Neon storefronts at night, slow push-in
CUTROOM

All tools / AI Subtitles

AI Subtitles.

Subtitles that write themselves, then let you fix the one word they missed.

Coming soon

Every other tool is open now. See all tools →

How it works.

Write out what is said and put it on screen as subtitles.

  1. Step 1

    Drag your video in

    It stays on your device while the words come back.

  2. Step 2

    Fix, don't retype

    Low-confidence words are flagged so you review ten words, not ten paragraphs.

  3. Step 3

    Save the subtitles

    Download them as a file, or burn them into the picture.

In detail

Speech recognition needs a model too large for a browser to carry, so this is the one bench that runs on our servers, and it opens with the Cutroom cloud version. Here is what it will do: upload a video and the transcript appears, timed to the frame. Edit any cue inline, pick a caption style, and export SRT, VTT, or a burned-in video. The AI does the typing; you keep the final cut.

What it gives you.

  • Word-level timing

    Cues split on natural phrase boundaries and land on the frame where speech starts.

  • Confidence flags

    Uncertain words carry a mark, so proofreading is targeted instead of total.

  • Inline cue editor

    Edit text next to its timecode. The preview overlay updates as you type.

  • Caption styles

    Clean broadcast, bold block, or outlined social style. Position top or bottom.

  • SRT, VTT, burn-in

    Sidecar files for YouTube and Vimeo, burned-in video for feeds that eat caption tracks.

  • 98-language recognition

    Auto-detected language with per-video override when the detection guesses wrong.

Straight answers.

Can I use the subtitle studio today?

Not yet. Every other Cutroom bench runs entirely in your browser, but transcription needs a speech model on a server, so it ships with the cloud version. Nothing else on the site is waiting on it: trimming, compressing, converting, reframing, and joining all work now, offline, with no account.

How accurate is the transcription?

Clear speech lands in the high nineties. Accents, crosstalk, and loud music pull that down, which is exactly what the confidence flags are for.

Can I edit the timing, not just the text?

Yes. Every cue's in and out points are editable, and the preview reflects changes immediately.

What is the difference between SRT and burned-in?

SRT is a separate file a player can toggle. Burned-in captions are drawn onto the pixels, so they survive every platform and every re-upload.

Does it handle multiple speakers?

Cues follow the audio regardless of who is talking. Speaker labels are on the roadmap, not in this cut.

Other tools