How it works — AI Image Generation
Generate images from text descriptions using DALL-E 3 — directly inside any NuBench conversation. Type /imagine followed by your description. Budget, classification, and audit apply the same as every other request.
Type /imagine followed by any description in the chat input to generate a high-quality image with DALL-E 3. You can also use natural language phrases like "generate an image of…" or "draw me…".
Images are generated at up to 1792×1024 resolution. You can specify size (square, landscape, or portrait) and quality (standard or HD) from the chat settings or by including them in your request.
Generated images appear directly in the conversation thread — no separate tab, no copy-paste. A Download button is always available. Hover or tap the image for the full prompt.
Images are stored in your organization's private S3 bucket with server-side KMS encryption. Presigned URLs expire after 15 minutes. Re-opening the conversation generates a fresh URL automatically.
Image generation runs through the same 16-step pipeline as text requests — DLP, budget enforcement, classification, and audit logging all apply. Content policy rejections are caught and surfaced as clear error messages.
Every generated image is recorded in the audit log with the user identity, organization, prompt (or revised prompt), model, cost, and timestamp. GDPR erasure applies to generated image records in both Phase 1 and Phase 2.
Each image is priced at standard DALL-E 3 rates ($0.04 standard / $0.08 HD at 1024×1024, higher for wider aspect ratios) and attributed to the active cost code. Budget caps apply identically to image spend.
Using your own OpenAI BYO key? Image generation uses your BYO key automatically. No NuBench markup applies when a BYO key is active — you pay your OpenAI rate directly.
Generate custom diagrams, conceptual illustrations, or placeholder visuals for client decks — all from a text description, attributed to the correct client cost code.
Quickly sketch layout concepts or visual directions before committing to a design system. Generate variations by rephrasing the prompt.
Add custom figures or illustrations to reports and internal documents. Images are stored in your organization's secure storage and linked back to the conversation for context.
Visualize architectural concepts, process flows, or strategic options described in text. Useful for strategy presentations, board materials, and stakeholder briefings.