HunyuanImage 3.0 Resolutions:
HunyuanImage3Resolutions, displayed in ComfyUI as “HunyuanImage 3.0 Resolutions,” is a preset-size selector for HunyuanImage 3.0 workflows. It lets you choose a size from the model’s resolution table instead of manually entering dimensions in several places. Its main benefit is consistency: the text encoder’s image geometry and the empty latent’s dimensions need to agree about the image grid, and a shared preset helps keep them synchronized. The node works by reading the selected preset’s dimension label and converting it into two whole numbers; it does not generate images, resize existing images, or perform model inference. It intentionally offers presets only, keeping standard model-table sizes convenient while leaving custom-size control to the encoder and empty latent nodes.
HunyuanImage 3.0 Resolutions Input Parameters:
resolution
A required COMBO selection that determines the image dimensions returned by the node. The available choices are generated by preset_options(BASE_SIZE), and the default is specified by DEFAULT_RESOLUTION. The supplied source does not show the expanded option list, the numeric value of BASE_SIZE, or the actual default label, so exact preset dimensions, minimum dimensions, and maximum dimensions cannot be confirmed from this excerpt. Each selectable label is interpreted as widthxheight, with a lowercase x separating two whole numbers. Selecting a different preset changes the dimensions passed to connected nodes, allowing you to choose the framing and size appropriate to your composition. This is a preset-only selector with no custom-size option; execution simply converts the selected dimensions to integers rather than calculating a new size or searching for a nearest match.
HunyuanImage 3.0 Resolutions Output Parameters:
width
An INT representing the selected image’s horizontal dimension in pixels. This is the first number in the selected widthxheight label, converted into a whole number without additional scaling or adjustment. Connect it to compatible width inputs on the HunyuanImage 3.0 encoder and empty latent nodes so they describe the same horizontal image grid. The value specifies a requested dimension; it is not image data and does not, by itself, resize or create an image.
height
An INT representing the selected image’s vertical dimension in pixels. This is the second number in the selected widthxheight label, converted into a whole number without additional scaling or adjustment. Connect it to compatible height inputs on the same downstream nodes that receive the width, keeping the complete size synchronized throughout the relevant workflow. Together, width and height define the selected image’s proportions: equal values describe a square, a larger height describes a portrait orientation, and a larger width describes a landscape orientation.
HunyuanImage 3.0 Resolutions Usage Tips:
- Connect both outputs to the corresponding width and height inputs on the HunyuanImage 3.0 encoder and empty latent nodes. Sharing both dimensions is more reliable than manually matching separate settings.
- Choose your orientation before generation: use an available preset with width greater than height for landscape compositions, height greater than width for portrait compositions, or equal dimensions for square compositions.
- When changing presets, check that every relevant downstream size input receives the updated dimensions. A remaining manually configured dimension can leave part of the workflow using a different image grid.
- Use the encoder’s and empty latent’s own width and height controls when you need an arbitrary custom size. This selector deliberately does not provide a custom entry, and modifying its preset label is not a substitute for configuring custom dimensions properly.
- If your workflow offers several preset sizes, remember that larger image grids generally require more downstream processing and memory. This selector itself only parses a label; choosing a smaller available preset can reduce generation demands, not the selector’s already minimal work.
HunyuanImage 3.0 Resolutions Common Errors and Solutions:
ValueError: not enough values to unpack (expected 2, got 1)
- Explanation: The selected value does not contain the lowercase
xseparator expected by the node. This could occur if a preset value is altered through an API request or a modified workflow; normal dropdown selections should already have the correct format. - Solution: Select a valid preset from the dropdown again. For API-based workflows, use an exact option exposed by the installed node rather than constructing your own label.
ValueError: too many values to unpack (expected 2)
- Explanation: The supplied label contains more than one lowercase
x, so the node cannot separate it into exactly two dimension values. - Solution: Restore a valid dropdown preset. Avoid appending extra dimensions or descriptive text to the selected value.
ValueError: invalid literal for int() with base 10
- Explanation: At least one part of the selected label cannot be converted into a whole number. Added units, descriptive text, or decimal dimensions can cause this Python error; its full message includes the offending value.
- Solution: Choose an unmodified preset. If you supply the selection programmatically, copy a valid option exactly and do not add units such as “px.”
Encoder and empty latent dimensions do not match
- Explanation: This is a workflow configuration problem, not an error message raised by this node. If the encoder and empty latent use different dimensions, they can disagree about the image grid, potentially causing downstream failures.
- Solution: Connect this node’s width output to both width inputs and its height output to both height inputs. Check that the two outputs have not been swapped and that no separate size controls override the intended dimensions.
Out-of-memory error during generation
- Explanation: This is a possible downstream generation issue, not a failure of the resolution selector. A selected image size may demand more memory than your generation setup has available.
- Solution: If the dropdown offers a smaller preset, select it and update all connected size inputs together. Diagnose any remaining memory error in the model or generation node that reports it; this node does not allocate image latents or run inference.
