Generate image prompts that actually work
Pick reference images from the library, choose exactly what to borrow from each — the lighting, the framing, the finish — describe your idea, and get a brand-new, original prompt built from what already delivers.
How it works
Pick reference images
Search the library by keyword or by meaning (semantic) and select up to 5 image prompts that already work. They shape your result — their wording is never copied into it.
Choose what to borrow
References marked “DNA” have been analysed into structured detail. Open one to see exactly what was found — lighting, composition, palette, subject — and tick which parts you want. Skip this and you borrow everything.
Describe your idea
Write your intent in plain language, e.g. “an ad for my chocolate brand with a model”. Optionally set a target model, tone, or constraints.
Generate, refine, publish
Each reference’s DNA is matched field by field against your idea to build one original prompt — and it tells you which reference supplied which part. Nudge it with a one-line instruction, then copy or publish.
What makes it different
Grounded in proven DNA
Instead of writing from scratch, the generator learns the structure of prompts that already produce great results and reuses their techniques — so output is defensibly grounded, not a generic wrapper.
Keyword & semantic search
Find references literally by title/tag, or flip on semantic search to match by meaning — great for fuzzy ideas where you don’t know the exact words.
Borrow only what you want
A reference’s lighting and framing are transferable craft; its palette is usually someone else’s brand. Open any analysed reference and pick which parts to take — keep the studio lighting, leave the colours behind.
Refine in one line
Not quite right? Keep the same result and nudge it — “more minimalist”, “add negatives”, “shorter” — without starting over or re-picking references.
See where every part came from
Results carry field-level attribution — “lighting ← ref 2”, “palette ← your intent”. Each one is checked against what that reference actually contains, so it’s a record of what happened rather than the model’s account of it.
Publish in a click
Love what you made? Send it straight to the share flow, pre-filled with a title and tags, and add it to your PromptsEdge profile.
Frequently asked
What is the AI Image Prompt Generator?
It writes a brand-new, original image prompt for your idea by learning from prompts that already work well in the PromptsEdge library — rather than generating generic text from scratch.
Does it copy existing prompts?
No. It reads two things from your references: their reusable structure, and an analysed breakdown of what they depict. It then writes fresh wording for your own subject. A reference’s brand names and specific subject matter are deliberately excluded so they can’t end up in your prompt.
What is prompt “DNA”?
Every image prompt in the library can be analysed into structured detail across five schemas — who appears, what appears, what is happening, how it looks, and how it was captured. That breakdown is the DNA. It’s what lets you borrow a reference’s lighting without also inheriting its colours.
Why do only some references show a DNA badge?
Analysis runs per prompt and is still being rolled out across the library. References without it still contribute their structure, just not the detailed breakdown. Claiming a locked prompt also generates its DNA if it doesn’t have any yet.
Can I generate video or text prompts?
Not right now — the generator is focused on image prompts. Video support is built and will be enabled once enough of the video library has been analysed.
How much does it cost?
Generating a prompt costs a small number of credits, and refining costs less. You can see your balance and the exact cost right in the tool.
Ready to generate your first prompt?
Pick your references, choose what to borrow, and get a polished, ready-to-use image prompt in seconds.
Open Prompt Generator