Apply ClipDrop Style

Let's get the most important thing out of the way: Apply ClipDrop Style does not call ClipDrop, does not call any Stability API, and needs no API key. ClipDrop is Stability's web product with the fancy style controls; this node is just the prompt-text version of its style list, shipped as local JSON. If you were worried about an API dependency or a key field - there isn't one. It's a text transform that runs entirely on your machine.

The set itself is 16 styles with friendly human names: Anime, Photographic, Digital Art, Comicbook, Fantasy Art, Analog Film, Neonpunk, Isometric, Lowpoly, Origami, Line Art, Craft Clay, Cinematic, 3D Model, Pixel Art, and Texture. If those look familiar, they should - this is essentially the official SDXL sai-* preset list, trimmed, with the hyphens and "sai-" prefixes stripped off.

So the honest take: if you already have Apply SDXL Style SAI from this same pack, this node is largely redundant. The differences are cosmetic (readable names, a "Texture" entry that SAI also has). It's only "extra" if you specifically want the plain-name list without digging through slugs. Since it ships in the same pack as SAI, you get it either way - but don't install the pack for this node.

How it works

Identical to its siblings. Pick a style from the dropdown, the node replaces {prompt} in that style's template with your positive and appends the template's negative to yours. Strings in, strings out, runs before CLIP Text Encode. No model loading, no network, no downloads - the whole pack is pure Python stdlib.

Inputs and outputs that matter

• positive_prompt / negative_prompt - your prompt strings (multiline).
• style - 17 entries: bypass plus the 16 styles. bypass is always first and passes both prompts through untouched.
• apply_negative_style (default yes) - off if you're keeping your own negative.
• log_prompt (default off) - prints before/after to the console.

Outputs are positive_prompt and negative_prompt (STRING), wired into your CLIP Text Encode node.

Install

Part of the ComfyUI_mozman_nodes pack. ComfyUI Manager (search ComfyUI_mozman_nodes), or:

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

Then restart. No requirements.txt, no model files - the rare custom-node install with literally nothing else to download. By Manfred Moitzi (the ezdxf author), MIT-licensed, with unit tests.

Where people get burned

The main trap is expecting something this node doesn't do. Because of the name, people assume it reproduces ClipDrop's actual stylization - it doesn't, and it can't; it's a prompt wrapper, not a style-transfer model. Set expectations accordingly: it adds the "analog film" phrasing, it doesn't grade your image.

And the usual family caveat: these are SDXL-CLIP prompts. On guidance-distilled or LLM-encoded models the negative half is inert at CFG 1 and the phrasing transfers poorly. Keep it on SDXL, Illustrious, Pony, or NoobAI and it behaves as advertised.