Want captions on a video? Open the caption generator.
FAQ
How do I transcribe audio to text?
Choose an audio or video file and click Transcribe. Review the words, timing, and speaker labels, then copy the transcript or download a file.
Can I transcribe a video?
Yes. The tool transcribes the audio track from your video. It does not add captions to the video; use the caption generator for that.
Which file should I download?
TXT is a plain transcript. SRT and VTT are subtitle files for editors and players. CSV includes text, speakers, and times. JSON includes transcript segments and timing data. RTTM stores speaker timings and names without transcript text.
Will downloads include my corrections?
Yes. TXT, SRT, VTT, CSV, and JSON include your corrected text. Downloads that contain timestamps or speaker labels also use your timing and speaker corrections. Deleted segments are left out of every download. RTTM contains only speaker labels and times, so there is no transcript text to update.
How do corrections affect word timestamps?
A word correction keeps its original timing. Adding or removing words estimates timings around that edit; changing a segment’s start or end adjusts its word timings to fit. JSON includes corrected words and marks estimated timings. Other formats use your corrected text and segment times. Review timing after larger edits.
Can I transcribe just part of a file?
Yes. After choosing a file, use Start and End under Section to transcribe to choose the part you want. Enter times in seconds. The first section is selected automatically: up to 30 seconds on mobile or 10 minutes on desktop.
Where do downloaded timestamps start?
Zero means the start of the section you transcribed. If you set Start to 120 seconds (2:00), speech at 2:05 in the original file has a download timestamp of 0:05. JSON also records where that section starts in the original file.
Can I label different speakers?
On desktop browsers, enable Identify multiple speakers before transcribing. Review assignments and rename the labels. You can add and assign speaker labels manually on mobile or desktop, up to eight total. Overlapping voices may need correction; the transcript follows one speech stream.
Which languages can I transcribe?
On mobile, Whistle supports English, French, German, Italian, Dutch, Polish, and Spanish. Desktop also offers Whisper Tiny and Base for 99 spoken languages. Automatic detection uses the languages supported by your selected model. Transcripts stay in the spoken language; they are not translated. Accuracy varies by language, accent, and recording quality. This language support applies to transcription; speaker identification uses a separate model.
Does speaker identification support all 99 languages?
Speaker identification uses Nemotron 3 Diarization to tell voices apart and label who spoke when. On desktop browsers, you can enable it with any transcription language. The model was trained and evaluated on multilingual audio, but accuracy is not guaranteed across all 99 languages. Review the speaker labels, especially when voices overlap.
What files and lengths are supported?
Choose audio or video up to 600 MB. Select up to 30 seconds on mobile or 10 minutes on desktop. Common formats include MP3, WAV, M4A, FLAC, Ogg, MP4, MOV, and WebM. Browser codec support varies. If needed, Prepare compatible copy creates an audio file for transcription.
Is the audio transcriber free and private?
Yes. No signup or payment. Your file and transcript stay on your device, and speech recognition runs in your browser.
Which browser and model should I use?
Whistle uses a small download and runs on mobile and desktop. Desktop also offers Whisper Tiny for speed and Base for improved recognition across more languages. Keep this tab open while processing.
Why does the first transcription take longer?
The speech model downloads on first use: about 18 MB including the runtime for Whistle, 44–154 MB for Whisper Tiny, or 80–294 MB for Base. On desktop, speaker identification adds about 106 MB. Your browser caches downloads when storage allows.