AV and streaming engineering for live events in Japan
I'm Shogo Kajiyama, a video, audio and streaming engineer with SEVENTHWELL, a small production company that runs technical operations for live events, broadcasts, academic conferences and expos. I work from Japan and write custom software for the field: Electron and PWA timers, multitrack streaming apps, and a Barco E2 position calculator. Our stack is Claude and Claude Code.
On the open source side, I maintain ffmpeg-skill, a local video editor for coding agents. It has 42 FFmpeg tools, a machine-readable contract and an MCP server, and works with Claude Code, Cursor and Codex. It runs with no cloud and no API keys. I'm still working out how to bring agent-driven editing into real event production.
What I'd love from this community: an intro to anyone using AI agents in live production or media workflows, or feedback on ffmpeg-skill.






Comments1
Welcome to Lounge, Shogo. Running ffmpeg-skill fully local with a machine-readable contract is a sensible fit for event work where venue networks are unreliable. Where do you think agent editing could be trusted first: post-event highlight cuts, or something closer to live, like clip extraction during a show?
1 comment
Welcome to Lounge, Shogo. Running ffmpeg-skill fully local with a machine-readable contract is a sensible fit for event work where venue networks are unreliable. Where do you think agent editing could be trusted first: post-event highlight cuts, or something closer to live, like clip extraction during a show?