Extending the love of tails into the digital realm.
1
Multi-Modal Virtual Character Synthesis: Robustness Testing
An intended test on the robustness of photo analysis by inputting low-quality, ambiguous images. The model successfully generated consistent, high-quality Pixar-style virtual character scenes from a blurry input image. It was then asked to create an profile picture for the same character. Model demonstrated its ability to handle challenging data and produce coherent outputs.
2
Multiple Virtual Characters Generation from a Single Source Image
Expands on previous work by testing AI's ability to generate multiple distinct virtual characters from a single source image containing multiple animals. The AI successfully identified and rendered separate characters with individual expressions and poses, showcasing its advanced scene understanding and character creation capabilities.
3
Group Virtual Character Generation from a Single Source Image
Much impressed by image to text capability for its abundant description of the source image. The attributes were then used to generate a poster picture of the animals it identified.
4
Compositional Virtual Character Synthesis from Individual Analyses
By inputting individual animal photographs, the model was asked to synthetically create an image with each animal as a virutal character in the style of Pixar-Disney. The model took into the consideration of physical attributes and arranged the animals proportionally.
5
Diverse Animal Avatar Generation
A quick run to generate profile picture for different species in different styles. Model once again impressed by high-quality high-res output regradless of low quality image.
6
Expression Generation for Virtual Characters
It begins by generating a primary character from an input image. A subsequent process of creating three additional expressions for the same virtual characters was requested.