Image Generation
AI systems that create new images from text descriptions, sketches, or other inputs, using diffusion models, GANs, or autoregressive approaches.
Technologies
Diffusion models: DALL-E 3, Stable Diffusion, Midjourney — iteratively denoise random noise into coherent images. GANs: StyleGAN for photorealistic faces. Autoregressive: Generating images token by token. Flow-based: Flux models.
Control Methods
Text prompts (natural language descriptions). ControlNet (guided by edges, poses, depth maps). Inpainting (modifying parts of existing images). Image-to-image (transforming style or content). Upscaling and enhancement.
Applications and Concerns
Design and creative workflows. Marketing and advertising content. Game and film concept art. Architectural visualization. Concerns include copyright, deepfakes, representation biases, and impact on creative professionals.