LoRA models

About LoRA models in biniou

• Since commit 540a437, biniou offers experimental support for LoRA models on Stable Diffusion SD 1.5/SDXL-based models :
  • Currently, only Stable Diffusion, LCM, img2img, IP-Adapter, ControlNet and IP-Adapter FaceID modules are compatibles.
  • You can find the LoRA options for these modules at the very bottom of their settings panels.
  • After selection of a model in the model list, only LoRA compatibles with the model version will be displayed in the LoRA models list.
  • Select a LoRA model, and, if necessary, set its strength.
  • If a trigger keyword is needed (and known !) for the LoRA model, it will be automatically added to the prompt.

Adding LoRA models

• You can add customs or downloaded from Hugging Face / CivitAI LoRA models, by copying their .safetensors file to the directory :

  • For SD 1.5 : ./biniou/models/lora/SD/
  • For SDXL : ./biniou/models/lora/SDXL/

• You should now relaunch biniou and refresh the webui to see your models in the list. Please also note that you must append the trigger keyword manually to the prompt if required.

Built-in LoRA models list (by Stable Diffusion model version):

• SD 1.5 :

  • Kvikontent/midjourney-v6
  • Norod78/SD15-IllusionDiffusionPattern-LoRA

• SDXL :

  • openskyml/dalle-3-xl : Mimic Dall-e 3 rendering
  • ByteDance/SDXL-Lightning : Speed up generation using this LoRA model
  • ostris/face-helper-sdxl-lora : Generate realistic photo portraits of fictional characters
  • artificialguybr/ps1redmond-ps1-game-graphics-lora-for-sdxl : PS1 graphics
  • ProomptEngineer/pe-old-school-cartoon-style : Generate old school cartoons images
  • goofyai/3d_render_style_xl : beautiful 3d render style
  • artificialguybr/StickersRedmond : Stickers-like style
  • TheLastBen/William_Eggleston_Style_SDXL : Generate photographies in the style of William Eggleston.
  • Pclanglais/Mickey-1928 : Mickey 1928 (Mickey Minnie Pete)
  • Norod78/SDXL-YarnArtStyle-LoRA : Yarn style
  • KappaNeuro/1987-action-figure-playset-packaging : Generate action figure packages
  • KappaNeuro/director-tim-burton-style : Give your creation a Tim Burton style
  • KappaNeuro/vintage-postage-stamps : Create vintage stamps
  • KappaNeuro/diorama : Create images of diorama
  • artificialguybr/movie-poster-redmond-for-sd-xl-create-movie-poster-images : Generate movie posters
  • KappaNeuro/movie-poster : Generate movie posters vintage style
  • DoctorDiffusion/doctor-diffusion-s-xray-xl-lora
  • openskyml/soviet-diffusion-xl : Generate vintage Soviet posters
  • HarroweD/HarrlogosXL : Logos creation
  • SvenN/sdxl-emoji : Generate Emoji "Apple-style"
  • KappaNeuro/cute-animals : Create cute animals
  • ostris/ikea-instructions-lora-sdxl : Create fake Ikea instructions notices
  • ostris/super-cereal-sdxl-lora : Create fake cereal boxes
  • Norod78/SDXL-JojosoStyle-Lora-v2 : Generate anime images in the style of JoJo : Stone Ocean.
  • Norod78/SDXL-simpstyle-Lora-v2 : Give your images a Simpson style.
  • Norod78/SDXL-Caricaturized-Lora : Create caricatures
  • Norod78/SDXL-StickerSheet-Lora : Create a full sheet of stickers

Examples :

• Module : Stable Diffusion
• Prompt : A photo portrait of a beautiful woman laughing, looks happy and fullfilled, RAW photo, ultra realistic, intricate, highly detailed, sharp, clean
• Negative Prompt : None (SDXL-turbo model)

Examples of effects of LoRA models on images generated with the same seed, SDXL-Turbo model and defaults settings. Results are NOT cherrypicked.

Settings	Result
Model=stabilityai/sdxl-turbo | XL model=True | Sampler=Euler a | Steps=1 | CFG scale=0 | Size=512x512 | GFPGAN=True | Token merging=0.6 | nsfw_filter=True | Seed List=8852281397 - 00:01:10 (70 seconds) - LoRA model=None
Model=stabilityai/sdxl-turbo | XL model=True | Sampler=Euler a | Steps=1 | CFG scale=0 | Size=512x512 | GFPGAN=True | Token merging=0.6 | nsfw_filter=True | Seed List=8852281397 - 00:01:13 (73 seconds) - LoRA model=openskyml/dalle-3-xl
Model=stabilityai/sdxl-turbo | XL model=True | Sampler=Euler a | Steps=1 | CFG scale=0 | Size=512x512 | GFPGAN=True | Token merging=0.6 | nsfw_filter=True | Seed List=8852281397 - 00:01:09 (69 seconds) - LoRA model=KappaNeuro/director-tim-burton-style
Model=stabilityai/sdxl-turbo | XL model=True | Sampler=Euler a | Steps=1 | CFG scale=0 | Size=512x512 | GFPGAN=True | Token merging=0.6 | nsfw_filter=True | Seed List=8852281397 - 00:01:14 (74 seconds) - LoRA model=KappaNeuro/1987-action-figure-playset-packaging
Model=stabilityai/sdxl-turbo | XL model=True | Sampler=Euler a | Steps=1 | CFG scale=0 | Size=512x512 | GFPGAN=True | Token merging=0.6 | nsfw_filter=True | Seed List=8852281397 - 00:01:06 (66 seconds) - LoRA model=Norod78/SDXL-Caricaturized-Lora
Model=stabilityai/sdxl-turbo | XL model=True | Sampler=Euler a | Steps=1 | CFG scale=0 | Size=512x512 | GFPGAN=True | Token merging=0.6 | nsfw_filter=True | Seed List=8852281397 - 00:01:09 (69 seconds) - LoRA model=Norod78/SDXL-JojosoStyle-Lora-v2
Model=stabilityai/sdxl-turbo | XL model=True | Sampler=Euler a | Steps=1 | CFG scale=0 | Size=512x512 | GFPGAN=True | Token merging=0.6 | nsfw_filter=True | Seed List=8852281397 - 00:01:05 (65 seconds) - LoRA model=SvenN/sdxl-emoji