Quick Answer
Pick a target model (Midjourney, Stable Diffusion, DALL·E, or Flux), describe the subject, and click chips for style, lighting, camera, mood, and palette. The finished prompt assembles live with correct formatting for the target — Midjourney gets --ar/--chaos/--stylize/--seed flags and --no exclusions, Stable Diffusion gets a proper negative prompt block and weighted terms, DALL·E and Flux get natural-language phrasing. Copy it in one click. The builder is fully free with no API key; an optional AI enhancement adds visual detail if you add a key.
How It Works
Pick the target model — Midjourney, Stable Diffusion, DALL·E, or Flux — the output formatting follows automatically
Describe the subject, then click chips for style, lighting, camera/composition, mood, and palette
Add parameters (Midjourney --ar, --chaos, --stylize, --seed), weighted terms, and a negative prompt as needed
Copy the finished prompt; optionally Enhance with AI for added visual detail (3 free runs a day)
Key Facts
- Four target models with correct formatting each: Midjourney parameters, Stable Diffusion negative prompts, natural language for DALL·E and Flux
- 60+ curated chips across style, lighting, camera/composition, mood, and palette
- Midjourney --ar, --chaos, --stylize, --weird, --seed, and --niji inputs with validation-friendly fields
- Weighted terms (term::1.5) for Midjourney and Stable Diffusion
- Negative prompts / --no exclusions per target model
- No API key needed — the builder is deterministic; AI enhancement is optional
Frequently Asked Questions
Do I need an API key?
No. Building and copying prompts is fully free and runs entirely in your browser. The optional "Enhance with AI" button uses a model call — 3 free a day on this site’s key, or your own key.
What are weighted terms?
Midjourney multi-prompts (lighthouse::1.5) and Stable Diffusion emphasis shift attention toward or away from parts of the prompt. The weights field accepts comma-separated weighted terms and appends them correctly for the target.
Why does DALL·E get sentences instead of tags?
DALL·E 3 and Flux are tuned for natural-language descriptions; tag soup works against them. The builder rephrases the same selections as flowing sentences for those targets and keeps tag/parameter formatting for Midjourney and Stable Diffusion.