Capture

Capture turns what you already have into tasks. Paste notes, a list or plain sentences, drop a PDF (typed or scanned) or photos of a whiteboard, paste a screenshot, or type one sentence in a task composer. Every line becomes a task you can edit before importing: its title, a due date and a priority read from the words ("by Friday", "urgent"), and who it is for, read from an @name or a "Maya to" attribution and matched to a member of the workspace. Where the workspace has an AI provider, the model can read the notes instead and find tasks a line-by-line reading misses. An import can be undone at once.

• Every line of pasted notes is read as a task. Lines marked TODO:, Action: or [ ] are chosen first; headings, meeting details and ticked boxes are skipped
• Due dates and priorities are read from the words, in your own time zone
• An @name or a "Maya to" attribution becomes the assignee when exactly one member fits, and the row says so when nobody or more than one person does
• Each task can be edited before import: title, due date, priority and assignee
• Undo removes exactly the tasks an import created
• PDF and photo text is read in your browser in up to three of the ten Atlas languages, scanned PDFs included, and the text can be edited before tasks are read from it, with the lines it was unsure of listed for checking
• Find tasks with AI reads notes, and Read with AI reads a photo, handwriting included, with your workspace AI provider, where one is set up
• A single sentence in a task composer becomes a task with subtasks, a due date, a priority and labels

• Reading lines, dates, priorities and names, and reading the text of PDFs and photos, happens without any AI provider. Only Find tasks with AI and Read with AI send the text or the picture to the workspace AI provider, and only when you choose them.
• Text in photos and scanned PDFs is read in the languages chosen under Text language, up to three at a time. Each language downloads once and is kept in your browser; reading in more languages at once is slower.
• A name that matches nobody, or more than one member, leaves the task unassigned. Choose the assignee on the row before importing.
• Undo is offered right after an import. Tasks removed by it go where any deleted task goes.
• Voice capture records in the browser. Turning speech into text is done by your browser, and some browsers send the audio to their maker online to do it. When you start a recording, the browser asks for permission to use the microphone: choose Allow. If the microphone is blocked, missing or in use by another application, Atlas says which, and what to do.

1. 1Open Capture and choose the project the tasks go to.
2. 2Paste your notes. A task appears for every line, with marked lines chosen.
3. 3Check each task: change the title, the due date, the priority or the assignee, and untick what should not be imported.
4. 4Optionally choose Find tasks with AI to have your workspace AI read the notes instead, then check what it found.
5. 5Choose Import. If something is wrong, choose Undo straight away to remove the tasks that import created.
6. 6For a PDF or photos, open Capture PDF or Capture image, choose the text language, drop the files, paste a screenshot or take a photo on a phone, check the text it read, and import the same way.
7. 7For a spoken note, open Voice capture and record. Save it as one task, or choose Review as several tasks to check a task for each line of what was said and import them the same way.

Why is a line I wrote not chosen?

When your notes mark some lines as tasks (TODO:, Action:, [ ] or an @name), only those start chosen and the rest are listed as context. Tick any other line to import it too. When nothing is marked, every line starts chosen.

Does Capture send my text to an AI service?

Not unless you choose Find tasks with AI or Read with AI. Reading lines, dates, priorities and names, and the text of PDFs and photos, happens without one. Find tasks with AI sends the text, and Read with AI sends the picture, to the AI provider your workspace has set up, and the page shows which one.

Which languages can a photo or a scanned PDF be read in?

Any of the ten languages Atlas is available in: English, Spanish, Hindi, French, German, Japanese, Chinese, Arabic, Portuguese and Hebrew. Choose up to three under Text language; it starts with your Atlas language and English. A scanned PDF is read page by page when it has no text of its own. When Chinese is chosen with another language and a short note comes back with no Chinese characters, it is read again in Chinese alone.

Why did @alex not become the assignee?

Either nobody in the workspace is called that, or more than one person is. Atlas does not guess between two people. The row says so; choose the assignee there before importing.

How do I stop Capture guessing subtasks I did not want, in a single sentence?

State them explicitly. An explicit list always wins over what Atlas would infer, so writing the subtasks you want produces exactly those.