AI Briefing#diffusion-transformer
The latest AI and developer news about #diffusion-transformer, with the original source and a short summary.
The latest AI and developer news about #diffusion-transformer, with the original source and a short summary.
Meta has introduced Muse Realtime Avatar, a system that generates synchronized, expressive video from speech tokens with subsecond latency, outperforming Runway and HeyGen in user preference tests.
Odyssey has released Odyssey-3, a general-purpose physical intelligence model designed to control robots, autonomous vehicles, and games.
Wan-Animate-2 supports video-based character animation and text-based viewpoint control.
Researchers proposed the DSR technique to reduce outlier tokens in DiTs, improving image generation quality.
Proposes a unified diffusion transformer model that simultaneously optimizes a robot's morphology and control policy.
Apple has unveiled a new audio synthesis architecture that generates high-quality speech in real time while minimizing memory usage.
Nvidia has announced SANA-Video 2.0, which accelerates high-resolution long-form video generation through hybrid attention technology.
Microsoft has released Mage, a lightweight multimodal model family with 4B parameters.
LingBot World has been released, a model that solves the drift problem occurring during long-term generation through MoBA attention masks and distillation techniques.
Krea has released Krea 2, an open-weight, 12B-parameter image generation model optimized for creative exploration and control.
A framework has been released for unified evaluation of Diffusion Transformer model performance, spanning from ImageNet to T2I.
Qwen-RobotWorld is a next-generation Embodied World Model that uses natural language as an interface to jointly train across diverse robot embodiments and environments.
Bonsai Image 4B, an ultra-lightweight image generation model that reduces memory usage to around 1GB through 1-bit and Ternary quantization, has been released.
PrismML has unveiled Bonsai Image 4B, a 1-bit/ternary-based model that can run locally in the browser via WebGPU.
NVIDIA has unveiled SANA-WM, which generates 1-minute 720p video with 6-DoF camera trajectories.
NVIDIA has released a 2.6B open-source world model for 1-minute 720p video.
A summary of the performance and limitations of local music generation on Apple Silicon by running ACE-Step 1.5 with MLX.
Lemon Slice partnered with ElevenLabs to launch a photo-based interactive avatar platform.
ActionParty solves the 'action binding' problem, where video generation models fail to accurately assign different actions to multiple subjects.
It summarized the major technical turning points and open-source model achievements in AI image and video generation in 2024.
OpenAI researched technology to simulate the physical world through its video generation model Sora.