ComfyUI-Krea2-AnimeStyle

Comfy-Krea2-AnimeStyle

ComfyUI custom node for grouped anime illustration style prompt injection.

The node appears under:

eastmoe -> Comfy-Krea2-AnimeStyle

Node

Krea2 Anime Style CLIP Text Encode

Inputs:

• clip: the CLIP/text encoder used for final conditioning.
• positive_text: user positive prompt.
• lora_trigger_text: LoRA trigger words prefixed directly to the final positive prompt, without translation or refinement.
• negative_text: user negative prompt.
• style_selection: saved selected style IDs. In the ComfyUI frontend this is rendered as grouped checkboxes.
• translate_prompts: when enabled, translates the user prompts to English through the connected CLIP-LLM before encoding.
• translation_max_length: max CLIP-LLM generation length, capped at 1024.
• refine_prompt: when enabled, expands and polishes the positive prompt through the connected CLIP-LLM before appending styles.
• refinement_max_length: max CLIP-LLM generation length for positive prompt refinement, capped at 1024.

Outputs:

• positive: positive conditioning with selected style English descriptions appended.
• negative: negative conditioning.

Krea2 Anime Style Prompt Text

Same prompt processing and style selector, but it returns strings instead of encoded conditioning. The clip input is optional and is only required when translate_prompts or refine_prompt is enabled. This is intended for subgraphs or any workflow that wants to pass the processed prompt text into another node.

Outputs:

• positive_text: positive prompt string with selected style English descriptions appended.
• negative_text: negative prompt string after optional translation.

Style data is generated from C:\Users\StarMoon\Downloads\anime_illustration_glossary.html into data/styles.json.