AI Shorts Creator

Long video to shorts

Long video to shorts, without the second edit

You already did the hard part when you recorded the full thing. This page is about the step most people skip: getting the short versions out while the episode is still fresh.

The 8-hour sleep rule is wrong

podcast · 34s

Why my first hire was a mistake

interview · 48s

Compound interest in 40 seconds

lesson · 40s

Track one number this quarter

keynote · 52s

Why long recordings rarely become shorts

Most creators have a folder of long recordings and a feed that goes quiet between uploads. The reason is not laziness. Scrubbing through forty minutes of footage looking for the two minutes that would work on their own is slow, boring, and happens at the end of the day when the energy for it has gone. Then there is the reframing: a widescreen recording has to become a tall one, and every clip needs captions because most people watch shorts with the sound off.

AI Shorts Creator does exactly that chore and nothing else. It is not a general editor and it does not generate footage. You upload a recording, it reads it, and it hands back a handful of finished vertical clips. The whole run usually takes a few minutes for a fifteen-minute source.

What happens to your upload, step by step

First the audio is pulled out of the file and transcribed with timestamps on every single word. That transcript is what everything else is built on: the picker reads it to choose moments, and the caption renderer uses the word timings so each caption lands exactly when the word is spoken.

Next, a language model reads the transcript the way a producer would. It looks for a hook in the first sentence, a complete thought in the middle, and a natural stopping point, and it avoids starting in the middle of a sentence. It proposes three to five windows between 20 and 60 seconds and writes a short title for each one so you know what the clip is about before you watch it.

Finally each window is cut from the original, reframed to 1080×1920, captioned, and given a thin progress bar along the bottom edge. The output is an ordinary MP4 with everything burned in, so it plays the same on every platform and in every app.

Framing choices for different kinds of footage

A single talking head is the easy case: centre crop keeps the speaker large and throws away the empty sides of the widescreen frame. For footage where the sides matter — a slide deck, a two-person interview, gameplay — switch to blur-pad. The full picture is kept and letterboxed in the middle of the vertical frame, and a blurred, slightly darkened copy of the same frame fills the space above and below so nothing looks empty.

Both modes are set once per job. If you are not sure, run centre crop first; the job is cheap enough to run again with the other setting while the upload is still on file.

Captions and the progress bar

Captions are rendered as short cards of three or four words in the lower third of the frame, large enough to read on a phone held at arm’s length. The default style sets the words in bold uppercase and lights up the word being spoken in a warm yellow. A quieter sentence-case style is available for content where shouting would feel wrong, and you can turn captions off entirely if your platform adds its own.

The progress bar is a small thing with a measurable effect: viewers can see that a clip is nearly over, which keeps them to the end instead of swiping at the first lull. It is drawn in the site’s cobalt blue, sits on the bottom edge, and never covers the picture.

What it costs and what you keep

A job is 40 credits regardless of how many shorts come out of it or how long the source is, as long as it is under fifteen minutes and five hundred megabytes. Every new account begins with two credits so you can see the interface, and the plans start at nine dollars a month for six jobs. Your upload is stored privately under a random name and deleted a day later; the finished shorts stay in your library and are yours to download, share, or delete.

Process

Three steps between a recording and a week of posts

  1. Step 1

    Upload the long version

    Drop in an MP4, MOV or WebM up to 15 minutes and 500 MB. It goes straight to private storage and is wiped 24 hours later.

  2. Step 2

    We listen to the whole thing

    The audio is transcribed with word-level timing, then a language model reads the transcript and marks the three to five stretches that stand on their own.

  3. Step 3

    Each moment becomes a short

    Every pick is cut at a sentence boundary, reframed to 9:16, captioned word by word and finished with a progress bar. You get MP4s, not a project file.

Options

Opinionated defaults, a few switches that matter

Two ways to go vertical

Centre crop keeps a single speaker large in frame. Blur-pad keeps the full widescreen picture, floating over a softened copy of itself, for slides and gameplay.

Captions that follow the voice

Pick bold uppercase with the spoken word lit up, a quieter sentence-case style, or none. Timing comes from the transcript, not a guess.

Lengths that fit the feed

Every short lands between 20 and 60 seconds, snapped to where sentences actually start and stop. Set a shorter ceiling if your audience scrolls fast.

Answers

Long video to shorts: common questions

How long can the source video be?
Up to fifteen minutes and five hundred megabytes, in MP4, MOV or WebM. Longer recordings should be split before uploading; most desktop players can export a range.
Does the picker ever choose badly?
Sometimes. It works from the words alone, so a visual moment with no commentary will be missed, and a dense technical section may be favoured over a lighter one. Each pick shows the sentence it was chosen for, and a re-run with a different clip count gives a new selection.
Who owns the shorts this makes?
You do, entirely. They are pieces of your own recording with captions over the top. The exact wording is in the terms of service. Copyright law is not the same everywhere, so for a specific commercial question a local lawyer is the right call.
Can I upload a video in another language?
Yes. The transcription step handles most widely spoken languages and the captions are rendered in the language that was spoken. Titles are written in the same language as the transcript.
Will the clip start mid-word?
No. Every window is snapped to word boundaries with a short lead-in, and the end is pushed to the end of the last word plus a small pause.

Also clip from

Pick the longest thing you published this month

Upload it, choose how many shorts you want, and compare what the picker chose with what you would have chosen yourself.

Upload a video