🖼️Image Block
Image blocks generate an image based on a text prompt or another image (via advanced controls) or both.
Title
Click the pencil icon next to the block’s title to edit it.
Tip: This is also the block’s variable name, which can be used to dynamically populate data in other blocks.
Prompt
Write a prompt that guides the AI to generate your desired image.
Use variables within your prompt to dynamically add data from other blocks.
When crafting a prompt for an image generator, it can be helpful to imagine that you’re talking to an alien who is visiting Earth for the first time. Break down the prompt into simple step-by-step instructions and be as descriptive as possible. Your new alien friend has never witnessed visual aspects of this planet before and needs a lot of context in order to draw an accurate picture of your prompt.
Variables
Variables can be used to dynamically populate data supplied by other blocks. Click on a variable name from the list to add it to anywhere in your prompt. Variable names match a block’s title.
Image Generation Model
Select the image generation model used to create your image. Currently supported models, as of 5/31/24, are:
Stable Diffusion XL (Stability AI)
DALL-E 3 (OpenAI) [Currently disabled]
Note: DALL-E also will not generate an image of a celebrity, per its usage policies.
Flux Pro (Black Forest Labs)
Flux Dev (Black Forest Labs)
Flux Schnell (Black Forest Labs)
Flux Pro v1.1 (Black Forest Labs)
Flux Pro v1.1 Ultra (Black Forest Labs)
Controllism
Note: Not really an image model in itself. It uses Stable Diffusion v1.5 with ControlNet QR.
Size
Select the desired aspect ratio of your image output.
Square (1:1)
Cinema (7:4)
Landscape (19:13)
Computer (9:7)
Portrait (7:9)
Tablet (13:19)
Advanced Image Block Controls
Negative prompts
Enter prompts that you do not want to be part of the image here. For example, add “hat” here if you want to exclude all hats from your image.
This setting is available for all supported image generation models except DALL-E.
Steps
The number of steps taken by the AI in refining its parameters.
Tip: Reduce the number of steps to speed up processing time. More steps will create a more refined output, but results in longer processing times.
This setting is available for all supported image generation models except DALL-E.
Prompt power
Prompt power refers to the influence that a prompt has over the generated output. The higher the prompt power, the more strictly the AI model will adhere to your prompt.
This setting is available for all supported image generation models except DALL-E.
Sampler
Currently supported samplers are:
pndm
unipc
k_euler_a
k_dpm_2_a
dpmpp
ddim
lcm
This setting is available for all supported image generation models except DALL-E.
Seed
Randomize Seed
The seed refers to the initial randomization or starting point for generating an image. Randomizing the seed will result in more diverse outcomes.
This setting is available for all supported image generation models except DALL-E.
Enable ControlNet
Enabling ControlNet allows you to specify an existing image as a reference for your generated image. ControlNet is a powerful and customizable tool that will stick to your prompt more closely than img2img. Paste the URL of the existing image that you want to use as the reference image. You can tweak the conditioning, canny upper/lower, and guidance start/end using the sliders.
Enable Input Image (Img2img)
Enabling img2img allows you to specify an existing image as a reference for your generated image. Img2img adheres to a prompt less closely than ControlNet. The resulting image may vary in style, color, and other aspects from the original. Paste the URL of the existing image that you want to use as the reference image. You can tweak the strength using the slider.
Restore faces
Enhances facial features in images. This setting is available for all supported image generation models except DALL-E.
This setting is only available for Stable Diffusion and Controllism.
Preview
Shows you a preview image. (Pretty self-explanatory!)
Last updated