If you searched for motion maxing github, you probably want the actual repository, not another explainer. So here it is up front. The open-source tool that does this is called OPEN VIDEO EDIT, and it installs as a set of Claude Code skills that turn a raw talking-head clip into an animated short. You clone it, run it locally, and ship a motion video in about the time it takes to render one file.
The repo is OPEN VIDEO EDIT (github.com/kevinbadi/open-edits). It renders with Python and ffmpeg on your own machine, so nothing leaves your laptop unless you move the file. That is the whole pitch. This post covers what the repo does, how to vet it before you install it, who it suits, and where the repo stops and a publishing system starts.
raw talking-head clip
|
v
[ OPEN VIDEO EDIT repo: Claude Code skills ]
|
+-- split-animated-talking-head -> vertical short
+-- longform-animated-talking-head -> long video
+-- fireship-style-edit / -short -> fast cuts
+-- split-animated-short
+-- reverse-engineer -> match an existing edit
|
v
Python + ffmpeg render, local file
|
v
Creator OS -> schedule to IG, TikTok, YouTube, X, LinkedIn
What Motion Maxing on GitHub Actually Refers To
Motion maxing is a specific editing style. A speaker is cut out of the background, held at the center or offset in the frame, and decorated with moving parts: animated captions, logo cards, cutout faces floating beside the speaker, GIF overlays, b-roll cards that slide in, and counters that tick up while the audio plays. The goal is that a viewer who scrolls past for one second still sees movement.
The repo that produces this on a talking-head clip is OPEN VIDEO EDIT. It is not a Premiere template or a CapCut preset. It is a set of Claude Code skills you invoke from the terminal. The repo lists six skills: split-animated-talking-head, longform-animated-talking-head, fireship-style-edit, fireship-style-short, split-animated-short, and reverse-engineer.
The names tell you the job. Split-animated versions do the cutout treatment on a short or long source. Fireship-style versions use the fast, stacked, high-cut-density look. Reverse-engineer works the other way: you hand it a finished video you like, and the skill reads its structure so you can rebuild a similar rhythm on your own footage.
Every skill runs locally. Python and ffmpeg do the render. You keep the source clip, the output, and the project files. If you want the wider picture of how short, layered edits map onto platform rules, the motion maxing overview covers the style itself.
Motion Maxing vs Motionmaxxing: Clearing Up the Spelling
You will see three spellings in search: motion maxing, motionmaxxing, and motion-maxxing. They point at the same thing. The doubled-x form follows the online habit of stretching a word for emphasis. The repo name does not use any of them, and that is the part that matters.
So the practical advice is simple. Do not search GitHub for “motion maxing.” Search for “open video edit,” or go straight to the URL. A style name is a fragile search term: it is short, common in unrelated motion graphics contexts, and easy to mistype. The repo name is the stable handle.
One more naming note. Motion maxing is the editing style. Creator OS is the product that publishes the output. They are different things, and mixing them up leads people to expect the repo to schedule posts. It does not. It renders files.
How to Find and Vet the OPEN VIDEO EDIT Repo
Vetting an open-source editing repo takes five minutes if you check the right things. Do it before you install anything, because these skills call ffmpeg and write files to your disk.
- Open the repo page and read the README first. You want the skill list, the dependencies, and a stated render step. If the README only has a GIF and no setup, keep looking.
- Check the commit history. Look at the last few commits. Recent commits with real messages tell you the author is still working on it.
- Read the skills themselves. Claude Code skills are readable files. Open one and read what commands it runs. You should see ffmpeg filters and Python, not downloads from unknown hosts.
- Confirm the dependency list. You need Python and ffmpeg installed locally. No account, no hosted render queue, no API key to start.
- Run it on a throwaway clip. Do a 20-second test with a file you do not care about before you feed it an hour of client footage.
That last point matters. Render times scale with length, and a mistake in a long video costs you a full render cycle. Test short.
Who the Motion Maxing Workflow Is Built For
This workflow fits people who talk to a camera and want the edit handled. If you record yourself once and want a vertical short out the other side without dragging keyframes for an hour, it fits. If you run a podcast and want clips that carry motion, it fits, and there is a separate walkthrough on motion maxxing for podcasters.
It also fits developers and technical creators. You are already comfortable in a terminal, so a skill that runs Python is not a barrier. It is faster than a GUI for repeatable output. If you post about code, the Codex workflow post shows the same pattern for the writing side.
It does not fit everyone. If you need frame-level craft on every shot, you still want a human editor. If you want a hosted, drag and drop editor with a browser timeline, the repo is the wrong tool. And if you only post stills, none of this applies.
