The Scene Relighting workflow modifies the lighting conditions of an existing scene to transition from one time of day to another.
This enables game developers and VFX teams to quickly visualize different lighting scenarios without re-rendering assets, saving time in production.
How To Use
- Upload an Image – Provide an initial image of the scene you want to relight.
- Describe the Scene – Enter a brief description of the scene, including key subjects, style, and textures.
- Select Lighting Conditions – Choose from a list of lighting options such as golden hour, fog light, bioluminescent glow, or volumetric lighting.
- Adjust Prompt Settings – Modify or refine the prompts to get the desired outcome.
- Choose a Model – Select the AI model (such as JuggernautXv10) to be used for processing.
- Generate and Review – The workflow applies edge detection, depth estimation, and relighting transformations to your scene.
- Save or Refine – Once satisfied, save your image to a board for further iterations or final use.
How It Works
1. Image Upload & Scene Description: The workflow begins by taking an input image that the user uploads. The user also provides a textual description of the scene, specifying elements like environment, subjects, and textures.
2. Lighting Selection & Prompt Processing: The user selects or enters lighting conditions to be applied to the scene. The workflow then processes these inputs, merging scene details with lighting descriptors to generate a complete relighting prompt.
3. Edge Detection & Depth Estimation: HED Edge Detection identifies structural lines in the image. Depth Estimation extracts depth data, allowing the workflow to apply lighting changes realistically based on depth and object positioning.
4. ControlNet Processing: The ControlNet module uses the extracted edge and depth data to guide the AI model in modifying the image while maintaining the scene’s structure.
5. Model Processing & Image Generation: The selected AI model (e.g., JuggernautXv10) processes the image using the refined prompt and ControlNet guidance. The denoising and rendering steps ensure smooth transitions between the original and newly applied lighting conditions.
6. Final Output & Save to Board: The relit image is generated and displayed on the canvas. The user can further refine it or save it to a board for future reference.
FAQs
.png)
