Voice Recording to Transcript

Sign in, choose a saved audio file, and select Upload & transcribe. The transcript arrives with timestamps and speaker labels; play the recording back and correct any words before you export.

Yes, you can try it without paying. New accounts get 2 free credits once, which covers about one hour of audio in total, and no card is required. After that, 1 credit covers 30 minutes of audio and packs start at $9.99 for 20 credits. Choosing a file is free; starting a transcription is what spends a credit.

Yes. In Voice Memos, share the recording and save it to Files. Choose that M4A file in the transcription tool.

Save the voice note to your device and select it here. If your browser cannot read the exported Opus file, convert it to WAV or MP3 first.

Try MP3, WAV, M4A, AAC, FLAC, OGG-Opus, AMR, MP4 or WMA. WebM and AIFF are not supported by the speech provider. Codec support also depends on your browser; if a recording fails to open, try MP3 or WAV.

Yes. Files up to 500 MB are accepted, and long recordings are charged in 30-minute blocks: a 2-hour interview uses 4 credits. Jobs run asynchronously, so you can close the tab and reload the page to pick the transcript back up.

We use the provider's recording-file models with punctuation and number formatting on: the standard model for free accounts, and the larger model for Pro, which also separates speakers. On clean recordings accuracy is high; background noise, strong accents, crosstalk, and unusual proper nouns are where errors appear. Replay uncertain segments and check names, numbers, and quotes before sharing — every segment stays editable.

The service is aimed at spoken English recordings; Pro runs the larger model, which also separates speakers. Other languages may work with reduced accuracy, so test a short section first before spending credits on a long file.

On Pro, yes: transcripts carry inferred speaker labels (Speaker A, Speaker B, and so on) that you can rename before exporting. Free accounts run the standard engine, which does not separate speakers. Check any label against the recording: short replies, overlapping speech and similar voices can be misattributed.

Yes, transcription runs in the cloud. Your recording is sent to Tencent Cloud, our speech-recognition provider, which processes it on servers in mainland China and returns the transcript. We delete our own copy of the audio as soon as the result is saved; the provider keeps its copy for a limited period under its retention policy. Nothing is uploaded until you press Upload & transcribe. See the privacy policy for details.

Usually a small fraction of the recording length. Jobs run asynchronously, so you can keep the tab open or reload the page later to pick the transcript back up. Long files take longer to upload on a slow connection.

Yes, with Pro. Free accounts can export TXT and SRT at any time. Pro unlocks DOCX for Word, VTT, JSON, and Print / PDF through the browser print dialog.

Yes. Each timestamped segment has its own editable text field. Your corrections carry over to every export.