Try it now
A working sandbox. No sign-up, no project. Sample data only.
- WAV, MP3, M4A, MP4, MOV, MKV and most other formats.
- Video is read and stripped to audio in the browser before upload, however large. A 10 GB camera original costs a few minutes of disk reading and about 20 MB of audio per hour goes up.
- Speakers separated automatically. Rename one and the name applies project-wide.
- Click any line to seek the player. Edits save as you type.
- An interview analysis per file, a rating from one to five, archive folders, and a 90-day trash.
- Fewer speakers than the room had? Re-diarize listens again and keeps every word.
How it works
Three steps from raw material to result.
Audio or video, up to a few hours per file. The audio is extracted automatically and routed to the transcription engine you picked.
Take me back to that night. Did you have any idea what was going to happen?
It wasn't planned. None of it was. We were sitting there… and the lights went out.
Speaker turns are detected and a voice fingerprint is captured for every voice in the file. Each line is clickable and seeks the waveform.
Correct typos inline, name the speakers once, and the renames propagate across every interview where their voice appears.
Frequently asked questions
What audio formats can I upload?
WAV, MP3, M4A, AAC, OGG, FLAC, AIFF, and most video containers (MP4, MOV, M4V, MXF, MKV, MTS, WebM, AVI). Audio is pulled out of video files for you before transcription runs.
How accurate is the transcript?
Word error rate depends on audio quality and accent. On clean studio interviews in major languages, accuracy lands in the high 90s; on noisy field recordings with strong regional accents, it drops into the 80s. You always get the source audio aligned to the text so you can quickly review and correct.
Can I edit the transcript after it lands?
Yes. The transcript opens in a three-panel editor next to the audio waveform. Click any line to seek the player, edit text inline, rename speakers, or merge identities. Changes save instantly.
Does the transcript include speakers?
Speaker turns are detected automatically. Identities are voice-fingerprinted and matched across every interview in the same project, so once you name a speaker, the change applies everywhere they appear.
How long can a file be?
Single uploads are capped at a few hours of audio per file in the free tier. Longer interviews are usually split into reels during recording; you can upload them as a folder and the project handles them as one session.
Do I have to upload the whole video file?
No. The audio is separated out in the browser before the upload starts, so what crosses the connection is the sound rather than the picture. On a day of rushes that is the difference between an overnight upload and a coffee break. Uncompressed audio is compressed locally for the same reason. If a file turns out to be unreadable on your machine it uploads unchanged and is handled on our side instead, which costs you upload time but not the transcript.
What about MXF, MKV or other camera formats?
MP4, MOV and M4V are read on your computer, and so is WAV or AIFF of any size, so the picture never leaves it. MXF, MKV, MTS and other containers a browser cannot read are uploaded whole, in resumable parts that retry on their own, and the sound is separated on our side. The upload banner says which path a file is taking.
What if the transcript has too few speakers?
Re-diarize. It listens to the audio again with the newest speaker model and relabels the lines in place, splitting any line that glued two people together. Every word, edit, note and paper cut row stays where it is. You can set an optional upper limit on speakers, and leave it empty to let the model decide.
Can I export the transcript as a spreadsheet?
Yes. A transcript exports as a Word document, a PDF or an Excel spreadsheet, each with a live preview beside the options. The Excel file has a column for time, speaker, transcript, translation and highlights, with a filter on every column so you can sort by who spoke.
Related capabilities
Further reading
Background guides and comparisons.
Word error rate is the standard accuracy metric, but it understates the problems that matter most for documentary audio: proper nouns, accents, crosstalk, and technical terms. Here is what WER measures and what it does not.
Otter.ai is built for meeting notes. PaperCuts is built for multi-speaker documentary interviews, speaker identity across files, and a post-production assembly layer. The gap is not about accuracy.