Repository navigation
Scanpath Studio shows how people read. It draws eye-tracking-while-reading data over the text at its position on the screen, replays each reading, compares participants, and exports figures for a paper. This is a beta: the app, the Python API and the CLI may still change during 0.x.
Get it
- Desktop app, for your own data: the Windows, macOS and Linux downloads under Assets below.
- With pip (Python 3.11–3.14):
pip install -U scanpath-studio, thenscanpath-studio. - In the browser: try the bundled demo online.
Documentation · Ask a question · Report a bug · Cite
Not included yet: reading measures computed by the app from fixations. Corpus Analysis shows the measures your dataset brings.
Added
- Word IDs can be text, such as
w1orHello_3: fixations are matched to the word with the same ID in their trial, so a dataset keyed by text IDs loads without fixation coordinates, and its IDs are shown and exported as the file spells them. (#415)
Changed
- Moving between trials sends about a quarter less to the browser and takes about a third less server time, and on a slow connection the app's styling now arrives within a few seconds of the page instead of about 20 seconds after it. (#506)
Fixed
- A layer switch you flip right after picking a design preset now stays flipped instead of snapping back, and the first option you pick in a layer's ▾ menu after a preset is no longer undone. (#485)
- Words that read "None", "NA" or "null", and blank word cells, in a CSV or TSV words table are kept as words under pandas 3.1 too, instead of being read as missing. (#524)
- A built-in dataset whose column mapping cannot load now shows Save changes on its Edit dataset screen, so a field fixed there takes effect. (#525)