If you want two prompts to describe one blended result, consider an averaging route only after checking the documentation for the specific averaging node you plan to use. If you want two conditioning inputs concatenated into one output, ComfyUI’s ConditioningConcat documentation confirms that it appends the conditioning_from vectors to the conditioning_to vectors. Neither description guarantees how a model will interpret the result visually. Decide what relationship you intend, then compare outputs without changing other workflow settings.
Choose the relationship first
“Both prompts should influence the image” is not precise enough to choose a node. Ask whether you want the two inputs represented as a blend, or whether you want their conditioning vectors joined in a single output. Those are different data operations, not interchangeable promises about image content.
Use ConditioningConcat when concatenating the inputs matches your intended representation. Its documented inputs are conditioning_to and conditioning_from, and its output is a unified set of conditioning vectors. The documentation does not say that the inputs receive equal weight, that one becomes more important, or that particular objects from both prompts must appear.
The official ConditioningAverage node page cited in this guide describes blending two conditioning sets and lists conditioning_to_strength with a range of 0.0–1.0, a default of 1.0, and a step of 0.01. Check the current documentation for the exact node and ComfyUI version in your workflow before relying on those requirements or behavior. This node documentation does not guarantee a particular visual result or establish the behavior of other nodes or versions.
Run a controlled comparison
Use the following sequence to distinguish a change in combination operation from unrelated workflow changes:
- Write two prompts that make the intended relationship clear. Keep them short enough that you can identify what each contributes.
- Record the model, text-encoding path, sampler settings, seed, dimensions, and other conditioning inputs in the workflow. Keep those settings unchanged across comparison runs.
- Save a baseline using the current conditioning path. Record the exact prompts and node connections so you can restore the same setup.
- Make one comparison branch with
ConditioningConcat. Connect the primary input toconditioning_toand the additional input toconditioning_from, following the documented roles for this node. - An averaging branch is optional, not a required counterpart. Only add one after identifying the specific
ConditioningAveragenode and checking its current documentation and input requirements for your ComfyUI version. The officialConditioningAveragenode page cited in this guide listsconditioning_to,conditioning_from, andconditioning_to_strength(range 0.0–1.0, default 1.0, step 0.01); do not assume other nodes or versions share these details. If you cannot verify the node documentation, compare only the baseline andConditioningConcatbranches. - Generate and label each result. Note whether the intended contributions are present, absent, or ambiguous, without treating one image as proof of a general rule.
This procedure controls the comparison; it does not predict which branch will look better. A fixed seed helps keep one variable under review, but does not make the visual outcome certain.
Hypothetical acceptance record
Suppose prompt A is “a red bicycle” and prompt B is “a glass greenhouse.” The creator wants to record whether each route meets a predefined composition brief. This is an author-recommended comparison and a hypothetical worksheet, not an official test or a prediction of what a model will draw. Fill the observation and decision cells only after running that branch; add the optional averaging row only after verifying the specific node’s documentation and inputs. Until then, leave its cells blank.
| Branch | Combination | Settings held fixed | Observation to record after a run | Decision against the brief |
|---|---|---|---|---|
| Baseline | Existing conditioning path | Model, seed, sampler, dimensions | Record what is present, absent, or ambiguous | Apply the predefined success criterion |
| A | ConditioningConcat | Same settings as baseline | Record what is present, absent, or ambiguous | Apply the predefined success criterion |
| Optional B | Specific documented averaging node and verified inputs; otherwise do not run | Same settings as baseline | Leave blank until run; record observation without predicting it | Apply the predefined success criterion |
The observations above are deliberately hypothetical. If your own results differ, record what happened rather than changing the table to match a presumed node effect. Decide in advance what counts as success: for example, both subjects recognizable, both present but with one dominant, or a deliberate spatial relationship. “Both prompts influenced it” is not a sufficiently observable acceptance criterion.
When the result is unclear
If one contribution is missing or the composition is hard to interpret, first check that each input reaches the intended node socket and that the output is connected to the same downstream path in both branches. Then verify that only the combination operation changed. Re-run the comparison before changing prompt wording, seed, sampler, or other settings; otherwise you will not know which change affected the observation.
If the outcome still does not express the intended relationship, revise the brief to describe the desired composition more explicitly, or choose a different documented workflow. Do not infer from one sample that concatenation or averaging always gives a particular prompt priority.
The official ConditioningConcat page describes vector concatenation, not a visual blending rule or a guarantee that every input detail survives. The cited official ConditioningAverage page describes blending two conditioning sets using conditioning_to_strength and lists that input’s range, default, and step. These descriptions are limited to the documented nodes; they do not guarantee a particular image effect or establish the behavior of other nodes or versions. Check the current documentation for the exact node in your workflow before relying on its requirements or behavior.
Common questions
Does `ConditioningConcat` give both prompts equal influence?
The documented behavior is to concatenate conditioning_from vectors into conditioning_to vectors. It does not specify equal influence or a visual priority rule. Do not treat concatenation as a weighting control.
Should I use averaging whenever I want a blend?
Only after verifying the exact averaging node’s documentation and that its inputs fit your workflow. “Average” in a node name is not enough to establish a formula or guarantee a blended-looking result.
What if I cannot verify an averaging node’s documentation?
Run the baseline and ConditioningConcat comparison only. Do not treat an undocumented averaging branch as an equivalent, reproducible route; leave it out until you identify the exact node and verify its current documentation.
Can I compare the routes by changing prompts and seed at the same time?
You can explore that way, but it will not isolate the combination operation. For a useful first comparison, keep the prompts and workflow settings fixed, change only the combination route, and describe the result as an observation from that run.
Sources and next step
- ConditioningConcat - ComfyUI Built-in Node Documentation
- ConditioningAverage - ComfyUI Built-in Node Documentation
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.