Skip to main content
The processing selector above the canvas controls the page scope and stages. Start with the smallest scope that covers the work.

​
Choose a scope

Process the page currently shown on the canvas.

​
Choose stages

Select any non-empty set of stages, then choose Run. Selected stages follow this dependency order: Detection creates regions and removal masks. OCR reads those regions. Translation writes target text. Inpainting reconstructs artwork under the lettering.
Omitted prerequisites are not added automatically. Include detection or OCR when a later stage needs fresh input. Run Inpainting means detection plus inpainting; it does not include OCR or translation.
The Process menu also offers Process Project, Process Selected Pages, and Run Detection, Run OCR, Run Translation, and Run Inpainting. A Run item runs every stage up to the named one and always covers the entire project, regardless of the scope selector.

​
Follow progress and stop work

Only one processing job runs at a time, and pages cannot be imported while one is running. The activity center shows the stage, model, progress, and failures. Each stage commits when it finishes. Completed work remains available if a later stage fails. Stopping prevents new work from starting; a native call already in progress returns at a safe boundary. Its result is discarded if stop was requested before the commit. Stages with no applicable input complete without inference, although the model is still loaded first. A page with no detected text does not need OCR or translation.
Rerunning a stage does not discard your edits. Detection skips any page that already has a text region, so delete the existing regions to detect a page again. OCR and translation replace generated text but keep source text and translations you edited. Inpainting skips a page that already has a cleanup layer unless you paint a new Remove mask.
Continue with text review or cleanup.