Overview
Describe
Deconstruct AI outputs to reveal the prompt, parameters, and settings that produced them
describereverse-engineertransparency
Docs
Describe deconstructs AI-generated outputs to reveal the components that produced them — the prompt (actual or inferred), parameters, model version, seed, and other settings. This reverse-engineering capability serves transparency and reproducibility: users can understand why an output looks the way it does, replicate results, or use the extracted prompt as a starting point for new generations.
Variants
- Panel view — A detailed card showing the output alongside extracted metadata, inferred prompts, and action buttons for reuse.
- Popover view — A compact overlay triggered by clicking or right-clicking an output, showing key details without navigating away.
- Inline view — Metadata displayed directly beneath the output in a minimal format, suitable for quick reference.
Use Cases
- Reverse-engineering AI image prompts to reproduce or iterate on results
- Understanding which parameters influenced a particular generation
- Sharing reproducible generation settings with team members
- Auditing AI outputs for compliance or quality assurance
- Learning effective prompt patterns by analyzing successful outputs
Best Practices
- Prefer exact extraction over inference when the original prompt is available.
- Provide a compact default view with expandable details for power users.
- Make results actionable — allow one-click reuse of extracted prompts.
- Clearly distinguish between exact prompts and inferred/reconstructed ones.
- Include model version and seed information for full reproducibility.