Know who said what. Every meeting, every line.
Otter, Granola and Fathom can leave you with “Speaker 1”. Whozit puts a name on every line it’s sure of, and a “?” on the rest.
Built for teams that meet in person. Put a phone on the table, upload the recording, name each person once; later meetings name the people Whozit recognises by themselves.
Recordings stay in your account; ElevenLabs transcribes and is asked to delete its copy. No bot joins your calls. Delete or export anything, anytime. First 5 hours free, no card.
- 04:12Maya Chen Can we lock the pilot list by Friday?
- 04:31Dev Patel I’ll send the shortlist Thursday.
- 04:40Rosa Alvarez Yes, by Wednesday night.
- 05:02Speaker 4? Does procurement need the review first?
- 06:15Maya Chen Then security goes first. Dev, flag it.
- 06:22Dev Patel On it.
- 07:48Rosa Alvarez I’ll ask who meets in person.
- 08:05Maya Chen Good. Three teams, not six.
How it works
Name each voice once, and Speaker 1 becomes a name.
Many transcript tools stop at Speaker 1 and Speaker 2. Whozit tells you who they are, and says so when it can’t.
- Upload any recordingA phone memo, a call recording, a file from a recorder. Whozit splits it into voices. No bot joins your calls.
- Name each person oncePlay a voice, type a name. That is the only setup.
- Known voices come back namedIn later meetings, the people you’ve named appear by name. Anything Whozit isn’t sure of is marked “?” for you to confirm.
- Unsure voices are marked, never guessedA “?” means Whozit won’t put a name on a guess. You confirm it, or leave it.
You need the right to record, and where the law requires it, people’s consent. Voiceprints count as biometric data in some places. Consent and voiceprints
For AI assistants
Your AI can’t do much with Speaker 3.
Put one phone on the table, then ask your AI what was promised. It reads your transcripts over MCP, a standard way for AI assistants to read your data, so the answer has a name in it.
Works today with Claude Code, Claude Desktop, Cursor and VS Code. Claude.ai and ChatGPT connectors aren’t built yet.
See how to connect your AIWho promised the pilot list, and when?
● whozit.find_lines(words="pilot list")
Dev Patel, by Thursday. Rosa Alvarez checks who meets in person by Wednesday.
- team-sync.md 04:31 · Dev Patel
- team-sync.md 04:40 · Rosa Alvarez
- 05:02 · Speaker 4?
An example, with made-up names.
Privacy
Your recordings stay yours.
Your recordings and voices stay in your account, separate from everyone else’s. Audio goes to ElevenLabs for transcription, and Whozit asks them to delete their copy. You can make Whozit forget a person, or delete or export everything, whenever you like.
Read the privacy pageWhat does the “?” mean?
Whozit isn’t sure. “Speaker 4?” is a voice nobody has named. “Sam Okafor?” with “unconfirmed” is a name Whozit filled in that nobody has checked. Either way, you confirm it or leave it.
Do I have to name everyone?
No. Name the people you care about, once each. Anyone you skip stays a numbered speaker, like Speaker 3.
Does a bot join my calls?
No. Whozit never joins a call. You upload the recording afterwards, from any phone, laptop or recorder.
What can I upload?
A recording from any phone, laptop or recorder, video files included. Whozit uses the audio track.
Can I get my transcripts out?
Yes. Every transcript is Markdown with the names in it, and it’s yours: download one file or all of them as a zip, or delete everything whenever you like.
How does my AI assistant use it?
Claude Code, Claude Desktop, Cursor and VS Code connect over MCP with a token from your account, and read your transcripts with the names in them. Claude.ai and ChatGPT connectors aren’t built yet. Your token only sees your own data.