Figma: Swapping objects in a pre-built AI workflow

Step-by-step

1. Replace the existing object description in the prompt node with 'an ultra realistic image of a banana'
2. Click Run Model to generate the initial 2D image
3. Refine the prompt with positional details, such as 'laying down', to improve the composition
4. Click Run Model on the 3D object node to convert the 2D image into a 3D asset
5. Run the image through the description node to allow the AI to analyze the camera angle
6. Execute the final image model node to generate the composite result

Options

• Run the model multiple times to iterate on the generated image results
• Tweak the original prompt with more context to fix placement issues (e.g., adding a table surface)

Watch out for

• Conflicting keywords in the prompt, such as 'floating', can confuse the AI and prevent it from placing objects correctly on surfaces

Tips

• Use specific positional descriptors like 'laying down' rather than just the object name for better visual results
• Leverage existing node connections to quickly swap subjects while maintaining the same camera angle and placement

Highlights

“since we already did the connections and everything, we can just click on run model”