A Shorts maker built around the opening three seconds
A Short lives or dies on what a thumb sees before it swipes. This scores that opening on its own, without the rest of the video to lean on, tracks whoever is talking into a full vertical frame, and times every caption to the word — then waits for you to approve the cut.
What it takes to create Shorts from a YouTube video
A clip and a Short are cut by the same engine, and they are not the same request. A Short is watched full-screen on a phone, with a thumb one flick from leaving — three things follow from that, and each has a mechanism behind it below.
An opening that survives a thumb
The first line has to stand on its own with nothing before it, because on a Short nothing ever comes before it.
A frame with nothing wasted at the edges
Vertical, full height, and pointed at whoever is actually talking — not a centre crop that guesses and is wrong half the time two people are on camera.
Captions sized for arm’s length
Read at a glance on a small screen, timed to the word rather than the sentence, so the eye never gets ahead of the voice.
Graded by a reader who has seen nothing else
A stranger reads the opening before you do. Every candidate goes through a cold read: the clip’s own lines, no transcript, no title, no timecodes. If the first line reaches for something the Short never shows — a "there", a "this", a name never introduced — that clip is sent back to be cut again, because the one thing a Short cannot afford is a viewer who needed the video before it to make sense of the first line.
The throat-clearing is found and skipped. A model reads the start of the recording and finds where the actual content begins, rather than assuming the first sixty seconds is intro. Nothing is ranked from inside it.
Hook is one of four scores you can see. Alongside flow, value and travel, hook measures specifically how much of the first three seconds earns the next three — a number you can act on, not a single overall score standing in for four different questions.
How a moment is chosen, in fullFull height, and it follows whoever is talking
A Short fills the whole screen, so a fixed centre crop is wrong the moment two people are in frame or the speaker moves. This tracks the voice instead.
Full 9:16 by default
Edge to edge, no bands. TalkNet-ASD reads who is actually speaking and the crop follows them across the clip.
Boxed 1:1, when one speaker earns it
A vision pass over the footage — not the transcript — decides when a single speaker carried by their own words is better served boxed, with room freed up for text.
A two-shot, held rather than picked over
When two people are far enough apart in frame, the crop stacks them for that stretch instead of cutting one of them out to keep a single face centred.
Timed to the word, sized for a screen held at arm’s length
Karaoke, word by word. Each word highlights as it is spoken, timed directly off the transcript rather than an estimated cadence — the treatment built for a full-frame Short, and the one every shape can carry.
The look is yours; the position is not. Font, size, colour, outline, shadow and case are all editable, and a look you build can be saved as a named style and reused on the next Short. Where captions sit in the frame is fixed, and there is no free-floating caption block or logo layer to add.
A vertical clip with text behind the speaker is not offered. That treatment is built against the boxed 1:1 geometry and is genuinely unbuilt on a full-frame crop, so a Short renders with karaoke — the options panel sets this rather than a render failing partway through.
Both caption treatments, and where each one stopsWhat you get out, and what you do with it
The Short is the deliverable. Everything after it is a decision you make, not one this makes for you.
An MP4, ready to upload
The finished Short lands in your library and downloads in one click — a file on your machine before it is anything on a platform.
A calendar that remembers the date
Plan a Short onto a day and mark it published once it is live, with the link. It records the plan; it does not upload anything for you.
A cut you can still take elsewhere
A clip can also leave as an XML timeline or CMX3600 EDL for Premiere or Resolve — the cut, not the look, if you want to finish it somewhere else.
Paste the address — you never download your own video to upload it again
A YouTube address is enough. A watch page, a Short, a live URL or an embed — youtube.com or youtu.be — is read on our side without the video being stored: the transcription takes an audio-only stream, and each render afterwards fetches only the seconds of its own window, about six megabytes rather than six hundred. What it trades for that is a dependency an upload does not have — the video has to still be at that address when you come back to it.
You confirm the video is yours to use before the import starts; nothing is assumed. An upload — mp4, mov, mkv or webm, up to 8 GB — skips that step entirely because nothing is being fetched on your behalf.
The full intake, in six stepsThe free trial cuts full Shorts, not a watered-down preview
Thirty minutes of video, once, with no card — every shape and caption treatment included, and as many Shorts out of it as you keep. Trial clips carry a small mark in the corner and are kept for seven days; no paid plan marks what it exports.
Questions people ask before making a Short
Is this an AI YouTube Shorts generator?
Yes, in the sense that is checkable rather than a badge: the recording is transcribed, ranked in parallel windows, and a cold read grades whether each opening stands on its own before anything is offered to you. What is not automated is the decision to render — nothing is cut until you approve a specific candidate.
How do I create a Short from a YouTube video here?
Paste the watch address and confirm the video is yours to use, or upload the file if it never went up in the first place. The whole recording is read once, every passage worth cutting comes back ranked with its own hook, flow, value and travel scores, and you pick which ones become Shorts.
Does it post the Short to YouTube for me?
No. There is no connection to YouTube for publishing — reading a video from it and posting one back to it are different things, and only the first exists here. The finished MP4 sits in your library, and you upload it yourself.
Will the captions actually be readable on a phone?
They are timed word by word off the transcript rather than the sentence, so the highlight never gets ahead of the voice. Font, size, colour, outline and shadow are all editable and can be saved as a style you reuse — what you cannot change is where the caption sits in the frame, which is fixed.
Can I get the boxed square look instead of full-frame vertical?
Yes, when a vision pass over the footage finds a single speaker carried by their own words — that is when the boxed frame gets chosen, freeing up room for text. Full 9:16 is the default and the shape used whenever the frame needs to follow more than one person.
How long is a Short allowed to be?
Between twelve and a hundred seconds. A cut that comes back outside that band is made again rather than trimmed to fit, because the length is a consequence of where the idea actually finishes, not a target aimed at.
Where a Short comes from
The same engine, seen from the input it starts with or the decision it makes next.
YouTube clip maker
Paste the watch URL and it is fetched once on our side, or upload the file. Only the passages you approve are ever cut.
Change video aspect ratio
Five shapes, and which one a clip gets is decided by watching the picture rather than by dragging a box.
TikTok clip maker
Vertical 9:16 with word-by-word captions. It hands you the MP4 — the upload to TikTok stays yours.