Transcribe a recorded phone call
Call recordings transcribed on your own device — useful for support calls, interviews conducted by phone and conversations you are entitled to have recorded.
Phone audio is the hardest common input, and this page is honest about what to expect from it.
Language and translation need the Multilingual model.
Drop an audio or video file
or , paste with Ctrl + V, or
- MP3
- MP4
- WAV
- M4A
- MOV
- WEBM
- OGG
- FLAC
- AAC
The file is read by this page and never uploaded — no account, no size limit but your own RAM.
Preset: balanced English model, timestamped output.
Download
Click a line to jump the audio there. Click the text to fix a word — your edits go into every download.
How it works
- Open this page — the tool is already set up for “Phone calls”.
- Drop the file in, browse for it, or paste it with Ctrl+V.
- The model runs on your own device; the transcript appears as it goes.
- Correct anything misheard, then download text, subtitles or notes.
Check you were entitled to record it
Recording-consent rules differ sharply by jurisdiction — some places require only one party to consent, others require everyone on the call, and the rules can differ again for business calls. This tool takes a file you already have and expresses no view on how you obtained it, but a transcript makes a recording far easier to circulate, which raises the stakes if the recording itself was not permitted.
If there is any doubt, resolve it before transcribing rather than after. That is a legal question for your own advisers, not one this page can answer.
Why call audio transcribes worse
Telephone speech is narrowband: traditional call audio carries roughly 300 Hz to 3.4 kHz, discarding most of the high-frequency information that distinguishes consonants. On top of that sit lossy codecs designed for intelligibility rather than fidelity, and often a speakerphone in a room. Whisper handles it — it was trained on plenty of telephone-quality audio — but accuracy is measurably below a close-microphone recording, and consonant-heavy words and names suffer most.
Practically: expect a usable draft rather than a clean one. Names, numbers and spelled-out references are where to look first, and they are exactly the parts of a support or sales call anyone cares about.
Both sides of the call
How the call was captured matters more than any setting. A recording made by the phone system usually has both parties mixed reasonably. A recording made by putting a phone on speaker next to a second phone captures your side loudly and the other side through two rounds of processing — that side will transcribe poorly and there is no fixing it afterwards.
If the recording is stereo with one party per channel, that survives here: channels are averaged rather than dropped, so both voices are present. Where you have genuinely separate per-party files, transcribing each on its own gives real speaker separation.
Things that save a re-run
- Check numbers, names and anything spelled out — narrowband audio hurts exactly those.
- A phone-on-speaker-next-to-another-phone recording will transcribe badly; use the system’s own recording if you can.
- Separate per-party files give real speaker separation — transcribe each and interleave by timestamp.
Read the transcript against the audio before you rely on it. Every speech model — this one and the paid cloud ones — mishears names, numbers and crosstalk, and it does so confidently. The player and editor here exist so that check takes minutes rather than an afternoon.
Frequently asked questions
Is it legal to record and transcribe a call?
Recording law varies by country and by US state — some require all parties to consent, some only one, and business calls can be treated differently again. This page does not advise on it. If you are unsure whether a recording was lawfully made, settle that before transcribing it.
Why is call audio less accurate than other recordings?
Telephone audio is band-limited and heavily compressed, which removes much of the detail that distinguishes similar-sounding consonants. It is a limitation of the recording, not of the tool, and affects every transcription service equally.
Is the call recording uploaded?
No. It is read by the page and transcribed by the model running in your browser. Nothing is transmitted, which is the point — a call recording usually involves someone who did not choose to have it processed elsewhere.
Can it separate the two people on the call?
Not automatically. If you have a stereo recording with one party per channel, transcribing each channel separately gives real separation. Otherwise use the manual speaker labels.