Conditioning Upscale (Inspire)

If you build workflows with area conditioning - prompts pinned to specific rectangles of the canvas via Conditioning (Set Area) - and then upscale, you've hit the annoyance this node exists for. The region coordinates are stored in pixels tied to the original resolution. Upscale the latent and those boxes don't grow with it, so your carefully placed regions end up in the wrong spot at the wrong size. Conditioning Upscale expands them by the same factor you're upscaling the image, so everything stays where you put it.

Like its sibling Conditioning Stretch, it comes from Davemane42's ComfyUI_Dave_CustomNode, folded into the Inspire Pack so you get the one node without adopting a whole extra dependency.

How it works

Area conditioning carries a box - a width, height, and offset - that scopes a prompt to part of the frame. This node multiplies that box by a scalar. Feed it the same number you scaled your image by (2 for a 2× upscale) and the regions land in the right place at the right proportion. The prompt text is untouched; only the geometry moves.

The catch worth stating plainly: if the conditioning has no area attached, there's nothing to scale and the node is a no-op. It only matters inside a regional / set-area setup.

The inputs that matter

There are only two:

• conditioning - your area-based conditioning going in; a single CONDITIONING comes out to wire into the sampler.
• scalar - the multiplier, default 2. Set it to match your upscale factor. (It's typed as an integer with a 0.5 step, a small quirk of the node - think of it as "how many times bigger.")

Pick Conditioning Upscale when you're doing a straightforward uniform upscale and one number says it all. If your aspect ratio changes or you want to specify exact from/to resolutions, use Conditioning Stretch (Inspire) instead - same idea, more explicit control.

How to install it

Through ComfyUI Manager: Install Custom Nodes → search ComfyUI Inspire Pack → install → restart. Or clone it directly:

cd ComfyUI/custom_nodes
git clone https://github.com/ltdrdata/ComfyUI-Inspire-Pack

Restart afterward. The pack is Dr.Lt.Data's - the ComfyUI Manager and Impact Pack author - so it's well cared for and installs cleanly. A few pip deps are fetched on first run; no model files needed for this node.

Common issues

The mistake people make is treating this as an image upscaler. It's not - it rescales conditioning boxes only. You still need a real latent-upscale (or a new empty latent at the target size) elsewhere in the graph; this node just keeps the regional prompts aligned once you have. And make sure the scalar here actually matches whatever upscale ratio you used on the latent. If it doesn't, your regions will be scaled to a resolution the image never reached, and they'll sit in the wrong place - which usually looks like the region "leaking" or missing its target entirely.