LUMIÈRE
Why a timestamp on every word matters
Most transcripts give you a time per paragraph. You still scrub the video to find the exact sentence. In Lumière every single word has its own time. Click a word and you land on that moment, and the same timing is precise enough to cut on.
The language is picked automatically. If the guess is wrong, set "Mainly spoken" to the language of your footage before you start.
Step by step in Studio
- Open Studio and create a project. Give it a name you will recognise later.
- Open Import, choose your files (MP4, MOV, MKV, WebM or AVI) and press "Upload and transcribe". Large files upload in pieces, so a long interview is fine.
- Transcription starts by itself when the upload is done. When it finishes, you land on Understand.
- Open the Transcript tab. You see every word with its timing.
Transcription costs nothing. It works in any browser, on a laptop, a tablet or a phone.
Find the quote
Search the transcript for a word and jump to every place it was said. If you would rather ask than search, use Ask on the Understand step: write a question in your own words, and the answer comes only from this footage, with the time you can click.
If you work as a journalist, LumiVision Press goes a step further. You upload a recording, read it in your own language, and ask what was said. The answer quotes the exact line and its time, and every quote is checked against the transcript before you see it. It costs 19 € a month with 120 ✦.
AI transcripts can contain mistakes. Before a quote goes into print, play the moment and hear the exact words.
Export, subtitle, cut
- Download the transcript as TXT, or as SRT, a timed subtitle file.
- Turn the same transcript into subtitles. Boxed, Minimal or Karaoke style, burned into the picture or as a separate file. Correct any line by clicking it.
- Add other languages. Translation is free and offers 32 languages.
- Cut the video by striking words in the transcript: click a word and it is gone from the video. Hand the cut to DaVinci Resolve or export an EDL for any editor.
Who said it, and where your recording goes
Speaker labels are not available yet. Today the transcript is one text for the whole recording, so for a two-person interview you add the names yourself. Speaker detection is planned and is described on its own tool page as coming soon.
Your upload goes to servers in the EU and is never used to train models. You can delete a project from Lumière's servers at any time. For sensitive material, the desktop app Lumière Local (early access) keeps the footage on your computer; only the transcript, which is text, travels to the cloud.