AI makes 4D videos • AI separates video layers • New open-source AI video • SORA leak • Realistic 3D models • AI news this week
Welcome to the AI Search newsletter. Here are the top highlights in AI this week.
CAT4D can create 4D scenes from real or generated videos. It uses a multi-view video diffusion model to generate videos from different viewpoints, and then reconstructs the dynamic 3D scene as deforming 3D Gaussians. This allows for interactive 4D scenes with separate camera and time control. Read more
Do you prefer to watch instead of read? Check out this video covering all the highlights in AI this week:
AI can predict neuroscience study results better than human experts, a new study finds. The study showed that large language models, a type of AI, can analyze patterns in scientific literature and forecast outcomes of experiments with superhuman accuracy. This has the potential to accelerate research and improve decision-making in experiment design. Read more.
Amazon has developed a new AI model that can process images and videos, not just text. This AI, code-named Olympus, can understand scenes in images and videos and help customers search for specific scenes using simple text prompts. This new AI model will make Amazon less reliant on AI startup Anthropic. Read more.
uPix is an AI Selfie Generator that allows users to turn into anyone in just one click. Select from a vast array of templates, ranging from superheroes to business portraits, and even anime characters. Try it out today!
Generative Omnimatte is a new AI model that can separate videos into individual layers, allowing for precise editing and removal of objects and effects. This model decomposes videos into layers, enabling users to edit and remove specific objects, shadows, and reflections. Read more
SAMURAI is a new AI model that can track objects in videos with high accuracy, even in crowded and complex scenes. This model uses a combination of machine learning algorithms and motion-aware memory selection to predict object motion and refine mask selection, achieving robust tracking without the need for retraining or fine-tuning. Read more
Researchers have created a new AI system called Material Anything that generates realistic materials for 3D objects automatically. This system uses advanced techniques to adapt materials to different lighting scenarios without needing complicated setups. It combines a smart diffusion model with confidence masks to improve the quality of materials consistently. Read more
xAlerts is a powerful tool that helps you track the activity of your favorite accounts. Stay informed from the latest activities of investors, celebrities, influencers, athletes, and more. Try it for free!
Researchers have created LTX-Video, a new AI model that generates high-quality videos in real-time. This model can produce videos at 24 frames per second and is capable of transforming text or images into engaging video content. It's designed to work quickly and effectively across various video styles and scenarios. Read more
QwQ is an open-source AI model designed to enhance reasoning and problem-solving skills through deep questioning and reflection. It tackles complex tasks like math and coding by continuously analyzing its own thought processes, which helps it improve over time. It achieves similar performance compared with OpenAI’s o1 models. Read more
A group of artists leaked access to OpenAI's Sora video generator as a protest against being exploited for their work. The leak allowed the public to briefly use Sora, but it didn't reveal any details about how the model was trained or its underlying code. Read more
Ai2 has introduced OLMo 2, a new open language model that performs better than many existing models. This model is designed to understand and generate text more effectively by using advanced training techniques and large datasets. OLMo 2 is available for the community to use and improve upon, promoting open science in AI development. Read more