Sources
Loading...
Additional media
Loading...
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
Stability AI has unveiled an early preview of Stable Diffusion 3, which outperforms other AI models in typography and prompt adherence. The model architecture includes the Diffusion Transformer and bidirectional flow of information between image and text tokens. The research paper highlights the model's ability to improve performance for multi-subject prompts, image quality, and spelling abilities.
In case you missed it, diffusion models can spell now! We released the Stable Diffusion 3 research paper yesterday, outlining the underlying technology powering the model. Read the paper here: https://t.co/kQtfJKPcSw https://t.co/zkPMYs7g8X
The Stable Diffusion 3 research paper is here. Source in this 🧵 https://t.co/iDXxhH8roH
The new @StabilityAI Stable Diffusion images are amazing! I decided to try comparing some images from the paper to what I get from Gemini and ChatGPT 4. Here's SD3, Gemini, and GPT4 for an example (in that order). See 🧵 below for more https://t.co/44YR3dSn13