Realtime AI video games, AI makes antibodies, video to 3D, colorize animations, new TTS

This week’s AI breakthroughs include tools like Calligrapher for precise text editing in images, X4D for creating 360-degree explorable scenes from images or videos, and Long Animation AI for automated, consistent coloring of animation sequences. Additionally, innovations such as Mirage enable real-time text-to-playable game generation, Chai 2 advances antibody design for drug discovery, and Q-Tai TTS offers high-quality open-source text-to-speech and voice cloning, highlighting AI’s growing influence across creative, gaming, and biomedical fields.

This week in AI has been packed with groundbreaking advancements across various domains. One standout tool is Calligrapher, an AI that edits text in images while perfectly preserving the original font and style, even allowing users to transfer styles from reference images or abstract visuals. This capability is highly versatile, supporting complex fonts and enabling precise micro-edits without altering the rest of the image. The tool is open source with an easy-to-use graphical interface, making it accessible for creative professionals and hobbyists alike.

Another impressive innovation is X4D by Pico and ByteDance, which transforms images or videos into fully explorable 360-degree scenes. Whether starting from a single image or a short video, X4D extrapolates unseen angles to create immersive environments. Although it requires substantial VRAM (48 GB minimum), the open-source nature suggests future optimizations for broader accessibility. This technology holds great promise for virtual reality, gaming, and immersive media experiences.

For animators, the Long Animation AI is a game-changer, automating the coloring of long animation sequences from just one reference frame. It maintains color consistency across frames and can even generate backgrounds based on textual prompts, significantly reducing manual labor in animation production. Compared to existing colorization methods, it delivers superior quality and sharpness, making it a valuable tool for studios and independent creators.

In the realm of gaming, Mirage offers a revolutionary approach by generating fully playable video games in real time based solely on text prompts. Demonstrations include GTA-style and racing games where players can control characters and vehicles interactively. While the graphics and responsiveness are not yet perfect, this technology hints at a future where game creation is instant and highly customizable, potentially transforming game development and user experiences.

Finally, significant strides have been made in AI-driven biomedical research and multimedia generation. Chai 2 designs antibodies from scratch with a success rate 100 times better than previous methods, accelerating drug discovery. Meanwhile, ByteDance’s Xverse excels at transferring reference images into new photos with high fidelity, and Depth Anything at Any Condition improves depth estimation in challenging environments. Additionally, the new Q-Tai TTS offers open-source, high-quality text-to-speech and voice cloning capabilities, rivaling commercial solutions like 11 Labs, though currently limited to English and French. These advances collectively showcase AI’s expanding impact across science, art, and entertainment.