ReAuraPatcherAdvanced

This is the "I want to know exactly which layers the style guide touches" version of ReAuraPatcher. The plain patcher applies RES4LYF's style-transfer machinery to your whole AuraFlow model, on or off, and that's genuinely all most people need. This one exists for when the default is close but not quite right - when the style is bleeding into detail it shouldn't, or composition, and you want to narrow where in the network that guide signal is allowed to act.

Why block selection matters

Transformer diffusion models like AuraFlow process an image through a stack of blocks, and roughly speaking, earlier blocks handle broad composition while later ones refine fine detail and texture. RES4LYF's style patchers can inject their guide signal into a specific subset of those blocks instead of all of them - restrict it to the early "double layer" blocks and you're steering composition and coarse style without stomping on fine detail; restrict it to late "single layer" blocks and it's closer to a texture/finish pass. This is squarely power-user territory: the payoff is real control over a stubborn style-transfer result, and the cost is that you need to actually inspect (or experiment with) what each block range is doing on your specific model, since there's no universal map of "these blocks are composition, these are texture" that holds across every AuraFlow-family checkpoint.

The inputs and outputs that matter

• model (MODEL, required) - your loaded AuraFlow model.
• doublelayer_blocks (STRING, default "all") - which of AuraFlow's double-layer transformer blocks get patched. Leave it at "all" until you have a specific reason to narrow it; when you do, you'll pass a comma-separated list or range of block indices.
• singlelayer_blocks (STRING, default "all") - same idea for the single-layer blocks.
• style_dtype (default float64) - precision for the style computation itself, same as the non-advanced patcher.
• enable (BOOLEAN, default true) - bypass switch.

Output: a patched model, same as ReAuraPatcher.

How to install it

Ships with RES4LYF, no separate install. Via ComfyUI Manager: search "RES4LYF", install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/ClownsharkBatwing/RES4LYF/
cd RES4LYF
pip install -r requirements.txt

Restart ComfyUI afterward. You'll need your own AuraFlow weights, since RES4LYF doesn't ship any model files - grab fal/AuraFlow-v0.3 from Hugging Face if you don't have a copy already.

Common issues & troubleshooting

You have no idea which blocks to pick. That's normal - start from "all" (the equivalent of the plain ReAuraPatcher), confirm the effect looks like what you expect, then narrow one range at a time rather than guessing both fields at once. Change one variable, keep the seed fixed, compare.

Malformed block strings silently do nothing or error. These are free-text fields with no dropdown validation, so a typo or an out-of-range index is on you to catch. If narrowing the blocks seems to have zero effect, double check the string actually parsed the way you meant before concluding the technique doesn't work here.

You're reaching for this before trying the plain patcher. Don't. doublelayer_blocks/singlelayer_blocks at "all" is functionally the same result as ReAuraPatcher with more surface area for you to misconfigure. Only step up to the advanced node once you've confirmed the basic style patch is working and you specifically need finer control over where it acts.

Reminder on the wider context: AuraFlow itself is a niche, largely dormant model outside of Pony V7 - if you're not deliberately working in that corner, the effort spent tuning block ranges here is probably better spent on a patcher for a model family with an actual active ecosystem.