ClownGuide Mean
Match the overall color and tone of a reference
- guide
- mask
- weights
- guides
- guides
ClownGuide Mean is the gentlest guide in the RES4LYF style family. It matches the mean of a reference latent - the overall color cast, brightness, and tonal center - and pulls your generation toward it, without touching structure or fine detail. Think of it as a very soft color-and-mood match rather than a full style transfer.
Like the rest of the pack's guides, it doesn't render anything on its own. It builds a GUIDES bundle that plugs into the sampler's guides input and nudges denoising toward the reference while sampling runs. Guides chain, so a Mean guide is a natural first or last link - a light tonal correction stacked under a stronger AdaIN or attention guide, or used alone when all you want is for a batch of images to share a consistent palette.
How it works
Where AdaIN matches both mean and variance and attention injection rewrites Q/K/V, Mean matches only the average. It computes the mean of the guide latent and shifts the generation's own mean toward it by the amount you set. That's a deliberately limited operation: it can shift a whole image warmer, cooler, brighter, or darker to sit in the reference's tonal range, but it won't import texture, brushstrokes, or composition. The default scheduler here is beta57 - the pack's signature schedule - and there's a cutoff that limits how far the matching pushes.
The inputs and outputs that matter
guide(LATENT, optional) - the reference whose tone you're matching to.weight(FLOAT, default0.75) - how strongly to pull toward the reference's mean. The main dial; 0.75 is already a moderate pull, so ease off for subtlety.cutoff(FLOAT, default1, range 0–1) - caps the extent of the shift, a guard against overcorrection.weight_scheduler(enum, defaultbeta57) - how the influence ramps across steps.start_step/end_stepandmask/invert_mask- the shared step-window and regional controls.
Output is a guides (GUIDES) bundle for the next guide node or the sampler.
How to install it
ComfyUI Manager: search RES4LYF, install, restart. Manual: cd ComfyUI/custom_nodes && git clone https://github.com/ClownsharkBatwing/RES4LYF/, then pip install -r requirements.txt in your venv (portable ComfyUI: embedded python's pip). Restart and hard-refresh with F5. For the nested sampler menu, add rgthree-comfy with "Auto Nest Subdirectories" enabled.
Common issues
The most common disappointment is expecting too much from it. This is not a style-transfer node - it will not make your Flux render look like a Rembrandt, it'll just steer the palette. If you wanted an actual look transferred, the AdaIN or attention-injection guides are the right tools, and Mean is the finishing touch on top. Set expectations and it's genuinely useful; expect a miracle and it'll seem to "do nothing."
Turn the weight up too far, on the other hand, and you can flatten or wash out the image as its tones get dragged toward the reference's average - that's what the cutoff is there to prevent, so lower weight or tighten cutoff if colors go muddy. And the pack-wide scheduler note holds: if you switch weight_scheduler to bong_tangent and things behave oddly, that's the known sigma-shift quirk, not this node. Keep a fixed seed while you dial it in so you can actually see the tonal shift rather than seed variance.
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| weight | FLOAT | 0.75-100–100 | Set the strength of the guide. |
| cutoff | FLOAT | 1.000–1 | Disables the guide for the next step when the denoised image is similar to the guide. Higher values will strengthen the effect. |
| weight_scheduler | COMBO | beta57 | 12 options: constant, simple, sgm_uniform, karras, exponential, ddim_uniform, +6 |
| start_step | INT | 00–10000 | — |
| end_step | INT | 15-1–10000 | — |
| invert_mask | BOOLEAN | false | — |
| guideopt | LATENT | — | |
| maskopt | MASK | — | |
| weightsopt | SIGMAS | — | |
| guidesopt | GUIDES | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| guides | GUIDES | — |