Video Edit Instructions
January 9, 2026 ยท View on GitHub
Created: 9 January 2026
A test pipeline for converting video recordings into precise code editing instructions.
Purpose
This repository provides a workflow where:
- A user records a screencast or video (with or without annotations) describing desired edits to an application
- The recording is processed by a multimodal AI
- The AI converts the video content into precise editing instructions for an AI coding agent working in a codebase to implement those edits
Structure
screenshare/- Converted video files (H.265/HEVC MP4)audio/- Audio-only extracts (Opus OGG)
Workflow
- Record screencast describing desired changes
- Place original recording in repository
- Convert to space-efficient format
- Extract audio for transcription
- Process with multimodal AI to generate editing instructions