ComfyUI’s ImageBatch uses image1 as the size reference: when the inputs differ in dimensions, it automatically rescales image2 to match before combining them. That can change the second image’s proportions and apparent composition. The node may still run successfully; execution alone does not show whether the batch suits your workflow. Check the input order and both images’ dimensions before choosing whether to standardize them, swap them, or accept the resize.
What the node changes
The official documentation describes ImageBatch as combining two images into one batch. It identifies image1 as the reference for dimensions and says that image2 is rescaled when their dimensions differ. The output is a batch, not a side-by-side canvas. ComfyUI: ImageBatch
For a hypothetical example, suppose image1 is 1024 × 1024 and image2 is 1600 × 900. Because the dimensions differ, ImageBatch rescales image2 to match image1’s dimensions before batching. The documentation does not specify the resize algorithm, so the output dimensions alone do not establish whether the second image’s proportions or visible composition changed. Inspect the resulting image to determine whether its content is acceptable.
Diagnose the actual inputs
Follow these steps in order:
- Identify which upstream connection feeds
image1and which feedsimage2. Trace the wires in the workflow; do not infer the order from a preview layout or file names. - Read each input’s actual width and height in the image-loading or upstream output information available in your workflow. Record the values immediately before
ImageBatch. - Compare the dimensions. If they differ, expect
image2to be resized to matchimage1according to the documented behavior. - Inspect the batch output’s dimensions and each item’s contents. If your workflow provides a way to preview or inspect batch items separately, check both rather than treating the batch as one ordinary image.
- Decide which input should govern the dimensions, or whether neither input should be resized automatically.
Choose a preparation route
Use one of three routes, based on the intended result:
- Standardize dimensions before batching: If both images must keep a particular canvas size or shape, use an upstream
ImageScalenode or another appropriate preparation step to set each input deliberately. Then inspect the prepared images before connecting them toImageBatch. This makes the size change an explicit part of the workflow; it does not guarantee that the original composition or all details survive the preparation. - Change input order: If one image’s dimensions should govern the batch and you are willing to have the other image resized, connect the intended reference to
image1. Confirm the actual wire order afterward. This is a choice about which input sets the dimensions, not a general fix for distortion or content loss. - Accept the automatic resize: If matching the first image’s dimensions is acceptable for your task, keep the order and verify the second image’s resulting content. This avoids an extra preparation step but accepts the documented size adjustment.
ComfyUI documents the image1 and image2 inputs as IMAGE and the output as an IMAGE batch. That description establishes the dimension-matching behavior, but it does not specify an aspect-ratio-preserving mode, a crop setting, or a way to exempt image2 from the adjustment. Check the controls of any separate scaling node you use; do not attribute its options to ImageBatch. ComfyUI: ImageBatch inputs and output
Verify the batch, not just execution
Use a comparison record to determine which image changed. The values below are hypothetical examples, not a measured run. “Observed output” means what you would record from your own workflow.
| Check | Hypothetical input or observation | Decision |
|---|---|---|
image1 dimensions | 1024 × 1024 | This is the documented size reference. |
image2 dimensions | 1600 × 900 | Dimensions differ; expect this input to be rescaled. |
| Batch output dimensions | 1024 × 1024 per item, if observed | Confirm actual output rather than assuming the graph’s successful execution proves it. |
image1 content | Compare subject placement and visible edges with its input | Record whether the reference item remains suitable. |
image2 content | Compare proportions, framing, and important details with its input | If unacceptable, standardize inputs or reconsider which image is image1. |
The output dimensions in the table are a hypothetical value to check, not a claim about a test result. Pass the workflow only when the batch contains the expected images at the intended dimensions and the content of each item is acceptable for its next operation. A successful node execution does not by itself confirm either condition.
If the result is still wrong
If both source images already have matching dimensions but the second image still looks different, first confirm that you are inspecting the correct batch item and that the values recorded are from the connections entering this ImageBatch. The documented resize behavior is triggered by differing input dimensions; it does not explain every later change in a workflow. Inspect subsequent nodes separately rather than attributing their effects to ImageBatch.
If the inputs differ and neither image should lose its intended framing, prepare both deliberately before batching and review the prepared results. If you swap the inputs, check again which one is now in image2; a different ordering can move the resize to the other image. No route can be assumed to restore details that were removed or altered during a prior transformation. The ImageBatch documentation does not establish a particular resize method or composition-preserving option, so judge the resulting images directly.
Common questions
Does ImageBatch place the images side by side?
The documented output is a batch of images, not a single side-by-side composition. A later workflow node would be needed for a different layout.
If I swap the inputs, will both images keep their original sizes?
No. When their dimensions differ, the image connected to image2 is the one the documentation says will be resized to match image1. Swapping changes which image is in that position; it does not remove the dimension-matching behavior.
Should I use `ImageScale` first?
Use it when you need to choose input dimensions explicitly before batching. Then inspect both prepared images and the batch output. This recommendation is a workflow decision, not a claim that a particular ImageScale setting preserves every image’s composition.
Sources and next step
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.