AI image prompt expansion solves a very specific pain point: you have a picture in your head, but the prompt you write is a single sentence. "Draw a ginger cat napping on a windowsill" — the subject is there, but style, lighting, and composition are missing, so the result is pure luck. The expansion prompt below fixes exactly that: give the AI your one-line idea, it asks for the missing key details first, then outputs a complete English prompt you can paste straight into image tools like Midjourney or Jimeng — plus swappable word groups for easy tweaking.
When expansion is worth using
- You can picture the scene but can't write the English detail words: lighting, lens, style, and composition terms are most people's weak spot, and expansion fills exactly that gap.
- You're generating a batch on one theme: expand once to lock the style words, then only swap the subject — the whole set stays consistent.
- You're making images for someone who doesn't speak English: the expanded result is a ready-made English prompt, just copy and paste.
- Two cases where it doesn't fit: you already write English prompts fluently (writing directly is faster); or the idea itself isn't formed yet — figure out "what to draw, what mood" first, then expand.
The prompt: copy this block
You are an AI image prompt specialist, skilled at expanding a one-line idea into a complete English image prompt.
Workflow:
1. Read my one-line idea first and identify the missing key details (subject details, style, lighting, composition, aspect ratio). Confirm with no more than 3 questions; do not output the expanded result before I reply.
2. After confirmation, output two parts: first, a complete English prompt covering subject, setting, style, lighting, and composition in order, ending with quality terms; second, "swappable word groups" organized by subject, style, and lighting, with 2 alternative English terms per group for easy swapping.
3. Output only the expanded result and the swappable word groups — no explanations, no translations.
Let's begin. My idea is: [one-line idea]Variables: only one to fill in
| Variable | What to fill in | Examples |
|---|---|---|
| [one-line idea] | The picture in your head — subject and mood in one sentence, the more specific the better | A ginger cat napping on a sunlit windowsill / A cyberpunk city nightscape |
This prompt deliberately keeps a single variable: the AI asks follow-up questions to fill in what's missing, instead of making you figure out everything at once — that's the difference between "expansion" and "writing the full prompt yourself."
A real expansion example: what input and output look like
Input: "A ginger cat napping on a sunlit windowsill"
AI follow-ups (example): "Photorealistic or illustration style? Landscape or portrait orientation?"
After confirmation:
English prompt: A ginger cat napping on a sunlit windowsill, warm afternoon light streaming through sheer curtains, cozy home atmosphere, photorealistic style, 85mm lens, shallow depth of field, ultra detailed, 8k --ar 16:9
Swappable word groups: subject: ginger cat / tabby cat; style: photorealistic / watercolor illustration; lighting: warm afternoon light / soft morning glow
Copy the whole English prompt and paste it into Midjourney's input box (or the prompt box of Jimeng / Tongyi Wanxiang) to generate. Want a tabby instead? Just swap ginger cat for tabby cat and keep the style and lighting untouched. If Midjourney blocks your prompt, see this troubleshooting guide on community guidelines and the --no parameter.
Tuning tips: getting results closer to what you want
- Set the aspect ratio before expanding: add "portrait orientation, good for a phone wallpaper" to your idea — much easier than cropping afterward.
- Lock the style when it drifts: add "keep the style fixed as Studio Ghibli animation, no photorealistic elements" to stop the AI improvising.
- Delete extras the expansion added: if it threw in curtains and dust on its own, reply "remove the curtains and dust, keep only the cat and the windowsill" and it will regenerate your version.
- The more specific your idea, the better the English output: "ginger cat" beats "cat", "3pm sunlight" beats "sunlight" — expansion supplies detail words, not a subject you haven't figured out.
- Save expansions you like: when batch-generating on one theme, keep the "swappable word groups" part — next time just swap the subject words and the whole set keeps its style automatically.
Q: I don't speak English at all. Can I still use the expanded English prompt directly?
A: Yes. The expanded result is made for copy-pasting — you don't need to understand every word, and the alternatives in "swappable word groups" are swapped in directly with no grammar needed. If you want to tweak something but worry about breaking it, tell the AI in your own language (e.g. "change the lighting to early morning") and let it edit the English part for you.
