SD3.5 +LORA 文生图 基础


Updated:

Workflow Preview

Showcases (Image/Video)

SD3.5 +LORA 文生图 基础流

采用新一代的SD3.5 large基模,超级清晰:迭代步数15-20,CFG根据需要自己填写3-6。权重可以自己调试哦建议在.7-1.1之间。

2.主采样器:dpmpp_2m + sgm_uniform,以SD3.5L为底膜推荐的初步分辨率:832*1248

主采样器:dpmpp_2m + sgm_uniform

CFG:3 + |步骤:30 +

(推荐):CFG:5 | STEPS:40

  1. LORA 正向:

USAChrisHMS,solo,short hair,brown hair,shirt,black hair,1boy,jewelry,jacket,male focus,cowboy shot,open clothes,pants,facial hair,ring,blue shirt,beard,hand in pocket,realistic,leather,

4.建议搭配

stable-diffusion-3.5-large(SD3.5欧美休闲时装男模USAChrisHMS)

SD3.5 large ( https://huggingface.co/stabilityai/stable-diffusion-3.5-large/blob/main/sd3.5_large.safetensors

5.召唤词:USAChrisHMS

Nodes Detail

14 Nodes
Primitive Node Types
7
  • PreviewImage1
  • ConditioningCombine1
  • VAEDecode1
  • CheckpointLoaderSimple1
  • CLIPTextEncode2
  • KSampler1
Custom Node Types
7
  • ConditioningSetTimestepRange2
  • ModelSamplingSD31
  • TripleCLIPLoader1
  • LoraLoaderModelOnly1
  • ConditioningZeroOut1
  • EmptySD3LatentImage1