Robota uses ScreenCaptureKit to capture system audio (other meeting participants). This requires the "Screen & System Audio Recording" permission on macOS. Robota never records video, and it only ever captures an image of your screen when you press the snapshot button yourself during a recording. On macOS Sequoia (15+), Apple may ask you to re-confirm this permission monthly.

Robota mixes both audio streams (mic + system) 50/50 before transcription, which eliminates most echo. If you're still seeing duplicates, make sure you're using headphones - without them, your speakers bleed into your mic and can cause duplicates even with the mixing approach.

Apple's SpeechAnalyzer models are managed by the system and downloaded on demand. The first time you use a language, macOS needs to download the model (typically a few hundred MB). This happens in the background. If the download is slow, check your internet connection and try again - subsequent uses will be instant since the model is cached.

1. In Robota settings, set AI provider to "GGUF model".
2. Choose Add model, then Download from Hugging Face. Enter the repository and exact GGUF filename, then click Download.
3. When the download finishes, click Use to make it active. Summaries now run entirely on-device.

This runs a full AI model inside Robota using a bundled copy of llama.cpp, with no server or separate app needed, and is fully offline after the one-time download.

Already have model files? Choose Add model, then Link a model folder. Pick a compatible text .gguf from the list. Robota uses it in place without copying or downloading it.

Everything in Robota runs on your Mac, so the limit is your Mac's memory. A model needs about as much memory as its file size, plus some room for the meeting it is reading. Your meeting app and browser need memory too, so aim for a model file under about a third of your Mac's memory.

• 8 GB: stay with Apple Intelligence, or use a small model of 1 to 3 billion parameters (a 1 to 2 GB file), such as Qwen3 1.7B or Llama 3.2 3B.
• 16 GB: a model of 4 to 8 billion parameters (a 3 to 5 GB file) is a good balance of quality and speed.
• 32 GB or more: models of 12 to 14 billion parameters (an 8 to 9 GB file) write clearly better summaries.

When a model comes in several sizes, pick the file with Q4_K_M in its name. It is about a quarter the size of the full model and loses very little quality.

Good to know:

• Long meetings. Apple Intelligence can only read a few pages at once, so Robota summarizes a long meeting in parts and then joins them. A downloaded model can read far more in one go (up to about 32,000 tokens, if the model and your memory allow), which often gives a more joined-up summary. If a meeting is still too long, Robota trims it and the model is told so.
• The first summary is slower. The model loads into memory the first time you use it. It is freed again after 5 minutes of rest. You can change this with "Free memory when idle" in Settings. Choose Never if you have memory to spare and want every summary to start at once.
• Thinking models. Models like Qwen3 "think out loud" before they answer, which is slow. Robota hides that text, but you can skip it entirely by adding /no_think as the prompt suffix in Settings.
• Live summary. The live summary during a call runs the model every minute, while your Mac is also transcribing. On a Mac with 16 GB or less, leave it off and let the summary run once the meeting ends.
• Disk space. Downloaded models stay on your Mac until you delete them. Remove ones you no longer use in Settings to get the space back.

While a recording is running, the snapshot button (the camera icon on the recording pill) saves a picture of whatever window is in front - usually the slide someone is sharing. The button tells you when the image is saved.

Each snapshot is filed at the moment you took it, so it appears in the meeting's Markers & snapshots card in the Library. Click a thumbnail to open the full-size image, or click its timestamp to jump to that point in the transcript. Snapshots stay in the app and are not included in exports.

Nothing is captured unless you press the button. Robota never takes a snapshot on its own.

Robota captures audio passively at the macOS system mixer level via ScreenCaptureKit. It does not inject into or interact with the meeting app in any way. Other participants cannot detect that Robota is running. Please ensure you comply with local laws regarding recording consent.

In Robota settings, set your Obsidian vault path and subfolder (defaults to Meetings/Robota). After transcription, click the "Obsidian" button in the review toolbar to export. You can also enable auto-save to export automatically after every meeting. Notes include YAML frontmatter, formatted transcript, summary (if generated), and bookmarks.

Robota requires macOS 26 (Tahoe) or later. This is because it uses Apple's SpeechAnalyzer framework for on-device transcription, which is only available on macOS 26+. Robota is an Apple Silicon app and does not run on Intel Macs.

When you reopen Robota after an interrupted recording or transcription, it offers to recover the meeting. Choose Recover to finish transcription, Keep saved transcript to use the last saved live text, or Later to leave the files in place. You can return through Recover recordings in the Meeting Library. Notes, agenda items, and bookmarks are kept with the recording.

Checkpoints run every five seconds and when notes are saved. An abrupt exit can lose edits since the last successful checkpoint. If saving fails, check your free disk space and retry recovery. Failed or empty transcriptions keep their audio until you recover or discard them. Discarding permanently removes the unfinished recording and its saved content. If removal fails, Robota lets you retry deletion.

Unfinished audio and recovery checkpoints are kept under Application Support/com.worldtiki.robota/recording-sessions/ in the app's container. Audio is deleted after the meeting archive is saved and checked, or when you discard the recording. Recovery folders are excluded from backup. Saved meetings - transcripts, summaries, notes, and any snapshots - live under ~/Library/Application Support/com.worldtiki.robota/meetings/ until you delete them. Settings are stored at ~/Library/Application Support/com.worldtiki.robota/settings.json. If you export to Obsidian, markdown files are saved to your vault. No data is ever sent to any server. See the privacy policy for full details.