How to Turn a Long Video Into Shorts With AI, Step by Step

You turn a long video into shorts by uploading the file to Clip Studio, letting it transcribe the whole recording, choose the strongest moments, cut them, reframe each one to vertical 9:16 and burn in word-timed captions. Uploads run from 1 to 90 minutes and up to 1 GB, and each job returns between 1 and 5 clips. It costs 2 credits per source minute, with a minimum of 4 credits.

Updated October 2026.

What Clip Studio does, in order

Clip Studio is a cutter, not a generator. It does not invent footage; it finds the parts of your recording that already work as a short and packages them. The order matters, because each step depends on the one before it.

1. Transcribe. The audio is extracted and transcribed with word-level timing. Every later decision is based on that transcript.
2. Pick moments. The tool looks for self-contained stretches of speech and gives each candidate a title, a hook line and a score.
3. Cut. The chosen windows are trimmed out of the source and capped at 75 seconds, so no clip runs past the short-form limit.
4. Reframe. The horizontal frame is converted to vertical 9:16, using follow-face, blur fill or a two-speaker stack.
5. Caption. Word-timed captions are burned into the picture in the style you chose.

The output is a set of finished vertical MP4s that you can download and post. Nothing about the process needs a timeline or manual cutting.

Limits before you start

Three limits decide whether a file will be accepted, and all three are checked early.

• Upload only. You pick a file from your device. There is no URL field, so a YouTube or podcast link cannot be used as the source.
• 1 to 90 minutes. A recording longer than 90 minutes is refused. If the episode is longer, split it into parts first.
• Up to 1 GB. A bigger file needs compressing before upload.

The file picker accepts video files only. An audio-only podcast cannot be uploaded, because the tool needs a video track to reframe and caption. If that is what you have, see audio to video for a route that starts from audio.

Choose how many clips and run it

Set the number of clips from 1 to 5. This is a target, not a guarantee of quality: asking for five returns five of the best candidates the transcript offers, and on a recording with little self-contained speech, the fifth may be weaker than the first. Start with three on a first episode, judge the output, then adjust.

Below the clip count are the caption style and the vertical framing controls. Set them before you run, because they apply to every clip in the batch. Then submit. The button shows the credit cost before you confirm, so there is no surprise charge. The job takes a few minutes and reports progress while it works.

A 60-minute source is 120 credits by the 2-credits-per-minute rule, which is the number to expect for a full episode. Shorter recordings scale down from there, always with the 4-credit minimum.

The post-production options

These options change how the finished clips look. Some are on by default; check them before a batch.

• Vertical framing. Smart picks between the three layouts automatically. Follow face crops to whoever is speaking. Blur fill keeps the whole horizontal frame and puts it on a blurred vertical background. Two speakers stacks two people top and bottom. Smart is the sensible default for a first run.
• Caption style. The same preset list used by AI Captions, including bold word-by-word looks and calmer full-line styles. For a talking clip, pick one that fits under the platform interface.
• Hook title. Adds the clip’s title as on-screen text for the first three seconds, which gives a scrolling viewer a reason to stop.
• Zoom punch-ins. Pushes in slightly on key words, which adds visual movement to a talking head.
• Cut silences. Jump-cuts the pauses in speech. It shortens a clip and tightens the pace, but it does not remove filler words such as “um”.

Each option is independent. A news-style interview might use Follow face, a calm caption style and silence cuts. A two-host podcast might use Two speakers, a louder caption and no punch-ins, to keep the conversation readable.

Review and discard weak clips before posting

The results grid shows every clip with its title, hook line and score. Treat the score as a ranking of the transcript, not a promise about performance. Watch each clip from start to finish and ask three questions: does the first second make sense without context, does the clip end on a complete thought, and is the caption readable at phone size?

Discard anything that fails the first or second question. A clip that starts mid-sentence or stops before the point lands will not recover in the edit, and a weak clip posted alongside good ones dilutes the set. It is better to publish two strong clips from a batch of five than all five. If a moment is almost right, the fix is usually a tighter source edit, not a longer caption.

This cut-a-long-recording route is different from generating a short from scratch. If you want a new vertical video rather than a cut of an existing one, see how to make a YouTube Short with AI. For a wider look at editing options, the AI video editors round-up and the Opus Pro alternatives list cover the alternatives.

Cost

Clip Studio charges 2 credits per source minute, with a 4-credit minimum. The price is shown before the job runs, and a failed job is refunded. The charge covers the whole batch: one job on a 60-minute recording produces up to five clips for the same 120 credits, so the cost per finished clip falls as the batch size rises.

One detail to watch: the upload is charged on the duration your browser reports, and the server re-probes the file when it downloads it. If the real duration is longer than the reported one, the difference is charged before processing continues, and the job stops with a top-up message if the balance cannot cover it. That is why a file that claims 58 minutes and is really 62 minutes can cost more than the on-screen estimate.

Credit packs are one-time purchases that never expire, which suits a batch workflow: buy credits when you have a podcast run to cut, and keep the remainder for the next one. Prices are on the pricing page.

Where it stops, and the sizes to post

Clip Studio is built for spoken content that is already in a video file. It works best on interviews, podcasts, webinars and talking-head recordings, where the transcript can identify a complete idea. It is weaker on music videos, action footage and anything without clear speech, because there is no transcript to rank.

All output is vertical 9:16 with captions burned in. If a platform wants a different shape, you are editing a finished vertical clip, which means cropping again. Our guide to the best aspect ratio for YouTube Shorts explains why 9:16 is the default for short-form feeds.

For a single clip that needs a different caption look or a re-render, AI Captions takes any video up to three minutes and gives you three renders per charge. That pair covers most of the pipeline: clip the long recording here, then re-style a chosen short there.

Working from a podcast episode? The podcast to short clips guide covers the parts that differ: picking moments from conversation, framing two speakers, and what to do when the episode exists only as audio.

Frequently asked questions

Can I paste a YouTube link?

No. Clip Studio accepts an uploaded video file only. There is no URL input, so download the file first if it is yours.

How long can the source video be?

1 to 90 minutes, up to 1 GB. The file picker accepts video files only, so audio-only recordings cannot be uploaded.

How many clips does one job return?

Between 1 and 5, depending on the number you choose. Three is a good starting point for a first episode.

How long is each clip?

Clips are cut to a maximum of 75 seconds, with a minimum of 10 seconds. The tool chooses the window, then caps it at 75 seconds.

What does it cost?

2 credits per source minute with a 4-credit minimum. A 60-minute source is 120 credits, shown before you run it. Failed jobs are refunded.

Open Clip Studio, upload your recording, set the clip count and the vertical framing, and run the batch. New accounts start with 10 free credits and no card required, enough to cover a short source and see how the tool reads your own speech before you put a full episode through it.