Audio to text converter with speaker names and timestamps
Typing up an hour of interview audio takes four or five hours by hand. Send the recording to your todo.is agent and it transcribes it, labels who is speaking, adds timestamps you can jump back to, and marks anything it could not hear clearly so you know what to double-check.
The prompt
- Transcribe the attached audio: [ATTACH THE AUDIO FILE]. It is [WHAT IT IS] in [LANGUAGE]. The speakers are [SPEAKERS]. Label each turn with the speaker's name and add a timestamp every time the speaker changes. Style: [TRANSCRIPT STYLE]. Mark unclear words as (inaudible 12:34) with the time. Give me [FORMAT], plus a list of names, places and terms you were unsure how to spell.
What to change
- [ATTACH THE AUDIO FILE]: Attach MP3, WAV, M4A, OGG, FLAC or another audio file, up to 50 MB.
- [WHAT IT IS]: E.g. "a research interview", "a podcast episode", "a client call" or "a lecture".
- [LANGUAGE]: E.g. "English", "Spanish", or "English and French mixed".
- [SPEAKERS]: Names and roles, e.g. "Dr. Ana Ruiz (interviewer), Tom Hale (participant)".
- [TRANSCRIPT STYLE]: E.g. "clean verbatim (remove ums and repeats)" or "full verbatim (every word, for research)".
- [FORMAT]: E.g. "a Word file", "plain TXT", "both" or "a Google Doc in my Drive".
Example result
- Transcript: research interview with Tom Hale
- 52 min · 2 speakers · clean verbatim · 8,140 words · English
- Excerpt
- [00:00:04] Dr. Ana Ruiz: Thanks for joining. To start, can you tell me how you first got into community gardening?
- [00:00:11] Tom Hale: Sure. It was 2019, we had just moved to the estate and there was this empty lot behind the car park. A neighbour, Priya, put a note through every door asking who wanted to help.
- [00:00:29] Dr. Ana Ruiz: And what made you say yes?
- [00:00:32] Tom Hale: Honestly, I was lonely. I worked from home and I did not know anyone. The garden gave me a reason to be outside on a Saturday.
- [00:14:08] Tom Hale: The council gave us a small grant in the second year, but the water was always the problem. We had a hose from [inaudible 00:14:15] flat, which was not really allowed.
- What the transcript includes
- • Speaker names on every turn, not "Speaker 1"
- • A timestamp at each change of speaker, so you can find any quote in the recording
- • Clean verbatim: ums, false starts and repeated words removed; meaning untouched
- • 4 [inaudible] marks where a bus passed or both people talked at once
- Please check these spellings
- • "Priya" (neighbour)
- • "Greenhollow Estate" (heard twice, could be "Green Hollow")
- • "Rotary Allotment Fund"
- Accuracy note
- Clear recordings usually transcribe very accurately. Accents, crosstalk and background noise lower accuracy, so check any quote before you publish it.
How to do it with todo.is
- Copy the prompt and fill in the language, speakers, style and format.
- Attach the recording in todo.is, or send the file to your agent on Telegram, WhatsApp or email.
- Your agent transcribes it, labels speakers and returns the transcript file.
- Ask follow-ups like "find every place Tom talks about funding" or "summarize the main themes".
Tips for a better result
- Give speaker names and roles up front. It turns "Speaker 1" into real names and makes the transcript usable right away.
- For research, ask for full verbatim. Pauses and hesitations can matter for analysis.
- Send a list of tricky names and terms before transcribing. Your agent will spell them right throughout.
- Record close to the mic in a quiet room. It does more for accuracy than any setting.
audio to text converter: FAQ
- How accurate is AI audio to text? For clear speech with little background noise it is usually very accurate. Strong accents, overlapping voices and noise cause more errors, and your agent marks unclear spots.
- Can it tell different speakers apart? Yes. Give the speaker names and your agent labels each turn. Very similar voices can sometimes be mixed up, so check key quotes.
- Which languages can it transcribe? Many languages, including recordings that switch between two. Say which languages to expect.
- Is my recording private? Your audio and transcript stay in your own agent workspace and only your agent uses them. Get consent from the people you record.
JavaScript is required to use the todo.is app.