Diffusion

What it means

A diffusion model denoises step-by-step from noise to pixels, conditioned on your prompt and settings like aspect ratio, style, or reference image.

Why designers should care

Image UX needs progress for multi-second runs, seed/history for reproducibility, negative prompts or filters, and clear rights or safety messaging on outputs.

Example

A marketing tool generates ad variants from a brief; users see step progress, pick from four thumbnails, edit prompt, and regenerate one slot without rerunning the whole batch.

Common mistakes

• • Treating image generation like instant chat with no wait or cancel states.
• • No gallery/history when outputs are non-deterministic and users need to compare runs.
• • Hiding that diffusion can produce artifacts, wrong text, or off-brand visuals without review.

Related terms

• Model: A model is a packaged set of learned weights and settings that maps prompts plus context to generated text, classifications, or tool calls.
• Inference: Each user request triggers inference: tokens in, tokens out, optionally interleaved with tool calls and retrieval.
• Latency: Network, queue, retrieval, tool calls, and generation length all add wait time before users can read, edit, or act on output.
• Multimodal: A multimodal model accepts or produces multiple media types, not just strings of text, often in one conversation or agent run.
• Moderation: Automated classifiers and human reviewers flag content categories (hate, violence, sexual content, PII) and enforce actions (block, blur, warn, escalate).