Nodes/ComfyUI-BrainDead/BD Qwen Identity Lock
ComfyUI Node

BD Qwen Identity Lock

Identity-focused Qwen encoder with strong character preservation using weighted multi-image input.

By BizaNator·Created 7 months ago·Updated 16 days ago· 14
BD Qwen Identity Lock
  • clip
  • vae
  • image1
  • image2
  • image3
  • conditioning
  • image1
  • image2
  • image3
  • latent
◄prompt—►
◄system_template<|im_start|>system CRITICAL IDENTITY PRESERVATION PROTOCOL: IMAGE 1 - PRIMARY IDENTITY SOURCE (MUST PRESERVE): Picture 1: <|vision_start|><|image_pad|><|vision_end|> This contains the CHARACTER IDENTITY that must be EXACTLY preserved: - PRECISE facial structure, bone structure, face shape - EXACT eye color, eye shape, eye spacing, eyebrow shape - IDENTICAL nose shape, nostril shape, nose bridge - EXACT mouth shape, lip fullness, lip color - PRECISE jawline, chin shape, cheekbone structure - IDENTICAL skin tone, skin texture, any marks/moles/freckles - EXACT hair color, hair texture, hairline, hair style IMAGE 2 - SECONDARY REFERENCE (APPLY SELECTIVELY): Picture 2: <|vision_start|><|image_pad|><|vision_end|> Apply ONLY non-identity aspects: pose, clothing, style, environment IMAGE 3 - ADDITIONAL CONTEXT (MINIMAL INFLUENCE): Picture 3: <|vision_start|><|image_pad|><|vision_end|> Use for composition/background only, NO identity influence GENERATION RULES: 1. Picture 1's face MUST appear UNCHANGED in output 2. Preserve Picture 1's exact identity regardless of other changes 3. Background should be clean, artifact-free, no noise 4. Maintain clear subject-background separation User request: {} Generate maintaining EXACT identity from Picture 1.<|im_end|> <|im_start|>assistant ►
◄image1_strength1.5►
◄image2_strength0.6►
◄image3_strength0.4►
◄image1_rolecharacter►
◄image2_rolestyle►
◄image3_rolereference►
◄enable_resizetrue►
◄enable_vl_resizetrue►
Category🧠BrainDead/Character

Inputs (15)

NameTypeDefaultDescription
clipCLIP—
promptSTRING—
system_templateSTRING<|im_start|>system CRITICAL IDENTITY PRESERVATION PROTOCOL: IMAGE 1 - PRIMARY IDENTITY SOURCE (MUST PRESERVE): Picture 1: <|vision_start|><|image_pad|><|vision_end|> This contains the CHARACTER IDENTITY that must be EXACTLY preserved: - PRECISE facial structure, bone structure, face shape - EXACT eye color, eye shape, eye spacing, eyebrow shape - IDENTICAL nose shape, nostril shape, nose bridge - EXACT mouth shape, lip fullness, lip color - PRECISE jawline, chin shape, cheekbone structure - IDENTICAL skin tone, skin texture, any marks/moles/freckles - EXACT hair color, hair texture, hairline, hair style IMAGE 2 - SECONDARY REFERENCE (APPLY SELECTIVELY): Picture 2: <|vision_start|><|image_pad|><|vision_end|> Apply ONLY non-identity aspects: pose, clothing, style, environment IMAGE 3 - ADDITIONAL CONTEXT (MINIMAL INFLUENCE): Picture 3: <|vision_start|><|image_pad|><|vision_end|> Use for composition/background only, NO identity influence GENERATION RULES: 1. Picture 1's face MUST appear UNCHANGED in output 2. Preserve Picture 1's exact identity regardless of other changes 3. Background should be clean, artifact-free, no noise 4. Maintain clear subject-background separation User request: {} Generate maintaining EXACT identity from Picture 1.<|im_end|> <|im_start|>assistant —
vaeoptVAE—
image1optIMAGE—
image2optIMAGE—
image3optIMAGE—
image1_strengthoptCOMBO1.510 options: 0.0, 0.2, 0.4, 0.6, 0.8, 1.0, +4
image2_strengthoptCOMBO0.610 options: 0.0, 0.2, 0.4, 0.6, 0.8, 1.0, +4
image3_strengthoptCOMBO0.410 options: 0.0, 0.2, 0.4, 0.6, 0.8, 1.0, +4
image1_roleoptCOMBOcharacter6 options: character, style, pose, environment, reference, control
image2_roleoptCOMBOstyle6 options: character, style, pose, environment, reference, control
image3_roleoptCOMBOreference6 options: character, style, pose, environment, reference, control
enable_resizeoptBOOLEANtrue—
enable_vl_resizeoptBOOLEANtrue—

Outputs (5)

NameTypeDescription
conditioningCONDITIONING—
image1IMAGE—
image2IMAGE—
image3IMAGE—
latentLATENT—