Increase robustness of image loading check before sending canvas data.
Now retries up to 5 times with 200ms delays (1 second total) instead
of a single 100ms wait.
This fixes the 'Failed to get confirmation from server' error that
appeared when executing workflows immediately after ComfyUI restart,
before images finished loading from IndexedDB.
Prevents workflow execution failures due to timing issues during
canvas initialization.
The canvas was getting corrupted to a small strip because of confusion
between two different dimension types:
- Output area dimensions (logical working area, e.g. 512x512)
- Display canvas dimensions (actual pixels shown on screen)
Root cause: Setting canvas.width/height attributes to match output area
while also using CSS width:100%/height:100% created conflicts. When
zooming or reloading, wrong dimensions would be read and saved.
Fix: Remove canvas element width/height attribute assignments. Let the
render loop control display size based on clientWidth/clientHeight.
Keep output area dimensions separate.
This prevents the canvas from being saved with corrupted tiny dimensions
and fixes the issue where canvas would only show in a small strip after
zooming or reloading workflows.
Fix critical issue where LayerForge was trying to send canvas data
even when the node was bypassed (mode === 4). This caused unnecessary
errors and blocked workflow execution.
Now properly checks node.mode before attempting to send data via
WebSocket, skipping bypassed nodes entirely.
- Add image loading validation before sending canvas data to server
Prevents 'Failed to get confirmation' error when images haven't
finished loading after workflow reload. Waits 100ms and checks
if all layer images are complete before rendering output.
- Improve layer loading error handling in CanvasState
Better logging when layers fail to load from IndexedDB.
Allows empty canvas as valid state instead of failing.
- Add ResizeObserver for canvas container
Fixes bug where canvas only shows in top half of node.
Watches container size changes and triggers re-render to ensure
canvas dimensions are correctly calculated after DOM layout.
Ensure the canvas element is actually connected to the DOM before
handling paste events. This prevents LayerForge from capturing paste
events when navigating in subgraphs where the canvas is not visible.
Adds check for canvas.isConnected and document.body.contains() to
verify the canvas is part of the active DOM tree.
Prevent LayerForge from intercepting Ctrl+C, Ctrl+V, and other keyboard
shortcuts when the canvas is not focused. This was causing unwanted
popups and interfering with other nodes in ComfyUI.
Changes:
- Remove document.body focus check from handlePasteEvent
- Add focus validation to handleKeyDown before processing shortcuts
- Modifier keys (Ctrl, Shift, Alt, Meta) are still tracked globally
- All other shortcuts only trigger when canvas is focused
Fixes issue where paste events were captured globally regardless of focus.
Added quick start and workflow example links for easier navigation. Improved installation instructions and clarified manual install steps. Documented known incompatibility with Vue Nodes and provided guidance for reverting settings. Enhanced support section with actionable items.
Implemented grab icon feature in transform mode to move selected layers without changing selection, even when behind other layers. Added hover detection, cursor updates, and visual rendering in CanvasInteractions.ts and CanvasRenderer.ts.
Added proper backend validation for both config.json and model.safetensors to confirm model availability. Updated frontend logic to use /matting/check-model response, preventing unnecessary download notifications.
Add crossOrigin='anonymous' to image elements in CanvasState._createLayerFromSrc() method. This prevents canvas tainting when images are restored from IndexedDB after page refresh, ensuring export functions work correctly.
Add crossOrigin='anonymous' attribute to image elements in CanvasLayers.ts to prevent canvas tainting. This resolves the "Tainted canvases may not be exported" error when using the Send to Clipspace feature.
Implemented a comprehensive notification system with smart deduplication for LayerForge's "Paste Image" operations. The system prevents duplicate error/warning notifications while providing clear feedback for all clipboard operations including success, failure, and edge cases.
Implemented drag-to-resize functionality for the output area with visual transform handles on corners and edges. Users can now interactively resize the output area by dragging handles instead of using dialogs, with support for grid snapping and aspect ratio preservation.
Implements one-click auto adjustment of output area to fit selected layers with intelligent bounding box calculation. Supports rotation, crop mode, flips, and includes automatic padding with complete canvas state updates.
Corrected the overlay alignment issue on the canvas so that the preview mask now matches the actual mask positioning. This ensures consistent visual accuracy during editing.
Introduce a three-state checkbox in CanvasLayersPanel header to control visibility of all layers at once. Supports automatic state updates and integrates with renderLayers() for seamless layer management.
Fixed LayerForge fullscreen mode displaying behind ComfyUI interface elements by increasing z-index from 111 to 999999. Fullscreen mode now properly overlays all UI components as intended.
Removed duplicate code from CanvasIO.ts and replaced it with unified helpers from ImageUtils.ts. All tensor-to-image conversions and image creation now use centralized utility functions for consistency and maintainability.
Refactored mask scaling and drawing into the scaleImageToFit method in ImageUtils.ts. Updated CanvasIO.ts to use this utility, reducing code duplication and improving maintainability.
Updated mask loading to immediately use available data from connected nodes and preserve existing masks if none is provided. Backend mask data is only fetched after workflow execution, ensuring no stale data is loaded during connection.
Enhanced LayerForge input handling to strictly separate image and mask loading based on connection type. Images now only load when allowImage=true and masks only when allowMask=true, preventing unintended cross-loading between input types.
Fixed batch image processing to prevent duplicates and layer deletion while ensuring proper mask loading from input_mask. Images are now added as new layers without removing existing ones, and masks are always checked from backend regardless of image state.
Implemented dynamic brush cursor with visual feedback for size (circle radius), strength (opacity), and hardness (solid/dashed border with gradient). Added overlay canvas system for smooth cursor updates without affecting main rendering performance.
Implemented centralized modifier state management with ModifierState interface and getModifierState() method. This eliminates inconsistencies between event-based and state-based modifier checking across mouse, wheel, and keyboard interactions.
