DALL·E 3
image
paid
OpenAI DALL·E 3 (2023) is a text-to-image model that creates high-quality images from natural language prompts. It was d...
Version: 3.0
Released: 2y 13d ago on 10/19/2023
Architecture
- parameters: Proprietary (undisclosed)
- context_length: Supports long detailed prompts
- training_data: Large curated dataset of images with captions
- inference: Transformer with diffusion and captioning enhancements
Capabilities
- Generates high-quality images from detailed text prompts
- Understands complex instructions (e.g. text in images, specific styles)
- Produces coherent visuals with fine details (faces, hands, text)
Benchmarks
- FID: State-of-the-art image quality (subjective comparison; no official values)
Safety
- Multi-tiered filtering for harmful or explicit
- will not reproduce art in the style of living artists or identifiable public figures
Deployment
- regions: global
- hosting: OpenAI servers
- integrations: available via ChatGPT (Plus/Premium), OpenAI API, Microsoft Bing Image Creator, Microsoft Designer tools
Tags
image-generationtext-to-imagegenerative AIcreative