fermatresearch / high-resolution-controlnet-tile

UPDATE: new upscaling algorithm for a much improved image quality. Fermat.app open-source implementation of an efficient ControlNet 1.1 tile for high-quality upscales. Increase the creativity to encourage hallucination.

  • Public
  • 615.4K runs
  • L40S

Input

image
string
Shift + Return to add a new line

Prompt for the model

file

Control image for scribble controlnet

integer

Image resolution

Default: 2560

number
(minimum: 0, maximum: 1)

Conditioning scale for controlnet

Default: 0.85

number
(minimum: 0.1, maximum: 1)

Denoising strength. 1 means total destruction of the original image

Default: 0.35

number
(minimum: 0, maximum: 1)

HDR improvement over the original image

Default: 0

string

Choose a scheduler.

Default: "DDIM"

integer

Steps

Default: 8

number
(minimum: 0, maximum: 30)

Scale for classifier-free guidance, should be 0.

Default: 0

integer

Seed

string
Shift + Return to add a new line

Negative prompt

Default: "teeth, tooth, open mouth, longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, mutant"

boolean

In this mode, the ControlNet encoder will try best to recognize the content of the input image even if you remove all prompts.

Default: false

number
(minimum: -5, maximum: 10)

Strength of the image's sharpness. We don't recommend values above 2.

Default: 1.25

number
(minimum: -5, maximum: 3)

Strength of the image's details

Default: 1

string

Format of the output.

Default: "jpg"

Output

We were unable to load these images. Please make sure the URLs are valid.

{
  "input": "https://replicate.delivery/pbxt/LKnw8rSgafZf4IlAVyPhzpX1TpTVcyfRa1saoaoiSfUYZLiL/fermat_app_a_living_room_modern_and_minimalistic_39b5a58a-e05b-4281-ac24-e87435256333-1.webp",
  "outut": "https://replicate.delivery/pbxt/YFlNVWt9oEpTMVEKg69wXDRGtjHSHKV02ESZeChKj69ffiYmA/output.jpg"
}
Generated in

This output was created using a different version of the model, fermatresearch/high-resolution-controlnet-tile:13c637d3.

Run time and cost

This model costs approximately $0.045 to run on Replicate, or 22 runs per $1, but this varies depending on your inputs. It is also open source and you can run it on your own computer with Docker.

This model runs on Nvidia L40S GPU hardware. Predictions typically complete within 47 seconds. The predict time for this model varies significantly based on the inputs.

Readme

High quality upscale from Fermat.app. Increase the creativity to encourage hallucination.