AI Image Prompt Architect & Generator

Status: Version 1.0

Last verified:

Description: Build engine-aware image prompts through four progressive-disclosure zones and generate an optional configured preview.

Instructions: Define the subject, then select styling, composition, and output options. Review the compiled prompt, add the selected provider key, and generate or copy.

Engine and output setup

AI engine

The formatted output and generation provider adapt to the selected engine.

Used only for this Gemini generation request. Optional when MyKMHub has a Gemini server key. It is not retained unless browser-session caching is enabled below.

Opt in to retain the masked provider keys across reloads in this tab. Closing the browser session discards them; shared devices are not recommended.

Aspect ratio
Preview output
Each preset combines provider quality and resolution.

Describe the central person, object, place, or idea.
Traits and apparel
Action or pose
Subject scale and quantity
Location or setting
Optional. Overrides the setting selection.
Time or era
Weather and atmosphere

Accessibility and privacy

What this version supports

  • Spectrum 2 controls provide visible labels, keyboard operation, focus treatment, and accessible states.
  • Four Spectrum 2 accordion zones use progressive disclosure to reduce cognitive load.
  • Option groups are capped and every control has a visible label and keyboard behavior.
  • Status messages announce copying, generation, manual override, and reset results without moving focus.
  • Engine output and the structured parameter breakdown remain visible for review.
  • Prompt content remains in the current browser session.
  • Nothing entered in the tool is persisted.
  • Each user-supplied provider API key is masked, isolated from keys entered for other engines, and sent only to the MyKMHub server for that provider request.
  • Provider keys are not retained by default. A user may explicitly cache them in sessionStorage so they survive reloads in the current browser session; closing that session discards them.
  • When generation is intentionally activated, the natural-language prompt is sent to the selected provider; selections and copied engine syntax are not stored.
Version notes

1.0

  • Promoted the accessible engine-aware prompt architect and generator to version 1.0.
  • Made Gemini the default and grouped provider, credential, quality, aspect-ratio, and resolution controls at the top.
  • Added opt-in, session-only caching for provider API keys.

0.3 engine-aware beta

  • Expanded the workflow into four progressive-disclosure zones.
  • Added OpenAI, Gemini, Midjourney, and Stable Diffusion/FLUX translation.
  • Added editable compiled output and a structured parameter breakdown.
  • Added provider-specific OpenAI, Gemini, and Black Forest Labs FLUX credentials and direct generation.
  • Kept Midjourney as a copy-ready prompt workflow without collecting a credential for an unsupported API.

0.2 MyKMHub modernization

  • Rebuilt the workflow with React and Spectrum 2 controls.
  • Kept the prompt workflow independent of any image-generation model.
  • Added a persistent compiled preview, common negative constraints, and iteration notes.

0.1 original prototype

  • Established a six-stage guided prompt-building workflow.
  • Used semantic HTML, Tailwind CSS, and vanilla JavaScript.