Find the best moments in a long video

Input

Upload MP4, MOV, WebM, MP3, WAV, M4A, FLAC — or paste a YouTube, TikTok, Facebook, X, Vimeo, Twitch, or Dailymotion link

What a pick contains

Start/end timecodes, the hook quote, why it works, a 0–100 score, platform fit, and the clip’s transcript text to copy

Clip length

15–75 seconds, on sentence boundaries; most picks land in the 20–45 second range

Cost

2 AI credits for the first scan of a recording; 1 credit each time you ask for more

Cutting

Uploaded files: cut any pick to an MP4 or audio clip in the browser, no upload (paid plans). Link-sourced recordings: picks only

Vertical export

Clip Studio reframes a clip to 9:16 and exports a 1080×1920 MP4 rendered in your browser

How to find the best moments in a long video

1. Transcribe the video

   In Transcribe Studio, upload the file or paste the link. DomeScribe produces a timestamped transcript — the clip finder works from those timestamps, so pasted text without timing is not eligible.

2. Run Clips

   Open Clips on the recording’s workspace and scan. The first scan costs 2 AI credits and builds the recording’s whole pool of moments; the app reveals them a few at a time, and each later “find more” costs 1 credit.

3. Review the picks

   Each pick shows its timecodes, the quote a viewer would screenshot, one sentence on the hook and the payoff, a score, and the platforms it suits. Click a timecode to hear the moment in the player before you commit to it.

4. Cut the clip

   If the original file is on hand in your browser (you uploaded it), cut the pick into an MP4 or audio clip right there — nothing is uploaded. This step is part of the paid plans.

5. Reframe for Shorts, Reels, or TikTok

   Open the clip in Clip Studio, choose a layout, steer the frame while it plays, add captions, and export a 1080×1920 vertical MP4.

What the clip finder is actually looking for

A good short clip is not the loudest minute of a recording — it is the minute that works with no context. DomeScribe’s clip finder scores every candidate moment against a short rubric: a hook that lands cold (a surprising number, a bold claim, a question, the start of a story, or a line with real emotional charge), a payoff the clip ends right after (a punchline, a conclusion, a concrete tip, a reveal), and self-containment, so a viewer who has heard none of the episode still gets it. Each pick holds one idea; a passage with two strong moments becomes two picks.

Length is constrained to 15–75 seconds, always cut on sentence boundaries. The prompt prefers dense 20–45 second moments and only goes longer when the payoff needs it — which matches what TikTok, Reels, and Shorts reward, and what LinkedIn tolerates.

Fewer picks, on purpose

The scan is told how many moments a recording of that length can plausibly hold — roughly one per three minutes, with a floor of two and a cap of 20 — and it is told that the number is a ceiling, not a target. Every returned pick must clear a score of 70 out of 100. A seven-minute talk may yield two or three; a thin hour may yield none, and an empty result is treated as the correct answer rather than padded with ordinary passages. That is deliberate: a weak pick costs a creator more time than a missing one.

Picks never overlap, and each one comes with a one-sentence reason naming the hook and the payoff, so you can judge the call without re-listening.

From pick to posted clip

Picks are a list of moments with timecodes; what you can do next depends on where the recording came from. If you uploaded the file, DomeScribe still has it in your browser, and each pick can be cut into a real MP4 (or an audio clip from an audio source) on your machine — the media is not uploaded to cut it. From there, Clip Studio turns a landscape clip into a vertical 9:16 post: pick one of six layouts (Full, Stack, Trio, Spotlight, Inset, Wide), drag the frame while the clip plays, generate burned-in captions from the transcript, and export a 1080×1920 MP4 rendered in your browser.

If the recording came from a pasted link, you get the picks — timecodes, quote, reason, and the transcript text for each clip — and you cut the moment in whatever editor holds your master. DomeScribe never produces a media file from a platform link.

• Uploaded file → picks + in-browser cut (paid) → Clip Studio vertical export
• Pasted link → picks + copyable clip text; cut from your own copy of the video
• Audio-only source → picks + audio clip or quote text; no vertical video is produced

Platform fit, explained

Each pick carries the platforms it genuinely suits rather than a blanket “everywhere”. Funny, emotional, or punchy moments are tagged for YouTube Shorts, Instagram Reels, and TikTok; a professional insight or a concrete method is tagged for LinkedIn (and Shorts when it is under a minute). The sweet spots the finder works to are 20–45 seconds for TikTok and Reels, 25–55 seconds for Shorts, and 30–75 seconds for LinkedIn. Treat the tag as a starting point — your audience may disagree, and the timecodes work the same wherever you post.

Languages, accuracy, and what the AI can’t see

The finder reads the transcript, not the pixels: it judges what was said, not camera cuts, facial expressions, or on-screen graphics. Transcription runs in 99 languages with automatic detection, and quotes and reasons are written in the recording’s language. Because the transcript is machine-generated, the finder is instructed to read through obvious mis-transcriptions and not reproduce them in quotes — but it cannot rescue audio it could not hear, so a clear recording gives better picks.

Privacy follows the usual DomeScribe rules: the uploaded original is deleted after transcription, the transcript is encrypted at rest and auto-deletes after 30 days, and cutting a pick from an uploaded file happens in your browser.

Questions, answered

Related

• Turn a podcast into short clips
• Clip Studio — reframe and export vertical clips
• Repurpose a long-form video into posts
• Transcribe Studio — where the picks are made
• Turn a podcast into short clips

Try it on your own recording

Free to start — 30 minutes of transcription and 5 AI credits, no card required.

Find clips in a recording · See pricing