# Koma

> Self-hosted AI video tool that creates summaries, chapters, searchable transcripts, and key frames from videos or MP4 links. Exports Markdown, SRT, JSON, CSV, and TXT.

- **Category:** Video Processing
- **Latest version:** 0.1.0
- **Open source:** oss
- **Languages:** Typescript, CSS
- **Platforms:** Docker, Node.js
- **Tags:** CLI, Yt Dlp, Subtitles, Ffmpeg, Typescript, Transcription, Computer vision, Self Hosted, Speech To Text, Structured data, Video analysis, Github oauth, Ai summarization, Video Understanding, Keyframe Extraction, Export Markdown, Export Srt, Export JSON, Video Summarization

## Description

Koma is a self-hosted web application for AI-powered video analysis. Users upload video files or provide direct MP4 URLs, and the system uses configurable speech-to-text and vision AI providers to produce summaries, chapter markers, searchable transcripts, and key frame descriptions. Results are interactive — clicking a chapter, tag, or subtitle jumps to the corresponding moment in the video. Exports include Markdown, SRT subtitles, JSON, CSV, and plain text. A GitHub App enables private workspaces with account sign-in, while an optional admin console manages providers and all jobs. The stack requires Node.js 22.13+, bundles FFmpeg, defaults to SQLite storage with MySQL as a production option, and supports local or Aliyun OSS for media. Without provider API keys, Koma runs in mock mode for demonstration.

## Links

- [Homepage](https://github.com/yuxino-labs/koma)
- [Website](https://koma.yuxino.cn)
- [Documentation](https://github.com/yuxino-labs/koma/blob/main/docs/CONFIGURATION.md)
- [Administration Guide](https://github.com/yuxino-labs/koma/blob/main/docs/ADMIN.md)
- [HTTP API Reference](https://github.com/yuxino-labs/koma/blob/main/docs/API.md)
- [GitHub Repository](https://github.com/yuxino-labs/koma)

---
Source: https://selfhost.directory/project/koma