What is auto-motion?

auto-motion is a workflow template that turns a subtitle file into a finished vertical video with animated motion graphics, using two separate AI coding tools working together. You provide a transcription.srt subtitle file, and the project's instructions tell OpenAI's Codex command line tool how to split that subtitle into a series of scenes based on meaning, so that the scenes together cover the whole timeline without gaps. For each scene, Codex calls Claude Code and gives it that scene's text, its duration, and where to save the output. Claude Code then writes an HTML animation for that scene using a set of pre-built animation building blocks called HyperFrames, and renders it out as a silent video file at 1080 by 1440 resolution and 30 frames per second. Codex repeats this one scene at a time rather than running scenes in parallel, and once every scene's video exists, it uses FFmpeg to stitch them together into a single final.mp4 file with the original subtitles reapplied on top. Each scene's run produces its own log files recording the raw output, any errors, and a more readable progress log, which is useful for figuring out what went wrong if a particular scene fails to render correctly. The project also includes an automated end to end test that copies in a sample subtitle file, runs the whole pipeline, and then checks that the resulting video has the right resolution, frame rate, total length, and no audio track. To run this yourself you need the Codex CLI and Claude Code both installed and signed in, Node.js version 22 or newer, FFmpeg and its ffprobe companion tool, the jq command line JSON tool, and an internet connection so Claude Code can search for assets or install anything it needs while building each scene's animation.