Today's curated cutting-edge AI news, enjoy reading 👇
💻 NVIDIA launches Edify 3D model, capable of generating high-quality 3D scenes directly from text or images, bringing major breakthroughs to gaming, film, and other industries.
🤖 SAMURAI develops a high-precision video object tracking model based on SAM 2 that requires no training and maintains accuracy even in occlusion and fast-moving scenes.
🎬 A free AI tool ai-video-composer supports multimedia input and automatically generates videos using prompts.
Cutting-edge Technology
1. NVIDIA releases Edify 3D model for direct 3D scene generation.
Quickly generate scalable high-quality 3D assets with detailed UV mapping, 4K textures, and PBR materials through text prompts or reference images.
Details: https://research.nvidia.com/labs/dir/edify-3d/
This is undoubtedly great news for industries like game design, virtual and augmented reality, and film production.
2. A video object segmentation model based on enhanced SAM 2: SAMURAI.
Achieves high-precision video object tracking without training or fine-tuning, including fast movement and complex scenes, even with occlusion.
Details: https://yangchris11.github.io/samurai/
GitHub: https://github.com/yangchris11/samurai
From the demonstration video, it's indeed impressive. The code and model are now open-source, worth trying if needed.
AI Tools
1. Sharing a free AI automatic video editing tool: ai-video-composer.
Supports uploading multiple videos, images, or audio files at once, controlling editing through prompts to complete a full video.
Try it online: https://huggingface.co/spaces/huggingface-projects/ai-video-composer
However, the current implementation is a bit rough with suboptimal user experience. Interested users can check it out.