process-image
Public
Forked from ceveyne/process-image
Forked from ceveyne/process-image
Project Files
docs / CHANGELOG.md
Notable changes to this project will be documented in this file.
Zoom Step x/y (n%) instead of bare Step x/y in refine, , and .maxWidth, maxHeight).inputs.canvas.original now correctly shows the Pass-1 backend_returned dimensions instead of the pre-Pass-1 source canvas.w + h ≤ 1792 px). zoom-in now reads attachment dimensions from stored state instead of re-loading the original file.frameAdjust (detect path): %-values now expand each bbox dimension proportionally (width by pct%, height by pct%), preserving aspect ratio.zoom-in audit log: now correctly shows the original source file, its pre-crop dimensions, and the prompt used during restoration.crop with detectLabel + imageFormat: now applies symmetric AR expansion, matching the existing zoom-in behaviour.inpaint / outpaint: when the canvas is a detect_object result with more than one detection and no detectLabel is given, a clear error is returned listing the available labels instead of silently using the full image as the mask region.inpaint / outpaint tool descriptions now document model, detectLabel, detectIndex, frameAdjust, and mask-coordinate auto-pickup from prior mask runs.zoom-in with a mask canvas (iN where sourceTool is mask): now correctly loads the source from cropSource, applies the stored crop coordinates, and does not use the annotated image as the zoom target.frameAdjust now applies in the manual crop path (without detectLabel) for crop, mask, and zoom-in. Previously it was silently ignored unless detectLabel was used.frameAdjust: extended accepted string formats to include % suffix and +/- prefix (e.g. "30%", "+5%", "-10px").zoom-in: added imageFormat parameter to control output aspect ratioinpaintoutpaintrender_target.needs_upscaler now reflects the actual zoom-pass decision instead of being recomputed from raw dimensions versus backend limits.width/height parameters now preserve their native aspect ratio instead of adopting the requested output AR.crop with detectLabel + imageFormat: the AR expansion (e.g. portrait 3:4) was silently skipped because all four crop sides were pre-filled by resolveDetectionCrop. applyCropFormat now applies symmetric AR expansion in this case, matching the existing zoom-in behaviour.