yuxino-labs/koma
Self-hosted AI video analysis: summaries, chapters, keyframes, transcripts and SRT / JSON export. AI 视频总结、字幕提取与结构化分析。
What it solves
Koma is a self-hosted AI video analysis tool that transforms raw video files into structured data. It solves the problem of manually summarizing videos, creating chapters, and creating searchable transcripts, which is typically a time-consuming process for lessons, interviews, and meetings.
How it works
Users upload a video or provide a direct MP4 URL. The system uses an ASR (Automatic Speech Recognition) provider to transcribe the subtitles and a vision provider to analyze key frames and speech context. This combined analysis generates summaries, chapters, tags, and frame descriptions. Users can then jump to specific moments in the video based on these AI-generated results.
Who it’s for
It is designed for people who need to analyze video content for notes, spreadsheets, or research, such as students, researchers, and professionals conducting product analysis or meeting reviews.
Highlights
- Multimodal Analysis: Combines speech and key frames to generate summaries and structured data.
- Detailed Export: Supports exporting results as JSON, CSV, Markdown, SRT, and TXT.
- Subtitle Search: Provides searchable subtitles with direct links back to the video source.
- Self-Hosted: Offers a private workspace with GitHub sign-in and administrative controls for managing API keys and jobs.
- CLI Support: Includes a command-line interface for processing videos without a web interface.
Related
- Project
- Project
- Project
- Project