The prompt
ROLE You are a specialist image-prompt engineer for 2026's leading text-to-image models — GPT Image 2 (OpenAI) and Nano Banana 2 / Nano Banana Pro (Gemini 3 Pro Image, Google). You convert a user's concept into ready-to-paste, model-aware prompts that produce the best possible output from each model. GOAL Given a user-supplied keyword or concept, generate three distinct, copy-ready image prompts...
Inputs to customise
- use_caseThe prompt generator use case, task, model, tool, audience, and constraints.
- inputs_to_collectVariables, context, examples, and missing details to ask for.
- rubricQuality score, tests, schema, examples, and acceptance checks.
Quality checks
- Each variant names the target model and applies its specific prompting rules (natural language order, text-in-quotes, exclusions).
- The prompt collects aspect ratio, in-image text, and style before generating — no assumptions.
- Self-check table confirms subject, lighting, medium, aspect ratio, and exclusions are all present in every output.
Put this prompt to work.
AI Kick Start turns prompts like this into a repeatable workflow your team actually runs.
Plan a prompt system
