OC Rendering Block

Every other part of a character prompt describes what's in the picture. This one describes how it's painted. OC Rendering Block is the ComfyUI-OC-Generator node for visual style - 245 options spanning art style, quality, brush language, color, and image effects, all in one dropdown. Want "1980s anime," "acrylic paint," "absurdres," or an "album cover"? Pick it; the correct tag comes out the other side.

There's an honest quirk worth knowing up front: the pack's author notes that the raw dataset doesn't yet have a strong dedicated artist-string collection, so this block pulls its rendering entries from the 画面 (image/picture) category that the rest of the data already ships. In practice that means the selection is broad but leans on established booru rendering tags rather than named-artist strings. You get style families and mediums, not "in the style of [specific artist]." If that's what you need, this node handles it; if you were hoping for a curated artist library, it isn't there yet.

Groups and modes

The catalog splits into six subgroups: 画质 (quality, like absurdres), 艺术风格 (art style, like 1980s anime and 4koma), 素描 (sketch), 画笔 (brush/medium, like acrylic paint), 颜色 (colors), and 画面效果 (image effects). Entries render as 艺术风格 | 四格 -> 4koma, keeping the Chinese label and English tag together.

The controls match every other block in the pack:

• mode - fixed, randomize, increment, or decrement. The last two, combined with the step input, let you sweep through style options across runs.
• option - the static full-catalog dropdown.
• group - restricts the active pool to one subgroup.

Optional enabled and seed round it out. Outputs are the standard oc_block / prompt_preview / debug_text triad; oc_block feeds OC Generator, which places rendering at the very end of the assembled prompt - the last thing the model reads, right where style modifiers belong.

The group/option mismatch, one more time

Yes, it's here too: the dropdown is static, so if your selected option falls outside your chosen group, execution falls back to the first valid item in that group. With six groups and 245 options this is easy to trip, and the only visible trace is in debug_text - requested option not in current group; fallback applied. When your "watercolor" run comes back looking like line art, that's the first thing to check.

Install

The pack continues its no-drama theme: zero pip dependencies (its requirements.txt says so in a comment), zero model downloads, pure JSON catalogs plus Python.

cd ComfyUI/custom_nodes
git clone https://github.com/Zhuozhuo-219/ComfyUI-OC-Generator

Or grab "ComfyUI-OC-Generator" from ComfyUI Manager and restart. The node appears under OC Generator → Blocks. And the usual caveat: this is a prompt-text builder, so run the output through an anime SDXL model (Illustrious, NoobAI, Pony) where booru style tags are actually trained in.