Flux.2 Workflow Guide: Tell the differences of normal image generation and flux image generation

FLux.2 workflow is another useful and good-quility image generation workflow. It use flux model which is created by a company called blackforest lab. FLux workflow have a bit of differences with normal image generation workflow including new nodes and features. Here are the differences:

1. load text to image(flux.2 kelin 4b) workflow

Enter the template folder and search: Flux.2 Klein 4B: Text to Image.(the picture is hedgehog).

2. new nodes and its usage

  1. The main difference between text-image workflow of z-image-turbo and flux workflow is that ksampler is devided into several parts.
  • Except Flux2schedular, other nodes are the same as normal image generation workflow. If you forget it, review it at the end of article
  • The function of Flux2schedular acts as the Temporal Controller for the Flux model's denoising process. It translates the number of steps into a specific sequence of Timesteps . Just like thick morning fog get to lift gradually. thick fog is pure noise. after wind(algorism) and time(step) influence, fog will be lifted.
  1. empty flux2 latent node
  • Only used for flux.2 model, because of different channels and structures inside the node.
  1. Width and Height node
  • It converts width and height value to int datatype which is suitable for empty flux2 latent image node.

Some notes:

Flux model is developed by blackforest lab.It is polular in comfyui community.

  • here is some links to blackforest lab's website:
  1. Official website
  2. blackforest lab huggingface page
  3. blackforest lab github