Nodes/ComfyUI-ArchAi3d-Qwen/ArchAi3D_Qwen_Encoder_Simple
ComfyUI Node

ArchAi3D_Qwen_Encoder_Simple

A ComfyUI node in ArchAi3d/Qwen with 15 inputs and 3 outputs.

By amir84ferdos·Created 10 months ago·Updated 4 months ago· 65
ArchAi3D_Qwen_Encoder_Simple
  • clip
  • vae
  • image1_vl
  • image2_vl
  • image3_vl
  • image1_latent
  • image2_latent
  • image3_latent
  • conditioning
  • latent
  • formatted_prompt
prompt
system_prompt
conditioning_strength1.00
image1_latent_strength1.00
image2_latent_strength1.00
image3_latent_strength1.00
debug_modefalse
CategoryArchAi3d/Qwen

Inputs (15)

NameTypeDefaultDescription
clipCLIPQwen-VL CLIP model for tokenization and encoding
promptSTRINGUser prompt text (vision tokens added automatically)
system_promptSTRINGOptional system prompt (uses ChatML format: <|im_start|>system...when provided)
conditioning_strengthFLOAT1.000–2Global conditioning strength multiplier (1.0=normal, >1.0=stronger, <1.0=weaker). Acts like CFG weight.
image1_latent_strengthFLOAT1.000–2[EXPERIMENTAL] Multiplies image1 reference latent values (1.0=normal, >1.0=stronger, <1.0=weaker)
image2_latent_strengthFLOAT1.000–2[EXPERIMENTAL] Multiplies image2 reference latent values (1.0=normal, >1.0=stronger, <1.0=weaker)
image3_latent_strengthFLOAT1.000–2[EXPERIMENTAL] Multiplies image3 reference latent values (1.0=normal, >1.0=stronger, <1.0=weaker)
debug_modeBOOLEANfalsePrint detailed info to console (conditioning shapes, strengths applied, etc.)
vaeoptVAEVAE for encoding reference latents (optional, needed for latent outputs)
image1_vloptIMAGEImage 1 for Qwen-VL vision encoder (RGB channels only, no resize)
image2_vloptIMAGEImage 2 for Qwen-VL vision encoder (RGB channels only, no resize)
image3_vloptIMAGEImage 3 for Qwen-VL vision encoder (RGB channels only, no resize)
image1_latentoptIMAGEImage 1 for VAE encoding as reference latent (RGB only, standard 4D format)
image2_latentoptIMAGEImage 2 for VAE encoding as reference latent (RGB only, standard 4D format)
image3_latentoptIMAGEImage 3 for VAE encoding as reference latent (RGB only, standard 4D format)

Outputs (3)

NameTypeDescription
conditioningCONDITIONINGEncoded conditioning with vision tokens and reference latents
latentLATENTimage1_latent encoded output in standard format (compatible with VAEDecode)
formatted_promptSTRINGFinal formatted prompt sent to the model (includes vision tokens and ChatML)