Diffusion and Compositing
I also created estimated depth maps for compositing using MiDaS, a deep learning algorithm for producing depth map estimations from 2D images. Then used this Z-Depth for true camera depth of field effects in Nuke, some volume rays and painted problem areas, color graded and corrected for the final image.
Introducing these new AI models and algorithms to my workflow and achieving high-quality images with minimal retouching. Getting results fast!
I also created estimated depth maps for compositing using MiDaS, a deep learning algorithm for producing depth map estimations from 2D images. Then used this Z-Depth for true camera depth of field effects in Nuke, some volume rays and painted problem areas, color graded and corrected for the final image.
Introducing these new AI models and algorithms to my workflow and achieving high-quality images with minimal retouching. Getting results fast!
Last but not least, the Astro Nation logo is created with Stable Diffusion by inputting a random font from Google Fonts. The image was then upscaled using a model trained on typography and vector graphics, eliminating any bitmap pixellations. This is something I wish I had while I was working as an art director in an advertising agency.
