A YouTube clipper that reads the whole video first
Every passage worth clipping comes back ranked, with the sentences that made it a candidate, and nothing is cut until you approve it. Paste the watch link and it is fetched once on our side, or upload the file you already have.
Two ways in, and the link is the short one
Paste the address of the video and it is fetched once on our side. The file is the other way in, and it is not the lesser one.
The address is enough
A watch page, a Short, a live URL or an embed — youtube.com or youtu.be. The title, the channel and the exact duration are read in one request, before anything is queued or charged, so the dialog can tell you how many minutes your plan will read.
You say the video is yours to use
The import does not start without it. What is recorded is the sentence you were shown, the version of the terms it belonged to, the time and the address — not a tick in a box we forget, because the only question anyone ever asks later is what exactly was agreed to.
Nothing of it is stored here
No copy of the video is kept. The transcription reads an audio-only stream, and each render afterwards fetches only the seconds of its own window — about six megabytes for a clip rather than six hundred for the recording — with those sections cached, so the same moment exported twice fetches nothing the second time. You never download six gigabytes in order to upload them again, and the trade is that a link goes on depending on the video staying where it is.
The one difference that survives: imports are capped by the day — three on the trial, twenty on Starter, fifty on Creator, a hundred on Studio — and uploads are not. Pasting costs you nothing and costs us a fetch; an upload rate-limits itself on your own connection.
How to clip a YouTube video, step by step
Six steps, and you are only needed in two of them: handing it the video, and the decision about what gets cut.
Give it the video
Paste the YouTube address, or drop the file if the recording never went up in the first place — mp4, mov, mkv or webm, up to 8 GB. Either way you confirm the video is yours to use before anything starts.
It gets the bytes, once
A link is fetched on our side and stored a single time. An upload goes from the browser straight to storage rather than through the app, so a long file is not queued behind anything.
It transcribes the audio
Word by word, with timings. Only the audio stream is read out of the video: two hours becomes about sixty megabytes of mono audio, deleted as soon as the words come back.
It reads the whole transcript
Windows of roughly thirty minutes in parallel, with three minutes of overlap, then one judge ranks the pooled finalists. A single pass anchors on the first half and never picks from the back.
You keep or you pass
Each candidate arrives with its score, its four pillars, the sentences behind it, a preview cut of its own and a twelve-frame filmstrip to trim against.
It cuts what you approved
Only the seconds inside the moment are read out of the stored video. The finished MP4 waits in the library for you to download — nothing is posted anywhere, on YouTube or anywhere else.
Reading the video is the expensive part, not cutting it
Transcription and selection cost in proportion to how long the video is. A cut does not, up to a generous per-plan batch of clips. Past that batch, a plan spends three credits per clip from the same balance the minutes come from.
- Transcription reads an audio-only stream — mono, 16 kHz, deleted as soon as the words come back
- A link is fetched in full exactly once, and then never again: every later cut is read out of the copy
- The render reads only the window you approved, and seeks to the exact frame rather than the nearest keyframe
- The transcript is cached against the source, so looking at the same video again costs nothing
- A generous batch of clips is included with every plan — an export past it spends 3 credits, from the same balance as the minutes
Shortening an hour into a short is a selection problem, not a trim
The boundary is a word, never part of one. The transcript is split into sentences first, and a run-on longer than fourteen seconds is broken on the speaker’s own pauses. The model then quotes the words it wants to open and close on, verbatim, and the times are read off those words.
Up to three pieces, in the order spoken. The setup from a minute earlier, then the moment, with the detour between them simply not in the clip. Drag the handles yourself and that assembly is dropped: the gaps were reasoned about a window that no longer exists.
Twelve to a hundred seconds, refused not sheared. A cut that comes back outside the band is made again rather than trimmed to fit. Length is an outcome of where the idea ends.
A stranger reads it before you do. The lines of the clip alone — no transcript, no title, no timecodes — go to a cold read. If it cannot be followed the cut is made again, and a clip that still fails keeps its place in the list with its score capped and a line saying what is missing.
The intro is excluded first. A model reads the opening and finds where the content starts, rather than skipping a fixed sixty seconds and hoping.
How it chooses, in fullTwo looks ship by default, and one pairing is not built
Boxed and full are opinionated on purpose — pick one and every moment you approve comes out the same way.
Boxed 1:1, text behind
The subject is matte-cut and composited over the words, so the line reads and the face is never covered.
Full 9:16, karaoke
Edge to edge vertical, with word-by-word highlighting timed off the transcript.
The frame follows the voice
Speaker tracking is TalkNet-ASD, so on a two-person interview the crop stays on whoever is actually talking.
9:16 with text behind is unbuilt
The compositing geometry is written for the 1:1 box. The options panel corrects the pairing instead of letting a render fail twenty minutes in.
A free YouTube clip maker for the first thirty minutes
Thirty credits, once, with no card — about thirty minutes of video, with ten rendered clips included. Every render style is included too. Once, not monthly: a recurring free allowance is a recurring bill, and one that gets withdrawn later. Trial clips carry a small mark in the corner and are kept for seven days; no paid plan marks anything.
Questions people ask first
Can I paste a YouTube link?
Yes. Paste the watch address — a Short, a live URL or an embed works too — and the video is fetched once on our side, so you never download a long recording in order to upload it again. You confirm you have the rights to use it before the import starts, and that confirmation is stored with the video. Links are YouTube only: for anything else the way in is the file.
Can I clip a YouTube video that is not mine?
That is your call to make and your declaration to sign, not a box this ticks for you. The import asks you to confirm you have the right to use the video and records what you confirmed, when, and from where. What this is built for is your own channel, your own recordings, and footage you were given the rights to — it is not a way to lift somebody else’s uploads, and the record exists precisely so that is not a grey area.
Is there a free YouTube clip maker in here?
Thirty credits, once, with no card, and three link imports a day while you are on it — about thirty minutes of video, with ten rendered clips included. Every render style is included too. Trial clips carry a small mark in the corner and are kept for seven days; no paid plan marks anything.
Does it download the whole video to make one clip?
A link is fetched in full once, and that is the only time. After that the copy is what everything reads: transcription takes an audio-only stream out of it, and a render reads just the seconds inside the moment you approved — about forty seconds of video for a forty-second clip. An upload skips the fetch entirely.
How long is a clip?
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 finishes.
Does it upload the short to YouTube for me?
No. Reading a video from YouTube and posting one back to it are two different connections, and only the first exists here. Nothing is published on your behalf anywhere — not to YouTube, TikTok, Instagram, Twitch or Facebook. The MP4 sits in your library and you download it.
The same engine, from where your video is now
Every one of these ends the same way: an MP4 in your library that you download and post yourself. What changes is where the recording came from.
Clipping software
What the whole pipeline does, end to end: transcript, ranked candidates, your decision, then the cut.
Podcast clip maker
Two hours read in thirty-minute windows with three minutes of overlap, so the back half is ranked too.
TikTok clip maker
Vertical 9:16 with word-by-word captions. It hands you the MP4 — the upload to TikTok stays yours.
Twitch clip maker
Export the VOD, upload it, and read a whole stream for the passages worth cutting. No Twitch account is connected.