ghosthand screentext
Usage
ghosthand screentext
Options
| Option | Description |
--json | Print the full JSON (lines with confidence, screen size) |
--language [tag] | OCR language, e.g. it-IT; en-US is always included. Env: GHOSTHAND_LANGUAGE |
--settle | Wait (up to 2s) until the screen stops changing |
--scale [n] | Image and coordinate size as a fraction of native pixels (default: screen points, 1/2 or 1/3; 1 = native). Env: GHOSTHAND_SCALE |
Also takes the global options.
Details
Output: one line per text,
x,y wxh text, top-left corner and size in screen
points. The center of a line is x+w/2, y+h/2. To tap text, use
ghosthand tap --text "...", which does this for you.
Add --language for non-English UIs (it-IT, de-DE, ...) or accents are misread.
Body text and labels read at about 99% of words. Not read: red badge counts
(Mail "6"), icons (read as stray letters like "Q", "A", "4"), text cut by
an edge. Text inside photos comes out garbled. Use a screenshot for these.Examples
$ ghosthand screentext -l 42 $ ghosthand screentext -l 42 --language it-IT --json
MCP tool
The same command is the MCP tool
screentext; flags are arguments.