Transcribe
Transcribe turns speech into text with timestamps. It runs on Rivul AI's own server in Germany and needs no AI key.
Transcribe files or a recording
Step 1: Open Transcribe in the sidebar, under Convert (rivul.ai/app/transcribe).
Step 2: Choose Accurate (the default) or Fast before adding files. Fast takes about a third of the time and makes a few more mistakes.
Step 3: Upload up to 10 audio or video files at a time (MP3, WAV, M4A, FLAC, OGG, WebM, MP4, or MOV, up to 500 MB each), or click Record to record up to 10 minutes in the browser. A recording can stop by itself after a few seconds of silence; choose how long next to Record.
Step 4: Files are transcribed one at a time, and the list shows each file's place in the queue and its progress. Keep the page open until they finish.
Use the transcript
Select a file to read its transcript; click a timestamp to play the recording from there. Copy the text, or download it as .txt, .srt or .vtt subtitles, or .docx.
Speech recognition can mishear names, numbers, and technical terms, so check quotations against the recording before citing them.
Limits and privacy
Each account can transcribe 2 hours of audio a day and start 10 files an hour. A file longer than what is left of the day is transcribed up to that point, and the transcript says so.
Transcription uses the Whisper speech recognition model on Rivul AI's own server: no AI provider and no API key are involved. The audio is deleted after transcription, is never stored or logged, and is never sent to an AI company.
Something here wrong or missing? Use the contact form.