Replace STRING widget with fresh COCO.PATH widget object to prevent
LiteGraph default "Value" prompt from appearing alongside custom file
browser. Bump version to 0.8.0. Simplify .gitignore and move dev
files to dev/. Add test images and example workflow for release.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Server-side /cocotools/browse API endpoint returns directory listings
with extension filtering. JS modal file browser injects browse buttons
on ImageLoader, LoadExr, LoadExrSequence, and SaverNode path widgets.
Zero changes to node INPUT_TYPES — fully backward compatible.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Update pyproject.toml version from 0.5.0 to 0.7.0 to reflect actual
semver history since 0.4.2.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds a clickable Canvas2D preview widget to CryptomatteLayer that displays
the connected beauty pass, maps mouse clicks to pixel coordinates for matte
generation, and overlays a diagonal zebra stripe pattern on selected pixels.
- Add preview_image optional IMAGE input and OUTPUT_NODE for UI data
- Create js/cryptomatte.js with click handler, crosshair, and hover coords
- Fix image output to always produce valid [1,H,W,3] IMAGE tensors
- Search all cryptomatte ranks and ID channels for hash matching
- Extract layer name resolution into _resolve_layer_name method
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add optional x_coord/y_coord inputs for pixel selection
- Sample cryptomatte float32 hash at clicked position
- Generate binary mask matching all pixels with same object ID
- Add JS click handler to map preview clicks to pixel coordinates
- Existing functionality preserved when coords are -1 (default)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
P8: preserve original layer name in fuzzy-match debug logs
P9: preserve negative values in ACES colorspaces
P10: use actual channel count in reshape_from_processing
P11: add max version counter (9999) to get_unique_filepath
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
P3: normalize int32 PIL I-mode values by dtype max instead of raw passthrough
P4: clamp alpha to [0,1] instead of min-max stretching
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Step-by-step processing details were flooding the console during normal
ComfyUI operation. Demote ~45 info-level logs to debug, consolidate
multi-line error blocks, and remove redundant duplicate messages. Only
one clean summary line per major node operation remains at INFO level.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
PIL's convert("RGB") silently truncates to 8-bit, destroying precision
for high bit depth TIFFs and PNGs. Bypass it for I;16/I/F modes by
reading directly into numpy and normalizing per detected bit depth.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add channels < 3 guard matching histogram pattern.
Replace all var declarations with const/let for consistency.
Add batch_size to image_info for parity with histogram node.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Wrap execute body in try/except with error logging
- Handle < 3 channel edge case by repeating channels
- Remove dead-code guard on bins assignment
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Detect bit depth from original image mode before RGB conversion.
Fix double-normalization when normalize=True.
Preserve constant images instead of zeroing them.
Fix alpha mask tensor shape to match ComfyUI MASK type [B,H,W].
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Only replace #### placeholders, not arbitrary digits in filenames.
Require 2+ consecutive # in filename portion for sequence detection.
Support variable-width frame patterns (###, ####, #####).
Prefer last numeric sequence for frame number extraction.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CoCo