Imagen 3
Google DeepMind · June 2024
● activeCloseddiffusionimage
Description
Google's highest quality text-to-image model. Produces photorealistic images with exceptional detail, richer lighting, and significantly fewer visual artifacts (distortions or errors) than its predecessor, especially for complex scenes with multiple objects.
Key Innovations
Diffusion
DiffusionGenerates outputs by gradually denoising random noise into coherent images/audio. The backbone of Stable Diffusion and DALL·E.
Text-to-Image
Text-to-ImageGenerating images from text descriptions — the technology behind DALL·E, Midjourney, and Stable Diffusion.
Related Research (1)
ImagenDiffusion
2022 · Google Brain
Demonstrated that large frozen text encoders (T5-XXL) with cascaded diffusion models produce photorealistic images, outperforming DALL·E 2.