AI edits Reels on its own: agent cuts takes and adds subtitles

The author of the 'AI Marketing | Zhenya Korzhavin' channel describes a pipeline where a coding agent uses open-source tools video-use, HyperFrames, and add-zooms to automatically edit vertical videos from raw takes: it reads the ElevenLabs Scribe transcript with word-level timestamps, cuts pauses, selects takes, adds subtitles, and self-checks every cut.

Compiled by Sergey KostenchukPublished 2026-09-23Updated 2026-09-23

2026-09-23 Coding Anthropic

🛠 AI edits Reels on its own: agent cuts takes and adds subtitles.

The author of the 'AI Marketing | Zhenya Korzhavin' channel built a pipeline using video-use from browser-use and HyperFrames from HeyGen. The LLM doesn't watch frames, but reads the ElevenLabs Scribe transcript with word-level timestamps: it cuts pauses, selects takes, adds subtitles, and self-checks the cuts.

🌍 Conversational video editing is becoming a pipeline on top of coding agents: everything is assembled from open-source skills without SaaS subscriptions, and rendering is left to ffmpeg. The latest Hypit v0.2.13 breaks down someone else's viral video into an editable template — in the demo, cloning a 20-second video cost $1.15.

👤 You can try it today: clone video-use, install ffmpeg, insert your ElevenLabs key, and drop takes into a folder. But the quality of the cuts is unstable — it's wiser to edit the first videos by hand.

Source 1: https://github.com/browser-use/video-use Source 2: https://github.com/hypit-ai/hypit

Sources