Speaker-labeled transcription
Transcribe an interview with optional speaker detection, then review and rename the labels.
Daily processing allowance
Checking current allowance...
Uploaded and processed audio is stored on SoundHalo. Audio access expires after 30 days; expiry does not confirm storage deletion. Delete the stored files here before leaving this page. Speech providers have their own retention policies. Privacy policy.
Uploads use an external speech provider. Public transcription accepts 100 MB and 30 minutes per file, with 60 audio minutes per shared network per UTC day. Users on the same IPv4 address or IPv6 /64 share this allowance. A separate processing-cost budget also limits repeated jobs. Speaker labels are optional and need review.
Output presets
Only settings are saved locally. Recordings and transcripts are not saved here; download your work before leaving this page.
Diarization is not speaker identification
The model groups speech by voice. It does not know the speaker’s name. Check short turns and overlapping conversation carefully. If automatic detection is unavailable, the tool must say so rather than inventing speaker labels.
Related tools
FAQ
Where is the audio processed? +
The recording is uploaded to SoundHalo and sent to an external speech provider. SoundHalo stores uploaded audio. Access expires after 30 days, which is not proof of storage deletion. Use Delete stored audio before leaving the page to remove SoundHalo files sooner. Provider copies and saved account transcripts have separate retention. Review the privacy policy before uploading sensitive recordings.
Can I trust every speaker label? +
No. Treat labels as a draft and verify them by listening. Very short turns, crosstalk and similar voices can be assigned incorrectly.