Visual Context & Restore
Whispree cares about both “what you’re looking at while you speak” and “where the result should go”. Visual context feeds vision-model correction; context restore handles the precise insertion location.
Visual context (vision)
Section titled “Visual context (vision)”When using a vision-capable correction model, capturing the screen during recording and passing it to the model lets you get more accurate corrections that reflect the terms, code, and names on screen.
Settings → LLM (shown when a vision model is selected):
| Setting | What it does | Default |
|---|---|---|
| Enable screenshot context | Captures the screen during recording (app switch / scroll / click debounced, up to 20 shots) and passes it to the vision model. Requires Screen Recording permission. | Off |
| Pass to agent | After inserting text, also pastes the captured screenshots into the target app as images. | Off |
Capture works with intelligent debounce — switching apps or scrolling/clicking resets the timer, and a 1.5-second pause triggers a capture (up to 20 shots, to protect memory).
Screenshot selection panel
Section titled “Screenshot selection panel”When transcription finishes, a panel appears for choosing which screenshots to paste (in FIFO order if multiple recordings are queued):
- ↑ ↓ to move · Space to select · Enter to confirm · ⌘Enter to preview (Quick Look style) · Esc to skip.
Instant toggle during recording (⌥)
Section titled “Instant toggle during recording (⌥)”While recording, hold the left Option key for ~0.5 seconds to toggle “attach screenshots to the target app” on and off. Img Attach ON/OFF flashes briefly in the overlay (only when a vision model is active).
Context restore
Section titled “Context restore”The spot you were in when you started dictating is restored exactly at insertion time. Rather than simply pasting into “whatever app is frontmost now”, it returns to the location at the moment recording started.
Browser restore (Chrome)
Section titled “Browser restore (Chrome)”Browser restore (on by default) under Settings → General. It remembers the Chrome tab and the focused input element and cursor position at the start of recording, then returns to that tab and that input field before inserting.
- Chrome only. Requires Automation permission.
- Element-level cursor restore requires Chrome’s “Allow JavaScript from Apple Events”. Even with it off, tab restore still works (only element restore is skipped).
Terminal restore (iTerm2 / tmux)
Section titled “Terminal restore (iTerm2 / tmux)”Terminal restore (on by default) under Settings → General. It remembers the iTerm2 session (split pane) and, if tmux is attached inside it, the tmux window/pane as well, then restores to that spot.
- iTerm2 only. Requires Automation permission. tmux supports only the default socket.
- Terminal.app · Alacritty · Kitty · Ghostty · Warp do not support split/session restore (app focus restore only).
For the required permission setup, see Permissions.