IP‐Adapter

Informations

• Module : IP-Adapter
• Function : Transform an input image, with a conditional IP-Adapter image, a prompt and a negative prompt using Stable Diffusion, IP-Adapter and ostris/ip-composition-adapter
• Input(s) : Input image, conditional IP-Adapter image, prompt, negative prompt
• Output(s) : Image(s)
• HF model page :
  • SG161222/Realistic_Vision_V3.0_VAE
  • RunDiffusion/Juggernaut-XL-Lightning
  • Fluently-XL-v3-Lightning
  • fluently/Fluently-XL-v4
  • recoilme/ColorfulXL-Lightning
  • playgroundai/playground-v2-512px-base
  • playgroundai/playground-v2-1024px-aesthetic
  • playgroundai/playground-v2.5-1024px-aesthetic
  • stabilityai/sdxl-turbo
  • SG161222/RealVisXL_V4.0_Lightning
  • cagliostrolab/animagine-xl-3.1
  • aipicasso/emi-2
  • dataautogpt3/OpenDalleV1.1
  • dataautogpt3/ProteusV0.4
  • dataautogpt3/ProteusV0.4-Lightning
  • digiplay/AbsoluteReality_v1.8.1
  • gsdf/Counterfeit-V2.5
  • stabilityai/stable-diffusion-xl-base-1.0
  • runwayml/stable-diffusion-v1-5
  • nitrosocke/Ghibli-Diffusion

Help

• Usage :
  • (optional) Modify the settings to use another model or generate several images in a single run
  • (optional) Select a LoRA model and set its weight
  • Select the IP-Adapter type to use : standard (image to image + prompt) or composition (import composition into the output image)
  • Upload or import an image as an Input image (if using standard IP-Adapter)
  • Upload an image as an IP-Adapter image
  • Set the balance between the input image and the prompts (Ip-Adapter image, prompts, negative prompt) by choosing a denoising strength value between 0.01 and 1 : 0.01 will mostly ignore the prompts, 1 will completely ignore the input image
  • Fill the prompt with what you want to see in your output image
  • Fill the negative prompt with what you DO NOT want to see in your output image
  • Click the Generate button
  • After generation, generated images are displayed in the gallery. Save them individually or create a downloadable zip of the whole gallery.
• LoRA models :
  • You could place huggingface.co or civitai.com Stable diffusion based safetensors LoRA models in the directory ./biniou/models/lora/SD or ./biniou/models/lora/SDXL (depending on the LoRA model type : SD 1.5 or SDXL). Restart Biniou to see them in the models list.

Examples :

• Input image :

• IP-Adapter image :

• Prompt : high quality

• Negative prompt : low quality, blurry

• Output(s) :