AI Image Prompt Generator

Good image prompts are not magic words — they are a structured list of descriptive terms assembled in the order each model weighs most. This generator assembles your selections deterministically (the same inputs always produce the same string) and then formats them for the specific model you pick. No image is generated and no data leaves your browser.

Term ordering (all models). The subject is placed first because all three official guides treat the opening of a prompt as the most weighted position. Remaining terms follow a fixed sequence:

Subject → Medium → Style → Art movement → Lighting → Composition → Mood / palette → Quality → Render engine

Midjourney. Descriptive terms are comma-separated, then parameters are appended last and space-separated: --ar W:H for aspect ratio, --stylize (clamped 0–1000), --chaos (clamped 0–100, omitted when zero), an optional --seed, the version flag --v 7, and exclusions as --no term1, term2. Every numeric value is clamped to the documented range before it is written, and the output is independently re-parsed to confirm each parameter sits inside its valid bounds.

Stable Diffusion (SDXL). Terms are comma-separated with no -- parameters. If you emphasise the subject, it is wrapped as (subject:1.3)using SDXL's documented weighting syntax (valid 0.1–2.0). The aspect ratio is reported as the nearest documented SDXL pixel dimension (for example 16:9 → 1344×768), and the negative prompt is emitted in its own block because SDXL takes it in a separate field.

DALL·E 3.OpenAI's guide states DALL·E 3 expects fluent natural language and ignores --flags and weight syntax, so the tool joins your selections into a sentence, expresses exclusions in prose (“without …”), and maps your ratio to one of the three supported sizes (1024×1024, 1792×1024, 1024×1792). Options that a model cannot use are hidden when that model is selected.

• Midjourney — Parameter List (--ar, --stylize, --chaos, --no, --seed, --v)
• Stability AI — Stable Diffusion / SDXL API reference (weighting, negative_prompt, dimensions)
• OpenAI — DALL·E 3 image generation guide (natural-language prompts, supported sizes)

The syntax rules, parameter ranges, and supported sizes on this page were last cross-checked against the official documentation above on 2026-06-08. They are reviewed whenever a model ships a new default version or changes its parameter set.