ComfyUI/docs/FluxProUltraImageNode/en.md

The Flux 1.1 [pro] Ultra Image node allows you to generate ultra-high-resolution images through text prompts, directly connecting to Black Forest Labs' latest image generation API.

This node supports two main usage modes:

1. Text-to-Image: Generate high-quality images from text prompts (when no image input is used)
2. Image-to-Image: Combine existing images with prompts to create new images that blend features from both (Remix mode)

This node supports Ultra mode through API calls, capable of generating images at 4 times the resolution of standard Flux 1.1 [pro] (up to 4MP), without sacrificing prompt adherence, and maintaining super-fast generation times of just 10 seconds. Compared to other high-resolution models, it's more than 2.5 times faster.

Parameter Description

Basic Parameters

Parameter	Type	Default	Description
prompt	String	""	Text description for generating the image
prompt_upsampling	Boolean	False	Whether to use prompt upsampling technique to enhance details. When enabled, automatically modifies prompts for more creative generation, but results become non-deterministic (same seed won't produce exactly the same result)
seed	Integer	0	Random seed value, controls generation randomness
aspect_ratio	String	"16:9"	Width-to-height ratio of the image, must be between 1:4 and 4:1
raw	Boolean	False	When set to True, generates less processed, more natural-looking images

Optional Parameters

Parameter	Type	Default	Description
image_prompt	Image	None	Optional input, used for Image-to-Image (Remix) mode
image_prompt_strength	Float	0.1	Active when image_prompt is input, adjusts the blend between prompt and image prompt. Higher values make output closer to input image, range is 0.0-1.0

Output

Output	Type	Description
IMAGE	Image	Generated high-resolution image result

How It Works

Flux 1.1 [pro] Ultra mode uses optimized deep learning architecture and efficient GPU acceleration technology to achieve high-resolution image generation without sacrificing speed. When a request is sent to the API, the system parses the prompt, applies appropriate parameters, then computes the image in parallel, finally generating and returning the high-resolution result.

Compared to regular models, Ultra mode particularly focuses on detail preservation and consistency at large scales, ensuring impressive quality even at 4MP high resolution.