Transcription
ScriptCut turns video and audio in 54 languages into a transcript with timecode on every word, so the text stays locked to your footage. It is built for editors, producers and podcasters who cut unscripted material by reading, then send the result to their editing software.
You cannot cut what you cannot search. Hours of unscripted footage, and a transcript that drifts the moment you edit it.
Languages
Pick the spoken language, or leave it on Auto select. Every language comes back word by word, with timecode on every word.
Then translate. AI Translate turns any transcript into 48 languages, so you can read the footage and pick moments in your own language while the cut stays on the original recording's timecode. Subtitles export as SRT or VTT in any translated language.
How it works
01
Upload video or audio from your computer or the cloud. Pick the spoken language from 54, or let it detect.
02
Every word arrives with its own timecode and a speaker label.
03
Play a short voice sample, type a name once, and every line updates.
04
Fix spelling, remove fillers and pauses. Timecode stays attached to the footage.
05
Download the transcript, or move on to selects and a timeline XML.
Speakers
Open Manage speakers, play a voice sample, type a name, and every line from that person updates.
Type a name that already exists and the two speakers merge. Set a role (Interviewer, Interviewee or Crew) to filter crew lines out in one click. A short automatic check re-listens to lines next to a speaker change, and a wrong line can be reassigned from the row menu.
Cleanup
Cleanup changes the edit and the export, never the recording or the per-word timecode.
Remove Fillers finds um, uh and hmm everywhere and previews each cut before it applies. Remove Pauses finds silent gaps from the word timecode; half a second always stays so the cut sounds human.
Both are reversible, word by word. Check Spelling swaps one word for one word, so timing never shifts. Retyping or deleting words does pull text out of step with the audio, so avoid it.
Pause threshold: 1s, 1.5s, 2s or 3s
Numbers and units like 35mm are never touched
Type what you remember into the search above your projects. ScriptCut looks inside the transcript of every project in your account at once, by the words or by the idea.
Results come grouped by project, closest first, each with its speaker and time. Click one and that project opens in Select on that exact line, ready to mark. On a Team account, everyone searches the shared projects.
A project joins the search the first time you open it. Clients on a review link cannot search.
Words, other word forms and small typos
Similar meaning, tagged
Quotes for an exact phrase
Several cameras
Where Several recordings is on, each file keeps its own name, length, frame rate and transcript.
Transcribe each file in ScriptCut or bring a transcript you already have. When you add a file, ScriptCut asks who is in it and offers to match speakers already in the project. It never merges on its own. The timeline XML joins the files, each at its own frame rate. This is not multicam sync.
Take it out
Download txt, PDF, docx, srt, vtt or a word-timed json. Every download is the transcript exactly as it reads in Cleanup, with your edits applied. Bring in an SRT, VTT, Resolve SRTX or Premiere transcript JSON too. Or carry on to selects and a timeline XML for your editing software.
Why ScriptCut
Fillers and pauses come out of the edit and the export only. The recording and the per-word timecode stay as they were, and every cut can be restored.
Rename a speaker and every line follows. Add another file and ScriptCut offers to match its speakers to people already named.
The same transcript carries on to selects and a timeline XML for your editing software, at the footage's own frame rate.
Upload the video or audio, pick the spoken language or let ScriptCut detect it, and start the transcription. Every word arrives with its own timecode and a speaker label.
Accuracy depends on the audio: clear speech and a good microphone give cleaner results than crosstalk or noise. We do not publish a percentage. Every word is editable, and Check Spelling flags likely errors.
54 languages: Afrikaans, Albanian, Arabic, Azerbaijani, Bosnian, Bulgarian, Catalan, Chinese, Croatian, Czech, Danish, Dutch, English, Estonian, Filipino, Finnish, French, Galician, German, Greek, Hebrew, Hindi, Hungarian, Icelandic, Indonesian, Italian, Japanese, Javanese, Kazakh, Korean, Latvian, Lithuanian, Macedonian, Malay, Maltese, Norwegian, Persian, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Sundanese, Swahili, Swedish, Tamil, Turkish, Ukrainian, Urdu, Uzbek, Vietnamese and Welsh. Pick the language or leave it on Auto select. Every language comes back word by word with timecode, and speakers are detected in nearly all of them. AI Translate then turns any transcript into 48 languages.
No. Remove Fillers cuts the words from the edit, not from the recording, so every line still matches the footage. You can restore any word later.
Yes, where Several recordings is on. Each file keeps its own name, length, frame rate and transcript, and the timeline XML joins them at each file's own frame rate. It is not multicam sync.
Yes. ScriptCut accepts SRT, VTT, Resolve SRTX and Premiere transcript JSON, and matches each to its file by name, length and timecode. Word-level files can be trimmed to the word. Sentence-level files import but cannot.
Deleting or retyping words pulls the text out of step with the audio and its word timecodes. In ScriptCut, fix spelling, split lines, use Remove Fillers, or delete a whole row. Those keep timing intact.
Yes. The search above your projects looks inside the transcript of every project in your account and lists the closest lines, with the project, speaker and time.
No. Other forms of a word and small typos still match, and lines that say the same thing in other words show with a Similar meaning tag. Put words in quotes when you want the exact phrase.
The project opens in Select on that line and highlights it, so you can mark it right away.
Yes, once you have opened them. A project joins the search the first time you open it in ScriptCut.
Yes. Word search works in the language of your transcript, and accents never get in the way: a search for cafeteria finds cafetería. Search by meaning looks across languages too, so a search in English can find the same idea said in German or Spanish.
A copy or a Social clip set shares the original's lines, so each line shows once, in the original. If the original goes to the Trash, the copy takes its place.
No. The search only reads your own account. A client on a review link sees only the project you shared.
No. To search by meaning, the text of your transcript lines goes to our AI provider, whose API terms say that data is not used to train models. Your video and audio are never sent for search.
Keep going
Let AI pick selects from your transcript, then refine them.
Send clients a link to pick and approve moments in the transcript.
Send a timeline straight into Resolve or Premiere.