Latent Mirroring is a technique applied during the image generation process in models like Stable Diffusion to introduce symmetry and balance into the generated images. By applying mirroring and flipping operations to the latent representations of images mid-generation, this method can produce compositions ranging from subtle balanced designs to perfect reflections.
Key Features:
- Symmetry Enhancement: Latent mirroring ensures that generated images exhibit symmetrical features, which can be particularly useful for creating balanced compositions or realistic reflections.
- Mid-Generation Application: The technique is applied to the latent images during the generation process, allowing for real-time adjustments and refinements.
- Versatility: Depending on the strength of the mirroring applied, the effect can range from subtle enhancements to pronounced symmetrical features.