Figma: Building a multi-model image composition in Weave

Step-by-step

1. Open the left panel to access the library of model nodes
2. Drag and drop a prompt node onto the canvas and enter your text
3. Connect the prompt node to an image generation model node like Imagen 4 or Ideogram v3
4. Click Run Model to generate the image
5. Drag and drop a Compositor node onto the canvas
6. Connect multiple image nodes to the Compositor node as separate layers
7. Click Edit on the Compositor node to manually move and position layers for the final composition

Options

• Choose from various video models like Seedance 2.0, Sora 2, or Higgsfield
• Connect generated results to modification nodes like crop, levels, or color fine-tuning
• Add additional layers to the compositor for more complex scenes

Watch out for

• Every individual action or modification must be its own separate node
• Safari may have loading issues; use Google Chrome for better performance

Tips

• Check the credit cost of models before running; some like Seedance 2.0 are significantly more expensive than others like Imagen 4
• Read the documentation or test models individually to understand their specific behaviors
• Update the text in an existing prompt node and re-run it to iterate on an image without creating new nodes
• Use the Compositor node to manually 'tell' the model how to layer subjects over backgrounds

Highlights

“the hardest part is to know all of these models and to understand how they work”