DALL-E
A 12 billion parameter GPT-3-VQ-VAE-2 combo for conditional image generation from natural text prompts. Breakthrough work, yields starkly accurate and creative outputs far beyond existing SOTA.
A 12 billion parameter GPT-3-VQ-VAE-2 combo for conditional image generation from natural text prompts. Breakthrough work, yields starkly accurate and creative outputs far beyond existing SOTA.