Available now

Creative Upscaler: creative upscaling

Creative Upscaler enlarges an image up to five times and redraws detail as it goes, guided by an optional prompt.

Model ID

Input

Fields marked * are required. Review the price before generating.

The image to upscale.

Uploading a file needs an account. Create account

JPEG, PNG, WebP · up to 25 MB

Be as descriptive as possible for best results. If no prompt is provide BLIP2 will be used to generate a prompt.

Example prompt. Edit it or write your own.

Additional settings

The type of model to use for the upscaling. Default is SD_1_5

The higher the scale, the bigger the output image will be.

How much the output can deviate from the original

How much detail to add

How much to preserve the shape of the original image

The suffix to add to the prompt. This is useful to add a common ending to all prompts such as 'high quality' etc or embedding tokens.

The same seed and the same prompt given to the same version of Stable Diffusion will output the same image every time.

The more steps the better the image will be but it will also take longer to generate.

If set to true, the image will not be processed by the CCSR model before being processed by the creativity model.

Allow for large uploads that could take a very long time.

The URL to the base model to use for the upscaling

The URL to the additional LORA model to use for the upscaling. Default is None

The scale of the additional LORA model to use for the upscaling. Default is 1.0

The URL to the additional embeddings to use for the upscaling. Default is None

Price$0.60 / run

You sign up first. Your settings are kept.

About this model

Overview

Upload an image, set the scale and decide how far the result may drift from the original. Creativity sets how much the output can change, detail sets how much new detail is added, and shape preservation holds the original outlines in place.

It runs on a Stable Diffusion 1.5 or SDXL base. Without a prompt, a caption of your image is written for you and used instead.

What you can build

  • Large, detailed versions of small generated images
  • Textures and backgrounds with added fine detail
  • Illustrations enlarged with a sharper finish
  • Soft photos given new surface detail

Prompting tips

  • Keep creativity low when faces, text or products must stay recognizable.
  • Describe the materials you want in the prompt, such as weathered wood or wet stone, to steer the new detail.
  • Raise shape preservation if objects start to change form at higher creativity.

What to check

Inspect faces, lettering and repeated textures for invented details.

Inputs

  • Model Type (optional): The type of model to use for the upscaling. Default is SD_1_5. Options: Sd 1 5, Sdxl.
  • Image (required): The image to upscale.
  • Prompt (optional): The prompt to use for generating the image. Be as descriptive as possible for best results. If no prompt is provide BLIP2 will be used to generate a prompt.
  • Scale (optional): The scale of the output image. The higher the scale, the bigger the output image will be. From 1 to 5.
  • Creativity (optional): How much the output can deviate from the original. From 0 to 1.
  • Detail (optional): How much detail to add. From 0 to 5.
  • Shape Preservation (optional): How much to preserve the shape of the original image. From 0 to 3.
  • Prompt Suffix (optional): The suffix to add to the prompt. This is useful to add a common ending to all prompts such as 'high quality' etc or embedding tokens.
  • Negative Prompt (optional): The negative prompt to use.Use it to address details that you don't want in the image. This could be colors, objects, scenery and even the small details (e.g. moustache, blurry, low resolution).
  • Seed (optional): The same seed and the same prompt given to the same version of Stable Diffusion will output the same image every time.
  • Guidance scale (CFG) (optional): The CFG (Classifier Free Guidance) scale is a measure of how close you want the model to stick to your prompt when looking for a related image to show you. From 0 to 16.
  • Number of Inference Steps (optional): The number of inference steps to use for generating the image. The more steps the better the image will be but it will also take longer to generate. From 1 to 200.
  • Skip Ccsr (optional): If set to true, the image will not be processed by the CCSR model before being processed by the creativity model.
  • Override Size Limits (optional): Allow for large uploads that could take a very long time.
  • Base Model URL (optional): The URL to the base model to use for the upscaling.
  • Additional Lora URL (optional): The URL to the additional LORA model to use for the upscaling. Default is None.
  • Additional Lora Scale (optional): The scale of the additional LORA model to use for the upscaling. Default is 1.0.
  • Additional Embedding URL (optional): The URL to the additional embeddings to use for the upscaling. Default is None.

Output and limits

  • Small or dense text in the image can come out misspelled.
  • Hands, crowds and complex poses can show artifacts; generate a few variations and pick the best.
  • Some prompts or outputs can be refused by safety checks.

Price guide

  • up to: $0.60 / run