XB-BOX - Image Params Master
One hub for image size, batch, and strength — with dims that can't go invalid
- Image Width
- Image Height
- Batch Size
- Float Control
- Int Control
- Scale Size
Image Params Master is the "stop scattering your image settings across six primitive nodes" node. Instead of hunting for the width INT, the height INT, the batch INT, and the denoise/strength FLOAT scattered around a workflow, you get one box that owns all of them and hands clean values out on labeled wires. It's the image half of the same "unified parameters" idea that drives the pack's video-side master nodes.
If your workflow is basically stable and you just want to twiddle resolution and strength in one spot - or you're building a template for someone else to use - this node earns its keep immediately. It's also the kind of hub that makes a Dashboard-style front-end or a parameter template trivially easy to build on top of.
How it works
The mechanic is dimension hygiene plus consolidation. You set width, height, and batch once. Pick an aspect ratio from the dropdown and the node snaps your dimensions to the closest multiple of 16 that matches the ratio - so Free passes values through untouched, while 16:9, 9:16, 4:3, 3:4, and 21:9 force the pair to a valid, aligned ratio. That's the "enforces safe steps to prevent invalid inputs" from the README, and it genuinely saves you from the classic "why is my latent 1024x1023" failure.
The strength input is offered twice - once as a float and once as an int - because some downstream nodes want a percentage-like float and others want a per-mille-style integer. Wire whichever your sampler wants and ignore the other.
The inputs that matter
- aspect_ratio -
Free(pass-through) plus six locked ratios. Choose a locked ratio when you want the node to own the geometry. - width / height - 64–8192, default 1024². The values you actually type.
- batch_size - 1–1000, default 1.
- strength_float - 0–10 float, default 1.0.
- strength_int - 0–1000 int, default 1.
The outputs
Six wires, all self-explanatory: Image Width, Image Height, Batch Size, Float Control, Int Control, and Scale Size. Scale Size is the small bonus - it's max(width, height), precomputed for any node that wants the longer edge as a single number (upscalers, tiled VAE decoders, that sort of thing). Wire the first four into your sampler's latent and conditioning and you're done.
Install
Standard XB_ToolBox install: ComfyUI Manager → search XB_ToolBox, or git clone https://github.com/wjluoxiao/XB_ToolBox.git into custom_nodes, restart. No dependencies, no models.
Worth knowing: it's a pure parameter node, so nothing executes on the GPU and there's no "wrong" way to wire it - the failure mode is forgetting to update width after switching the ratio and wondering why the output is the size it is. Check the ratio dropdown first, always.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| aspect_ratio | COMBO | Free | 7 options: Free, 1:1, 16:9, 9:16, 4:3, 3:4, +1 |
| width | INT | 102464–8192 | — |
| height | INT | 102464–8192 | — |
| batch_size | INT | 11–1000 | — |
| strength_float | FLOAT | 1.000–10 | — |
| strength_int | INT | 10–1000 | — |
Outputs (6)
| Name | Type | Description |
|---|---|---|
| Image Width | INT | — |
| Image Height | INT | — |
| Batch Size | INT | — |
| Float Control | FLOAT | — |
| Int Control | INT | — |
| Scale Size | INT | — |