Dezgo now supports Flux LoRAs created with a wider range of training software.
This significantly reduces Flux LoRA loading issues reported by our users and increases the variety of LoRAs that can be used with Dezgo.

Generate an image from a text description, while matching the structure of a given image

powered by Stable Diffusion / ControlNet AI (CreativeML Open RAIL-M)

Prompt

Describe how the final image should look like

Control model

What type of structure should be extracted from the control image?

Canny Edges

Init image

Upload the image used to guide the generation

Max: 512x512px (auto-resized)

Drag and drop files here or click

Model

The AI used to generate the image.

RealDream 12 (realistic)

Available in Power Mode

Count

Number of images to generate

Resolution

Choose image resolution

Portrait

Square

Landscape

Width

Height

More options

LoRA

Choose extension models

Control scale

How much should the control image influence the result? 0% = no influence, 100% = full influence.

100%

Control preprocess

Should the control image be preprocessed? If disabled, the control image will be used as-is and is assumed to already be in the correct format.

Yes

Negative prompt

Describe what you DON'T want in the generated image

Guidance

Adjusts how much the AI tries to fit the prompt (higher = stricter, lower = more freedom). The sweet spot is between 6-10, extreme values may produce more artifacts.

7.0

Available in Power Mode

Steps

Number of sampling steps. More steps = more details but also longer computation time.

Sampler

Defines the sampling method used to generate the image.

DPM++ 2M Karras

Seed

Unique image seed number. If not provided, the image will be random.

Available in Power Mode

Upscale

Upscale the image by this factor using the Real-ESRGAN model. Currently only a factor of 2 is supported.

Format

The encoding format of the generated image.

PNG