Apply Fooocus Style

Fooocus was lllyasviel's bet that the way to win new users was to hide every knob and ship good defaults. It worked - through 2024 it was the standard "what should a beginner install" answer, and its styles were part of the recipe. Apply Fooocus Style hands you those six styles as prompt templates you can drop onto any SDXL-family workflow in ComfyUI.

Here's the quirk that surprises people: two of the six - Fooocus Enhance and Fooocus Negative - have no positive template at all. They're pure negative-prompt packs. Because the node defaults an empty positive template to your prompt unchanged, selecting "Fooocus Enhance" leaves your positive alone and simply tacks a ~740-character negative onto yours. That's not a bug; that's the whole design. The Fooocus negative is a carefully tuned block - quality keywords with attention weights, anti-watermark, anti-blur, anti-text - and you get to steal it for your own workflows.

The rest of the list: Fooocus Cinematic and Fooocus Sharp wrap your prompt in "cinematic still... emotional, harmonious, vignette" and are decent for photoreal-ish renders, Fooocus Masterpiece slaps on (masterpiece), (best quality), (ultra-detailed) for the anime-SDXL crowd, and Fooocus Photograph pushes you toward a 50mm photo look with its own long negative.

How it works

Identical mechanism to every node in this pack: pick a style from the dropdown, the node replaces {prompt} in the template's positive and appends the template's negative to yours. Strings in, strings out, runs before CLIP Text Encode. No models, no downloads, no API.

Inputs and outputs that matter

Same five inputs as the rest of the family:

• positive_prompt / negative_prompt - your prompt strings (multiline).
• style - 7 entries: bypass plus the six Fooocus styles.
• apply_negative_style (default yes) - matters more here than anywhere else in this pack. With "Fooocus Negative" or "Fooocus Enhance" selected, flipping this off makes the node a literal no-op. It's the only knob that decides whether these styles do anything at all.
• log_prompt (default off) - dumps before/after to the console; worth enabling once to see a 700-character negative land.

Outputs are positive_prompt and negative_prompt (STRING). Positive goes to your CLIP Text Encode positive, negative to its negative. If you're running a dual-encoder setup (SDXL-style), the negative string wants to feed the same encoder chain as your positive.

Install

Part of the ComfyUI_mozman_nodes pack - install it once and you get all seven style nodes. ComfyUI Manager (search ComfyUI_mozman_nodes), or:

cd ComfyUI/custom_nodes
git clone https://github.com/mozman/ComfyUI_mozman_nodes

Then restart. There's no requirements.txt, no model files, pure stdlib - by Manfred Moitzi (the ezdxf author), MIT-licensed, with tests. The rare custom node that can't break your environment.

Where people get burned

The Fooocus styles are SDXL through and through. On guidance-distilled or LLM-encoded models, the negative half is inert at CFG 1 and the (term:1.4) weight syntax gets thrown away - which guts "Fooocus Enhance" and "Fooocus Negative" specifically, since they're negative-only. Save this node for SDXL, Illustrious, Pony, or NoobAI.

And one honest caveat: Fooocus the app froze in 2024 (bug-fixes-only since then), so these styles are not being tuned for newer checkpoints. They still hold up on the SDXL family because the model family they were written for never really moved. If you're already on the modern-generation train, skip this one.