samizdat
5ae463eb3b
- Implement verified Q2 2025 addDOMWidget() API for visual diagnostics - Add real-time histogram display with color-coded brightness levels - Add channel balance bars (R/G/B) and clipping warnings - Rename node from "why did i break?" to "why did it break?" - Canvas-based visualization with statistical analysis overlay - Creative team accessible interface with intuitive color coding
1.9 KiB
why did it break?
Diagnostic node that tells you what's actually happening to your image data when workflows start acting weird.
What it does
Analyzes IMAGE tensors and gives you both visual feedback and actual numbers about:
- Brightness distribution (min/max/mean/std)
- Clipping detection (pixels hitting 0 or 1)
- Channel balance (R/G/B analysis)
- Tensor health metrics
When to use it
Drop this node anywhere in your workflow when:
- Images look too bright/dark and you don't know why
- Colors seem off or washed out
- You suspect clipping or data corruption
- You need to measure the effect of processing nodes
- Your 400-node workflow is doing something weird
Parameters
image (IMAGE): The tensor you want to analyze
show_overlay (BOOLEAN): Display diagnostic info overlaid on the image
True: Shows brightness stats as text overlayFalse: Returns original image unchanged
analysis_depth (["quick", "thorough"]): How much to analyze
quick: Basic stats (brightness, clipping, channels)thorough: + percentiles, histogram data
Outputs
image (IMAGE): Original image with optional diagnostic overlay
report (STRING): Human-readable diagnosis like:
- "Very bright image (avg 0.89) - check for overexposure"
- "WARNING: 15.2% pixels clipped to white"
- "Color imbalance detected (R:0.45 G:0.67 B:0.32)"
stats (STRING): Machine-readable JSON data for logging/comparison
Usage example
Place after ImageBlend nodes to measure brightness amplification:
Load Image → ImageBlend (screen mode) → why did it break? → your workflow continues...
The overlay shows immediate visual feedback, the report explains what's wrong, and the stats give you precise numbers for systematic analysis.
Perfect for debugging FOUC, measuring control signal processing, or just understanding what your pipeline is actually doing to your data.