If a conditioning area should stay in the same relative place and proportion after you change the canvas size, use ConditioningSetAreaPercentage. If it should keep the same pixel position and dimensions, use ConditioningSetArea. The two nodes use different coordinate systems: ConditioningSetArea takes integer width, height, x, and y; ConditioningSetAreaPercentage takes those values as percentages of the image dimensions. Pixel values do not automatically scale when the canvas changes.
Choose what must stay fixed
Use pixel inputs when the target is tied to a particular absolute region, such as a fixed-size patch measured in pixels. On a larger canvas, those same values still describe the same pixel-sized area at the same pixel offset, not the same fraction of the canvas.
Use percentage inputs when the target should move and resize with the canvas. The node defines x and width relative to image width, and y and height relative to image height. For example, a region starting 10% from the left and 20% from the top occupies the same relative placement on differently sized canvases when you retain those percentages.
This choice concerns the conditioning area's coordinates, not whether a generated image will obey the condition exactly. The documentation describes the nodes' inputs and outputs; it does not promise a particular visual result.
Calculate both canvas settings
Assume the first canvas is 800 × 600 pixels. The intended rectangle begins at (80, 120) and measures 240 × 180 pixels. These are hypothetical values for calculation, not observed ComfyUI output.
The rectangle occupies 30% of the canvas width and height, starts at 10% of its width, and starts at 20% of its height. If the second canvas is 1200 × 900 pixels, using those same percentages gives a rectangle starting at (120, 180) with dimensions 360 × 270 pixels. Its relative placement and size are preserved; its absolute pixel dimensions are not.
| Hypothetical canvas | Intended region in pixels: x, y, width, height | Percentage inputs: x, y, width, height | Expected region in pixels after applying percentages |
|---|---|---|---|
| 800 × 600 | 80, 120, 240, 180 | 10%, 20%, 30%, 30% | 80, 120, 240, 180 |
| 1200 × 900 | 120, 180, 360, 270 | 10%, 20%, 30%, 30% | 120, 180, 360, 270 |
For comparison, keeping the pixel inputs at 80, 120, 240, 180 on the second canvas leaves the region at that absolute offset and size. It no longer occupies 30% of the canvas width and height.
Set the node values
- Write down the canvas width and height for each intended workflow size, in pixels. Decide whether the region's absolute pixel bounds or its relative placement and size are the requirement.
- For fixed pixel bounds, add
ConditioningSetAreaand enter integerwidth,height,x, andyvalues. Check that the bounds fit the intended canvas: the right edge isx + width, and the bottom edge isy + height. - For a region that scales with the canvas, add
ConditioningSetAreaPercentage. Convert each pixel measurement using the corresponding canvas dimension:x ÷ canvas width × 100,width ÷ canvas width × 100,y ÷ canvas height × 100, andheight ÷ canvas height × 100. - Enter each value in the matching node's units. Do not copy percentage numbers into the pixel node or pixel counts into the percentage node. Keep the
strengthchoice separate from this geometry decision; neither node's documentation says that changing strength rescales the area. - Repeat the calculation for the second canvas size and compare the expected bounds before running the workflow.
The pixel-area node documentation identifies the four geometry inputs as integers. The percentage-area node documentation defines the corresponding values as percentages of image width or height.
Check bounds and correct mismatches
Use the table as a geometry check, not as evidence that a generated subject or detail will land precisely inside the area. For the 1200 × 900 hypothetical canvas, percentage inputs of 10%, 20%, 30%, and 30% imply left and top offsets of 120 and 180 pixels, with width and height of 360 and 270 pixels. If your calculation sheet still shows 80, 120, 240, and 180, you have carried over pixel values rather than recalculating the expected pixel bounds.
If an area extends beyond the canvas, check the relevant end bound: x + width must not exceed the canvas width, and y + height must not exceed its height. If the area is in the wrong relative place, recalculate the percentages from the intended rectangle and the correct canvas dimension; do not assume a width-based percentage also determines a vertical value.
These checks validate values against the documented input definitions. They do not test whether generation follows the conditioning as intended. Treat any visual discrepancy as a separate model-result review, not proof that the coordinate conversion was wrong.
Common questions
Should I use percentages if only the canvas width changes?
Use percentages if the region should retain the same fraction of each canvas dimension. Calculate horizontal values against width and vertical values against height; do not use the width as the denominator for every input. If only the canvas width changes while its height stays the same, the y and height percentage inputs and their expected pixel ranges stay unchanged; recalculate or recheck only the x and width pixel ranges against the new width.
What if I want the same pixel-sized region on both canvases?
Use ConditioningSetArea with the same integer bounds on both. That preserves absolute position and dimensions, not relative position or size.
Do percentage inputs guarantee the same visible result?
No. They define the conditioning area's geometry as percentages of the image dimensions. The documentation does not guarantee how strongly a generated result will follow that condition.
Sources and next step
- ConditioningSetArea - ComfyUI Built-in Node Documentation
- ConditioningSetAreaPercentage - ComfyUI Built-in Node Documentation
This guide interprets public documentation and uses explicitly hypothetical acceptance examples, not application tests. Controls and availability can change; verify the cited manual against your installed version.
For work created in Imagild, carry the input requirements and acceptance checklist into the photo editor. The external software controls discussed above belong to their respective products.