Nodes/ComfyUI-BrainDead/BD Qwen Multi-Image
ComfyUI Node

BD Qwen Multi-Image

Multi-image Qwen encoder with weight mode control for character-consistent image editing.

By BizaNator·Created 7 months ago·Updated 16 days ago· 14
BD Qwen Multi-Image
  • clip
  • vae
  • image1
  • image2
  • image3
  • conditioning
  • image1
  • image2
  • image3
  • latent
prompt
system_template<|im_start|>system You are analyzing multiple reference images for character-consistent image editing. MULTI-IMAGE ANALYSIS PROTOCOL: Picture 1: <|vision_start|><|image_pad|><|vision_end|> - Primary reference for character identity Picture 2: <|vision_start|><|image_pad|><|vision_end|> - Secondary reference for style/pose Picture 3: <|vision_start|><|image_pad|><|vision_end|> - Additional context reference CHARACTER CONSISTENCY (FROM PICTURE 1): - Face structure, facial features, eye color/shape - Hair style, color, texture - Skin tone and distinctive marks - Body proportions and build STYLE TRANSFER (FROM PICTURE 2/3): - Pose and composition - Clothing style (if requested) - Environment and lighting - Art style and rendering TASK: Combine elements while maintaining character identity from Picture 1. User request: {} Generate image maintaining character consistency, Ultra HD, 4K quality.<|im_end|> <|im_start|>assistant
enable_resizetrue
enable_vl_resizetrue
weight_modediffusers
Category🧠BrainDead/Character

Inputs (10)

NameTypeDefaultDescription
clipCLIP
promptSTRING
system_templateSTRING<|im_start|>system You are analyzing multiple reference images for character-consistent image editing. MULTI-IMAGE ANALYSIS PROTOCOL: Picture 1: <|vision_start|><|image_pad|><|vision_end|> - Primary reference for character identity Picture 2: <|vision_start|><|image_pad|><|vision_end|> - Secondary reference for style/pose Picture 3: <|vision_start|><|image_pad|><|vision_end|> - Additional context reference CHARACTER CONSISTENCY (FROM PICTURE 1): - Face structure, facial features, eye color/shape - Hair style, color, texture - Skin tone and distinctive marks - Body proportions and build STYLE TRANSFER (FROM PICTURE 2/3): - Pose and composition - Clothing style (if requested) - Environment and lighting - Art style and rendering TASK: Combine elements while maintaining character identity from Picture 1. User request: {} Generate image maintaining character consistency, Ultra HD, 4K quality.<|im_end|> <|im_start|>assistant
vaeoptVAE
image1optIMAGE
image2optIMAGE
image3optIMAGE
enable_resizeoptBOOLEANtrue
enable_vl_resizeoptBOOLEANtrue
weight_modeoptCOMBOdiffusers3 options: comfy, diffusers, raw

Outputs (5)

NameTypeDescription
conditioningCONDITIONING
image1IMAGE
image2IMAGE
image3IMAGE
latentLATENT