Ooro

Ooro runs beside a call or an in-person conversation. It listens to the microphone and the computer's audio and writes down who said what. While you talk, it keeps a notebook: what has been decided, what is still open, and what would be worth asking now. You can stay in the conversation, and afterwards you have a record you can check line by line. Everything runs on your computer: the app needs no account, API key or server.

Follow the conversation as it happens

The transcript appears while people talk, with each speaker kept apart. Conversations can be in English or German. You can pick the language for each conversation or let Ooro detect it. Each finished line is transcribed a second time for accuracy and replaces the first version about 0.3 seconds later.

A notebook that keeps up

As the conversation goes on, the assistant works through your agenda and keeps the notebook up to date:

• the bottom line
• decisions
• action items, with owners and due dates
• key facts
• open questions

It also suggests what is worth asking right now: for example, who approves the extra cost if on-site training pushes the pilot past 48,000 euros. When the conversation ends, it writes a final summary in English or German.

Every note has its source

Every item the assistant writes cites the transcript lines it is based on, and each citation is one click away. If a citation doesn't point to a real line, the item is dropped instead of shown.

Name a voice once

Name a speaker once, and Ooro recognises that person by voice in later conversations. A match stays a suggestion until you confirm it, because a wrong name is worse than no name.

A record you can search

Every conversation is kept in a searchable library. Each person has their own page that shows their voice status and every conversation they took part in.

You can also:

• import recordings, which go through the same pipeline as a live conversation, faster than real time
• edit transcript lines and reassign speakers
• export to Markdown, text, JSON, SRT or VTT, or make an encrypted backup

Local by default

Audio, transcripts, people and the built-in models stay on your machine:

• The vault is encrypted with SQLCipher, and audio is encrypted with AES-256-GCM.
• The encryption key lives in the operating system's credential store.
• You choose the audio sources for each conversation.

The assistant runs on a built-in model. If you'd rather use a model on your own server, or Claude with your own API key, only the transcript text and the notes leave the machine, never the audio. Conversations that use one of these options are labelled as such.

The built-in models come from a fixed list:

Ooro downloads them once and checks each file against a pinned SHA-256 hash before using it.

Where it stands

Ooro is in pre-release. Builds are shared by invitation, and there is no public download yet. The main target is macOS on Apple Silicon; Windows and Linux builds exist but haven't been validated on real devices.

Please get consent from everyone in the conversation before you record. Ooro reminds you to.