SD3.5 Medium + Flow-GRPO

Our model is trained separately for different tasks, so it’s best to use the corresponding prompt format for each task.

User Guide:

  1. Select a LoRA model (choose “None” to use the base model)
  2. Click “Sample Prompt” to randomly select from ~1000 task-specific prompts, or write your own
  3. Click “Run” to generate images (a 2×2 grid of 4 images will be produced)

Note:

  • For the Text Rendering task, please enclose the text to be displayed in double quotes ("), not single quotes (')
LoRA Model
0 2147483647
256 1024
256 1024
0 10
1 50