Stream clipping kit
The Clip Pipeline
Clip channels are made by hand: scrub the VOD, find the moment, crop it to vertical, hope the webcam is not cut off, write a caption, upload, repeat. It is an hour of work for forty seconds of video, and the work is identical every time.
This kit is that hour, automated. It reads a Twitch channel's top clips through the official API, or takes a YouTube URL, downloads the footage, finds the streamer's webcam in the frame, and renders a vertical short with the facecam on top and a second video underneath. It writes the caption and the hashtags, queues the result, and posts it. Point it at a different creator by changing one line.
What it does
Finds the facecam, so the crop keeps the face
A stream is a wide frame with gameplay filling most of it and a small webcam box in a corner. Crop that to vertical without thinking and you lose the streamer's face or keep it half sliced. The pipeline samples frames and runs a bundled face detector, pooling detections across frames rather than trusting any single one, so a moment where the streamer looks away does not move the crop. When there is no facecam to find, it falls back to a centre-weighted crop instead of guessing.
Two sources, one path
Twitch clips come through the official API, ranked and deduplicated. A YouTube video works too, and needs no Twitch credentials at all. Give it a start and end time and only that range is downloaded, so cutting forty seconds out of a two-hour upload fetches roughly forty seconds.
The split screen, built for you
The top half carries the clip, framed on the detected facecam. The bottom half carries a slice of a background video, taken from a random position each time so consecutive posts do not show the same footage. Swap the pool for your own footage in one list.
Captions, hashtags, and a queue
Every render gets a title, a description, and a hashtag set written for it, layered from your community tags, the clip's game, the content type, and broad reach tags. Finished videos go into a queue, and the included publisher posts from it.
How you use it
- Install the pipeline and its dependencies (Python, ffmpeg, and yt-dlp, all free).
- Add a free Twitch application's two credentials, or skip that entirely and use the YouTube route.
- Set the channel or channels you want clips from. That is one line.
- Run it on a single clip first and watch the file appear.
- Connect the account you will post to, once, in a browser.
- Put it on a schedule and let the queue drain at a measured pace.
Plan on about half an hour end to end. You do not need to write code, and you do not need to know ffmpeg.
Before you buy
The safest use of this kit is your own stream. The next safest is a creator who has given you written permission, and many streamers actively welcome clip channels. Re-uploading a creator's clips without permission is a copyright matter regardless of how the tooling works, and platforms separately demote or demonetise channels built on content you did not make. The guide has a full section on this. The kit gives you the mechanism; it does not give you the rights, and it cannot tell you what is true for your situation.
Who it is not for
If you want a hosted service with a dashboard, this is not that. It runs on your machine and you own every part of it. Delivered as-is: platform behaviour, moderation outcomes, API availability, and third-party pricing are outside anyone's control, and nothing here is legal advice.