What You Get Out of a Repo-to-Video Pipeline
You get a repeatable command. Here is the shape of a real session, using the split short skill on a phone shot.
cd ~/clips
# source: 4 minute talking-head, 4K, one take
# target: 45 second vertical short
# in Claude Code, inside the repo checkout:
# "use split-animated-short on ./raw/take-04.mp4,
# cut from 00:12 to 00:57, keep my face in the
# upper third, captions bold white, logo card at
# 00:03 and 00:38"
# the skill writes the project files and runs
# python + ffmpeg locally
# output: ./out/take-04-short.mp4What lands in that file: your cutout on a styled background, animated captions timed to speech, your logo card, and whatever b-roll or counter you asked for. Swap split-animated-short for split-animated-talking-head and you get the longer treatment on the same source. Swap in a fireship-style skill when you want density instead of a calm frame.
The output is a local MP4. That is a good thing. You can watch it before it goes anywhere, fix a caption, and re-render. Nothing publishes itself.
When the file is right, the publish step is where a second system helps. You can post the same file to Instagram, TikTok, YouTube, X, LinkedIn, Facebook, and Threads from Creator OS, and the platform rules are handled for you. A vertical short uploaded through Creator OS becomes a YouTube Short automatically. Reels get their cover and caption rules applied. You can send the same render to several platforms in one post.
For a fuller run through, watch the walkthroughs on the KevBuildsApps YouTube channel, and the launch video covers the tooling end to end.
Where the Repo Stops and Creator OS Starts
The repo renders. It does not schedule, publish, reply to comments, or read analytics. That is the clean line. If you stop at the repo, you have a folder of MP4s and a manual upload routine.
Creator OS covers the rest, and it does it from the same terminal or from an AI app. Three surfaces matter here.
- The CLI.
@creatoros/cligives you commands for posts, media, analytics, inbox, automations, Skool, blog, and ads. You can install the Claude Code skills withnpx @creatoros/cli@latest init, which adds 14 of them, including post-everywhere, post-longform, and schedule-posts. Runcreatoros synclater and updates land as .new files so your edits are not overwritten. - The hosted MCP server. At https://mcp.creatoros.ca/mcp, with nothing to install. You paste the URL into Claude, ChatGPT, Cursor, Windsurf, VS Code, or add it to Claude Code, then sign in and pick a workspace. Read only or read and write. Tool names include create_post, upload_media_from_url, get_best_time_to_post, reply_to_comment, and get_post_analytics. Destructive tools are flagged so the AI asks first. Setup lives at the MCP docs.
- The open-source agent harness. Social Agents (github.com/kevinbadi/social-agents) runs your accounts on Creator OS. Start it with
npm start creatoros social-agents. It interviews you about your brand, then posts, replies to comments and DMs, runs automations, and reports to a local dashboard. You bring your own model. The Social Agents walkthrough goes deeper.
One upload detail worth knowing. POST /v1/media uploads a file once and returns a med_ id you can reuse on every platform. So you render once, upload once, and reference the same media id in the post. The base URL and full request shapes are in the Creator OS docs.
Common Setup Mistakes and How to Avoid Them
Searching for the style name instead of the repo
You want OPEN VIDEO EDIT. Search that name or open github.com/kevinbadi/open-edits directly. A style name returns noise and slows you down.
Skipping the ffmpeg check
Run ffmpeg -version before your first skill invocation. A missing or broken ffmpeg is the most common reason a render fails silently.
Rendering the whole source first
Test on 20 seconds. Long renders are slow, and you want the framing and caption style right before you commit an hour of footage.
Assuming the repo will publish
It will not. Check your accounts are connected before you plan a launch. creatoros accounts:list and creatoros accounts:health tell you fast. If you post to a network that is not connected, the API returns it in missing_platforms and the rest still go out, which can look like a partial success.
Stretching one aspect ratio across every platform
Render the vertical for Reels and Shorts, then check what each network accepts. creatoros validate:post-length --text "your caption" catches caption overflows before you schedule.
Letting an agent write without boundaries
Connect read only when you are exploring. Writes get refused from read only connections, which is exactly what you want while you test a new prompt.
Get Started With Creator OS
Two steps, in order. First, clone the repo and render one short from a clip you already have. You will know within an hour whether the style matches your channel. Second, connect your accounts so the render has somewhere to go.
The Creator plan is $14.99/month or $59.99/year, with 11 platforms plus Skool, one account on each. API keys, the MCP server, the CLI, and the agent skills are included on every plan, so the free step is just the repo. Start at sign up, then read the Creator OS explainer if you want the full map of what the platform does.
Render locally, publish centrally. That is the whole pipeline.
Get started for $0
Free trial for new subscribers. Cancel anytime.