main.py # Entry point
├── pyproject.toml # Project config & dependencies
├── src/ffmpeg_tools/
│ ├── __init__.py
│ ├── server.py # MCP server setup & tool registration
│ ├── utils.py # Shared utilities (probe, clip prep, etc.)
│ └── tools/
│ ├── audio.py # Audio processing tools (6 tools)
│ ├── video.py # Video & video-audio tools (13 tools)
│ ├── overlay.py # Text/image/subtitle overlay tools (3 tools)
│ └── compose.py # Concatenation, B-roll, transitions (4 tools)
├── tests/
│ ├── test_video_functions.py # Test suite
│ └── sample_files/ # Test media files
└── assets/ # Images for documentation
```
## Testing
```bash
uv run pytest tests/ -v
```
## Contributing
Contributions are welcome! Here are some ways to help:
- Add new FFmpeg tools (e.g., video stabilization, noise reduction)
- Improve error handling and validation
- Add more test coverage
- Improve documentation
## License
MIT License — see [LICENSE](LICENSE) for details.
## Acknowledgments
Based on [video-audio-mcp](https://github.com/misbahsy/video-audio-mcp) by misbahsy.
---
<div align="center">
**Built with [FastMCP](https://github.com/jlowin/fastmcp) and [FFmpeg](https://ffmpeg.org)**
</div>
TDQS
Scored across 27 tools
Several tools have overlapping purposes, such as the single-property setters (set_audio_sample_rate, set_audio_bitrate, etc.) versus the all-in-one converters (convert_audio_properties, convert_video_properties). The descriptions help clarify usage, but an agent could easily select the wrong tool when simple format conversion is needed because both convert_audio_format and convert_audio_properties perform that task.
Tool names mostly follow a verb_noun pattern (set_*, convert_*, add_*), with clear prefixes for different actions. There are minor deviations like 'change_aspect_ratio' and 'change_video_speed' instead of 'set_*', and 'health_check' does not fit the pattern, but overall the naming is predictable and readable.
At 27 tools, the server is on the higher end, but the breadth of FFmpeg operations justifies many of them. However, the count is inflated by redundant single-property setters that could be consolidated into the existing property conversion tools, making the set feel heavier than necessary.
The server covers a wide range of common media operations: audio/video format conversion, property adjustments, trimming, speed changes, overlays, subtitles, concatenation, and silence removal. Minor gaps exist, such as video rotation or cropping to specific dimensions, but the core workflows are well represented.