obinnanwachukwu1/computer-use-capture — explained in plain English
Analysis updated 2026-05-18
Automatically record and edit a polished demo video while an AI agent completes a task on screen.
Ask Codex in plain language to demo a feature in an app and receive a finished video link.
Re-render a completed demo with a corrected cursor, background, or motion setting.
Discard or cancel recordings and renders that are no longer needed.
| obinnanwachukwu1/computer-use-capture | aiduckman/claudeusage_latest_may2026 | arnabau/thermalpulse | |
|---|---|---|---|
| Stars | 0 | 0 | 0 |
| Language | Swift | Swift | Swift |
| Setup difficulty | hard | easy | moderate |
| Complexity | 4/5 | 2/5 | 3/5 |
| Audience | developer | vibe coder | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Apple Silicon macOS, Codex with Computer Use, Node.js, and granting Screen Recording and Accessibility permissions before recording works.
Computer Use Capture is a Mac only tool that turns a Codex Computer Use session, where an AI agent controls your computer to complete a task, into a polished, professional looking product demo video. While the AI is clicking around and performing a task on screen, this tool records it, then rebuilds a clean version of what actually happened and renders it as a smooth 60 frames per second video, complete with a native Mac cursor, gentle camera movement, motion blur, and the boring waiting time cut out. It does not replace or interfere with how the AI performs its task. Instead it runs as a separate local helper program that Codex can call on to start recording, stop recording, and manage the resulting video. Because it runs as one shared background service, a recording started during one task stays visible even if Codex restarts its own connection partway through. To use it, a developer installs it through a single command that adds it to Codex, then simply asks Codex, in plain language, to record a demo of some task in an app like Safari. Codex is expected to check that recording is possible, start the recording once the right screen is ready, perform the task normally, stop the recording, wait for the video to finish rendering, and then hand back a link to the finished file. Several small commands exist under the hood for checking readiness, starting and stopping recordings, checking on a render's progress, requesting a specific correction to a finished render, canceling a render, or permanently deleting a recording. The tool is careful about only showing what it can verify actually happened. If it cannot confirm exactly where a click landed, it renders that moment without a visible cursor rather than guessing its position. Recordings are limited to a single selected app window, and moving or resizing that window mid recording is not supported. There is no audio capture, only video. Everything, including capturing, processing, and rendering, happens locally on the user's own Mac, with recordings and their related files stored in a folder under the user's Application Support directory until the user deletes them. It requires Apple Silicon, a recent version of macOS, Node.js, and screen recording and accessibility permissions to work.
A macOS tool that records an AI agent's Computer Use session and turns it into a polished, edited product demo video automatically.
Mainly Swift. The stack also includes Swift, Node.js, ScreenCaptureKit.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.