Why voice notes usually go nowhere
Recording a thought takes five seconds. Using it later takes much longer: a block of words with no punctuation, no date and no list, in an app nobody opens again. So people stop dictating, or keep dictating into a void.
What Noteer does with a voice note
- You talk. Tap the mic, speak, tap Stop. A recording can run up to ten minutes. On a phone the mic sits next to the field where you type.
- It is written down in your language. The audio goes to Whisper (large-v3) and comes back as text in the language you spoke. Ukrainian and Russian are both tried and the surer version is kept, so one isn’t written as the other. The audio is not stored.
- The agent reads it. Like any note, it may get one line: a due date from “before the end of the month”, who a name refers to, or a yes/no guess.
- It offers the tidying. A spoken list (“the chargers, the adapter, sunscreen…”) gets “Make a checklist?”. A long note that reads like raw speech gets “Polish text”, a readable version shown side by side; only “Use this” replaces your words, with Undo. A note that mixes three subjects gets “Split?”.
Tips for better voice notes
- Say the day. “By Thursday”, “next Friday”, “on the 12th”: the agent reports it, and code turns it into a date in your time zone.
- Say names. The agent learns who Marcus is the first time you tell it, and uses that from then on.
- Don’t worry about structure. Ramble. Split, checklist and polish are offered afterwards.
Privacy
Audio is passed to the transcription service (Groq, with zero data retention switched on) and never stored by Noteer; only the text is kept. One honest caveat: Noteer detects secrets in text, so a password you say out loud reaches the transcription service before it can be caught. The text that comes back is checked like any typed note.