One long video in. A month of posts out.
Four stages run after you hand over a video, and you are only involved in the first one. Here is what each of them actually does.
Transcribe
The first thing that happens is not editing — it is listening. Your video is transcribed to word-level timing, which is what makes everything downstream possible: captions that land on the syllable, cuts that never clip the front of a sentence, and a searchable record of every point made in the video. It also means the system works from what was said, not from visual motion.
Score
Every moment is then rated on how well it should perform as a short, using GPT-4o against the transcript. This is the stage that replaces the hour you would otherwise spend dragging a playhead looking for the good bit. You get the ranking too — each finished clip carries its score, so you know which three to post first instead of guessing.
Cut, reframe, caption
Winning moments are cut at sentence boundaries, reframed to 9:16 with the speaker tracked so they stay in frame when they move, and captioned word by word in whichever of the 22 styles you picked. Filler words and dead air are removed, B-roll can be dropped in where a clip needs something to look at, and everything renders at 1080p with no watermark on paid plans.
Publish
Finished clips go to the autoscheduler, which posts them to YouTube, TikTok, Instagram and X on a cadence you set once. Cross-posting one clip to four platforms is four posts, handled for you. Upload on Monday and the rest of the week goes out without you opening anything.
What you never have to do
Scrub a timeline hunting for a moment. Re-cut a clip to vertical by hand. Type captions. Cut out every “um”. Upload the same clip to four platforms one at a time. Remember to post on a Thursday.
What it is good at, and what it is not
It is built for long, spoken video — podcasts, interviews, talks, webinars, sermons, coaching calls, commentary. Because the scoring reads the transcript, talk-led content is where it performs best. Silent gameplay with no commentary is the honest exception: there is very little for the model to score, and it will return fewer clips.
Prefer the short version? The AI clip maker page covers what it does to each clip, and pricing explains how upload minutes are counted.