Lumina Image 2.0 Native Support in ComfyUI

ComfyUI now enables native support for Lumina Image 2.0 models!

Lumina-Image-2.0 is a 2 billion parameter flow-based diffusion transformer capable of generating images from text descriptions. It offers capabilities in photorealistic rendering, text generation, and diverse artistic styles.

Try the Lumina2 Native Workflow

1. Update ComfyUI or download the latest ComfyUI Desktop

2. The easiest way is to use the all-in-one checkpoint we provide:

   1. Download lumina_2.safetensors and put it in your ComfyUI/models/checkpoints directory.

3. Try dragging in the example workflow!

Here are some examples from the basic workflow:

Some Notes on Running the Model

1. Sampling: The official way to sample this model is: shift 6 with 36 steps. Sampling it with lower steps also works, but you might have to lower the shift value as well to reduce the artifacts. For example, 20 steps with shift 3 helps avoid artifacts according to our tests.

2. Prompt: In the examples workflow, the "You are an assistant... <Prompt Start> " text before the actual prompt is the one used in the official example. The model still works if you remove it.

Enjoy creation!