Talking-Head Cutter

Daily check-in: 15 free credits

Talking-Head Video Cutter: Rough Cut in Minutes

The talking-head cutter is an AI rough-cut tool for talking-to-camera recordings: it transcribes an MP4 or MOV of up to 31 minutes word by word and strikes every long pause, filler word and retake. Keep any cut with one click, then render the jump cut. The cutter uses purchased credits.

Start a rough cut

  1. Upload your talking-to-camera recording (MP4 or MOV, up to 31 minutes and 5 GB).
  2. Review the cut: every pause, filler word and retake is struck through in the transcript. Click any cut to keep it.
  3. Render the jump cut, with optional captions and a subtle punch-in, or export the timeline to your editor.

One recording per project: MP4 or MOV, up to 5 GB and 31 minutes, spoken in English or Chinese.

Uses purchased credits. A 10 minute recording costs about 7 credits to analyze and about 11 credits to render.

How the talking-head cutter works

  1. 1

    Upload the recording

    Drop in one MP4 or MOV, up to 31 minutes and 5 GB, straight from your camera or phone. It uploads directly to storage, so a large file is fine.

  2. 2

    AI finds the cuts

    The recording is transcribed with word timings. Pauses over your threshold, filler words and abandoned takes are marked as cuts.

  3. 3

    Review in the transcript

    Every cut is struck through in the transcript. Click a cut to keep it, move the pause slider, and preview the result in the browser.

  4. 4

    Render or export

    Render one jump-cut MP4 with optional captions and a subtle punch-in, or export an editable timeline for Premiere Pro, Final Cut Pro or DaVinci Resolve.

What it cuts and how

  • Silence removal with padding

    Cut every pause longer than 0.2 to 3 seconds, and keep a little air around each word (0.1 s by default) so the cut never clips a breath.

  • Filler words

    Removes um, uh, erm and hesitation sounds, plus like and you know when they stand alone. In Chinese it removes 嗯, 呃, and 那个 or 就是 used as fillers.

  • Retakes, last take kept

    When you restart a sentence, AI reads the transcript and cuts the earlier attempts, so only your last complete take stays.

  • Every cut is reversible

    Nothing is final until you render. Click any struck word or pause chip to bring it back, or undo all kept cuts in one click.

  • Captions and punch-in

    Burn in captions built from the words you kept, in three styles, and add a subtle 1.08x punch-in on every other cut to hide the jump.

  • Editable timeline export

    Export the kept ranges as FCPXML and Premiere XML with SRT captions, then relink to your original file and finish the edit in your own software.

Who uses it

  • YouTube creators

    Turn a 20 minute rambling take into a tight video before you add B-roll and graphics.

  • Course and tutorial makers

    Clean up lesson recordings without scrubbing through the waveform by hand.

  • Podcasters with video

    Cut the dead air and false starts from a solo episode recorded on camera.

  • Founders and marketers

    Record a product update in one go, then let the cutter remove the retakes.

What a rough cut costs

  • The talking-head cutter uses purchased credits. Each step is charged on its own and refunded if it fails.
  • Analysis: about 2 credits to read the recording, about 1 credit for every 2 to 3 minutes of transcript, and about 1 credit for the retake check. A 10 minute recording costs about 7 credits.
  • Render: about 2 credits for every 2 minutes of the finished cut plus 1 credit to join the parts. A 5 minute cut costs about 7 credits.
  • Changing the cut, the preview and the timeline export are free. You can render again after every change.

More AI video tools

A Guide to Talking-Head Rough Cuts

What is a talking-head rough cut?

A talking-head rough cut is the first pass of an edit of a talking-to-camera recording: the long pauses, the filler words and the failed takes come out, and the speech plays as one tight jump cut. B-roll, graphics and music come after it.

The talking-head cutter makes this pass for you. Upload one MP4 or MOV of up to 31 minutes and 5 GB, with the speech in English or Chinese, and you get back a transcript where every planned cut is struck through.

How are pauses and filler words found?

The recording is transcribed with a timing for every word. Every gap longer than your pause threshold, from 0.2 to 3 seconds, becomes a cut, and a little air (0.1 seconds by default) stays around each word so no breath is clipped.

Filler words are matched in context: um, uh and erm always, like and you know only when they stand alone. In Chinese the cutter removes 嗯 and 呃, and 那个 or 就是 when they are used as fillers.

How are retakes detected?

When you stop and say a sentence again, Claude Haiku reads the transcript and marks the earlier attempts, so only your last complete take stays. You see each marked attempt in the transcript before anything is rendered.

Every cut is reversible. Click a struck word or a pause chip to keep it, or click again to cut it; the browser preview updates at once.

Rough cut or full edit: which tool to use

Use the talking-head cutter when the recording itself is the video and it only needs the dead air removed, or when you want to finish the edit in your own software: the timeline export opens in Premiere Pro, Final Cut Pro and DaVinci Resolve.

Use B-roll Cut when the speech needs other footage on top of it, and the AI video editor when you want captions, reframing or a color grade from a plain-language instruction.

What it costs

The talking-head cutter uses purchased credits, and each step is charged on its own and refunded if it fails. Changing the cut, the preview and the timeline export are free, and you can render again after every change. The pricing section above gives the credits for each step.

Talking-head cutter questions

What does the talking-head cutter make?
One MP4 jump cut of your own recording in its original shape (up to 1920 px on the long side, 30 fps), with every cut you approved removed, the sound normalised, and optional burned-in captions. You can also export the same cut as an editable timeline.
Which files and languages does it accept?
One MP4 or MOV per project, up to 5 GB and 31 minutes, with a sound track. The speech can be English or Chinese.
How does it decide what to cut?
The recording is transcribed with a timing for every word. Pauses longer than your threshold are cut with padding kept around each word, filler words are matched in context, and Claude Haiku reads the sentences to mark earlier attempts of a sentence you said again.
Can I keep a cut it made?
Yes. Every cut is struck through in the transcript. Click it to keep it, or click again to cut it. The preview updates at once and nothing is charged until you render.
Does it work with Premiere Pro, Final Cut Pro and DaVinci Resolve?
Yes. Export timeline saves a ZIP with FCPXML (Final Cut Pro and DaVinci Resolve), Final Cut Pro 7 XML (Premiere Pro) and SRT captions. Relink the clip to your original recording and every cut is in place.
Is it free?
No. The cutter runs on purchased credits. A 10 minute recording costs about 7 credits to analyze and a 5 minute cut about 7 credits to render. A failed step is refunded.
Why does the preview not play my MOV?
Some phones record HEVC MOV files, which only Safari plays. The cut, the render and the export still work; the rendered MP4 plays everywhere.