Skip to main content

StabilityAI

Properties used to connect to Stability AI.

stabilityAI​

Service Types​

TextToImage​

  • Type: true | { StabilityAICommon, width?: number, height?: number }
  • Default: {engine_id: "stable-diffusion-v1-6", width: 512, height: 512}

Connect to Stability AI's text-to-image API.
StabilityAICommon properties can be used to set the engine Id and other image parameters.
width and height is used to set the image dimensions. They must be multiples of 64 and pass the following:
For 768 engines: 589,824 ≤ width * height ≤ 1,048,576 and for other engines: 262,144 ≤ width * height ≤ 1,048,576.

Example​

<deep-chat
  directConnection='{
    "stabilityAI": {
      "key": "placeholder key",
      "textToImage": {"engine_id": "stable-diffusion-v1-6", "height": 640, "samples": 1}
    }
  }'
></deep-chat>

ImageToImage​

  • Type: true | {
         StabilityAICommon,
         init_image_mode?: "image_strength" | "step_schedule_*",
         image_strength?: number,
         step_schedule_start?: number,
         step_schedule_end?: number
    }

  • Type: {
         engine_id: "stable-diffusion-v1-6",
         init_image_mode: "image_strength",
         image_strength: 0.35,
         step_schedule_start: 0.65,
         weight: 1
    }

Connect to Stability AI's image-to-image API.
StabilityAICommon properties can be used to set the engine Id and other image parameters.
init_image_mode denotes whether the image_strength or step_schedule properties control the influence of the uploaded image on the new image.
image_strength determines how much influence the uploaded image has on the diffusion process. A value close to 1 will yield an image very similar to the original, whilst a value closer to 0 will yield an image that is wildly different. (0 to 1)
step_schedule_start and step_schedule_end are used to skip a proportion of the start/end of the diffusion steps, allowing the uploaded image to influence the final generated image. Lower values will result in more influence from the original image, while higher values will result in more influence from the diffusion steps. (0 to 1)

Example​

<deep-chat
  directConnection='{
    "stabilityAI": {
      "key": "placeholder key",
      "imageToImage": {"engine_id": "stable-diffusion-v1-6", "init_image_mode": "image_strength", "samples": 1}
    }
  }'
></deep-chat>

ImageToImageMasking​

  • Type: true | {
         StabilityAICommon,
         mask_source?: "MASK_IMAGE_WHITE" | "MASK_IMAGE_BLACK" | "INIT_IMAGE_ALPHA"
    }

  • Default: {engine_id: "stable-diffusion-xl-1024-v1-0", mask_source: "MASK_IMAGE_WHITE", weight: 1}

Connect to Stability AI's image-to-image-masking API.
StabilityAICommon properties can be used to set the engine Id and other image parameters.
mask_source is used to define where the source of the mask is from. "MASK_IMAGE_WHITE" will use the white pixels of the mask image (second image) as the mask, where white pixels are completely replaced and black pixels are unchanged. "MASK_IMAGE_BLACK" will use the black pixels of the mask image (second image) as the mask, where black pixels are completely replaced and white pixels are unchanged. "INIT_IMAGE_ALPHA" will use the alpha channel of the uploaded image as the mask, where fully transparent pixels are completely replaced and fully opaque pixels are unchanged - in this instance the mask image does not need to be uploaded.

Example​

<deep-chat
  directConnection='{
    "stabilityAI": {
      "key": "placeholder key",
      "imageToImageMasking": {"mask_source": "MASK_IMAGE_WHITE", "samples": 1}
    }
  }'
></deep-chat>

ImageToImageUpscale​

  • Type: true | {engine_id?: string, width?: number, height?: number}
  • Default: {engine_id: "esrgan-v1-x2plus"}

Connect to Stability AI's image-to-image-upscale API.
engine_id is the engine that will be used to process the image.
width and height are used to define the desired with of the result image where only EITHER ONE of the two can be set. Minimum dimension number is 512.

Example​

<deep-chat
  directConnection='{
    "stabilityAI": {
      "key": "placeholder key",
      "imageToImageUpscale": {"width": 1000}
    }
  }'
></deep-chat>

Shared Types​

Types used in stabilityAI properties:

StabilityAICommon​

  • Type: {
         engine_id?: string,
         samples?: number,
         weight?: number,
         cfg_scale?: number,
         sampler?: string,
         seed?: number,
         steps?: number,
         style_preset?: string,
         clip_guidance_preset?: string
    }

  • Type: {
         samples: 1,
         cfg_scale: 7,
         seed: 0,
         steps: 50,
         clip_guidance_preset: "NONE"
    }

Object that is used to define the target engine and other image processing parameters.
engine_id is the identifier for the engine that will be used to process the images.
samples is the number of images that will be generated (1 to 10).
weight defines how specific to the prompt the generated image should be (0 to 1).
cfg_scale defines how strictly the diffusion process should adhere to the prompt (0 to 35).
sampler is the sampler that will be used for the diffusion process. If this value is not set - the most appropriate one is automatically selected.
seed is the number for the random noise (0 to 4294967295).
steps is the number of diffusion steps to run (10 to 150).
style_preset guides the image model towards a particular style.
clip_guidance_preset is the clip guidance preset.

Example​

<deep-chat
  directConnection='{
    "stabilityAI": {
      "key": "placeholder key",
      "textToImage": {
        "engine_id": "stable-diffusion-v1-6",
        "weight": 1,
        "style_preset": "fantasy-art",
        "samples": 2
  }}}'
></deep-chat>