Quick ConDelta

This pack gives you several ways to get a concept into your conditioning, and each one trades convenience for control differently. GetConDeltaFromPrompt builds a proper delta by diffing your prompt against a baseline pool. ApplyConDelta needs a file saved to disk first. QuickConDelta skips both of those steps: type a prompt, set a strength, and it blends straight in - the fastest path in the whole pack from "I want to try adding this concept" to seeing it in your generation.

It's the positive-side mirror of CFGlessNegativePrompt in this pack. That node encodes a negative prompt and subtracts it from your conditioning at a strength; this one encodes a prompt and adds it, at a strength - both sharing the same default of 0.6 and the same shape of inputs, just pushing in opposite directions.

How it works

You give it your existing conditioning, a clip model, and a prompt describing whatever you want to blend in. The node encodes the prompt and adds the result into your conditioning, scaled by strength. No baseline subtraction, no file save - it's the crudest and quickest way to nudge a generation toward a concept, which is exactly the point of its name.

The inputs and outputs that matter

• conditioning - the conditioning you're adding to.
• clip - the CLIP model used to encode prompt.
• prompt - multiline text for the concept you want to blend in.
• strength - default 0.6, range -100 to 100. Positive blends the concept in; negative pushes away from it, so this one node can do double duty if you don't want to reach for CFGlessNegativePrompt separately.

Output is a single CONDITIONING.

How to install it

Via ComfyUI Manager: search "ComfyUI-ConDelta", install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/envy-ai/ComfyUI-ConDelta

then restart. No model files, no extra dependencies.

Common issues & troubleshooting

This node is for quick experiments, not your final workflow. Because it skips the baseline-subtraction step that GetConDeltaFromPrompt does, what you're adding is closer to "the raw encoded prompt" than a clean, isolated concept - it'll carry along whatever else is implicit in how you phrased it. If the result is muddier than you expected, that's the tradeoff for skipping the baseline; switch to GetConDeltaFromPrompt if you need something cleaner.

Keep strength modest. The default of 0.6 is already a meaningful nudge, not a small one - this isn't a 0–1 blend weight, it's added directly on top of your existing conditioning. Push much higher and you'll see the same burn and structural breakdown that shows up anywhere else in this pack when a strength gets cranked too far.

Once you've found a strength and prompt combination that works, consider promoting it. If you're reusing the same quick concept across multiple generations, it's worth building it properly with GetConDeltaFromPrompt and saving it with SaveConditioningDelta instead of retyping the prompt into this node every time.