L2Lesson 4Free

Describe an image so any image tool nails it

After this, you'll understand that Claude writes the prompt while a separate image tool draws the picture, and you'll be able to have Claude write a well-structured image prompt you paste into whatever image tool you use.

Before you start

Complete Build a moodboard brief first; with your look captured in words, this lesson turns that brief into precise image prompts your own image tool can follow.

The idea

Claude cannot draw, but it can write the exact words that make your image tool draw the right thing. That split is the whole point of this lesson, so hold onto it.

A one-line image prompt points at a fuzzy target.

Claude is not an image generator (a separate tool that turns text into a picture, like the popular ones you may have heard of). What Claude is brilliant at is writing the image prompt: the description you paste into that tool. You bring the idea, Claude writes the precise prompt, your image tool makes the picture.

Why this matters is that image tools are picky readers. Vague prompts ("a nice logo for a cafe") give vague, generic pictures. The same tool, fed a precise prompt, gives you something close to what you pictured. The quality lives in the words, and writing those words well is exactly Claude's strength.

A strong image prompt has a shape. Claude knows it: subject (the main thing), then concrete details (color, material, expression), then style or medium (photo, flat illustration, 3D), then lighting, then composition (close-up, wide, centered), then mood. State the job in one line, then stack four to six of those high-signal details.

Here is the before and after: You type "a coffee cup" into an image tool and get a generic stock-looking mug. You ask Claude instead, and it writes "a warm overhead photo of a single flat-white in a matte clay cup on a linen cloth, soft morning light, shallow focus, cozy and unhurried," which gives your tool everything it needs.

Let Claude art-direct the words. You do not need to know the structure by heart. Tell Claude what you want and the feeling, and ask it to write you three prompt versions to try. Paste each into your image tool, see which lands, and ask Claude to adjust the words from there.

Claude writes; your image tool draws. Bring the idea and the feeling, let Claude turn them into a precise, structured prompt, and the picture you get back finally matches the one in your head.

Describe an image so any image tool nails it mapThe concept system works when input, review, and human taste stay connected.

Try it (10 min)

Watch out for

Expecting Claude to produce the image itself. It writes the prompt; a separate image tool makes the picture, so always take Claude's words somewhere else to render them.
Pasting a one-line idea into your image tool. Vague in, vague out; let Claude expand it into a structured prompt with details, light, and mood first.
Treating the first prompt as final. Generate a few versions, test them in your tool, and have Claude adjust the words based on what actually came back.
Claiming a specific tool will do a specific thing. Tools differ and change, so describe what you want generally and let your own tool interpret it.

Paste this into Claude

I want you to write me image prompts I can paste into my own image tool. You don't make the picture; you write the words that make my tool make it.

What I need a picture of: [DESCRIBE, e.g. "a hero image for my candle brand's homepage"].
The feeling and look (from my moodboard brief if I have one): [PASTE OR DESCRIBE].
Where it will be used and any shape needed: [e.g. "wide website banner"].

Please:
1. Write me 3 different image prompts, each as one rich paragraph, using a clear structure: subject, concrete details, style or medium, lighting, composition, mood.
2. Briefly say how the 3 differ, so I know which to try first.
3. Remind me what to send back to you after I test them, so you can adjust the wording.

What a good response looks like

Here are three image prompts for a candle-brand hero (paste into your image tool):

**Version A (warm photo):** "A wide overhead photo of three matte-clay candles on a cream linen surface, one lit with a soft small flame, scattered dried herbs, warm late-afternoon light from the left, shallow focus, calm and unhurried mood."

**Version B (flat illustration):** "A wide flat vector illustration of three simple candles in earthy tones on a cream background, minimal shapes, one warm glow, lots of negative space, modern and gentle."

**Version C (moody close-up):** "A close-up photo of a single candle flame in a dark clay vessel, deep shadows, one warm highlight, rich and quiet mood, cinematic."

**How they differ:** A is homey and real, B is clean and brand-flexible, C is dramatic. Try A first for a homepage.

**Send me back:** which one came closest and what felt off (too busy, wrong light, etc.), and I will rewrite the words to fix it.

What good looks like

You can explain that Claude writes the prompt while a separate image tool draws the picture
You have 3 structured image prompts ready to paste into your image tool
Each prompt names a subject, concrete details, style, lighting, composition, and mood
You know to report back what the tool produced so Claude can adjust the words

When this breaks

Breaks when you ask Claude to 'make' the image. It will write words, not pixels, so a request aimed at the wrong tool only wastes a turn; send the words to an image tool.
Breaks when the prompt skips structure. A subject with no details, light, or mood leaves the image tool guessing, and guessing is what produces the generic stock look you were avoiding.

AI can help with this

Use Claude, ChatGPT, Gemini, Midjourney, Firefly, Canva, or another approved image tool based on where the asset will be made. Have the assistant write the image prompt, then critique the render against subject, composition, style, and brand fit.

A structured prompt names subject, setting, style, composition, and avoid list.

You can now

You can complete the lesson outcome against a real creative job, brand, asset, or campaign.

You can explain that the creative assistant writes the prompt while a separate image tool draws the picture.
You have 3 structured image prompts ready to paste into your image tool.
You can verify that each prompt names a subject, concrete details, style, lighting, composition, and mood.
You know to report back what the tool produced so the creative assistant can adjust the words.

Key takeaways

Claude does not generate images; it writes the structured prompt your image tool needs. Bring the idea and the feeling, let Claude stack subject, details, style, light, composition, and mood, and the picture comes back matching what you pictured.

Claude writes the image prompt; a separate image tool draws the picture from those words.
Image tools are picky readers, so the quality of the picture lives in the precision of the words.
A strong prompt has a shape: subject, concrete details, style or medium, lighting, composition, mood.
State the job in one line, then stack four to six high-signal details for the tool to follow.
Generate a few versions, test them in your tool, and have Claude adjust the wording from what comes back.

Go deeper

Was this helpful?

Up nextHeadlines, taglines, and copy that doesn't sound like AI

← Back to AI for Creative Work

I want you to write me image prompts I can paste into my own image tool. You don't make the picture; you write the words that make my tool make it. What I need a picture of: [DESCRIBE, e.g. "a hero image for my candle brand's homepage"]. The feeling and look (from my moodboard brief if I have one): [PASTE OR DESCRIBE]. Where it will be used and any shape needed: [e.g. "wide website banner"]. Please: 1. Write me 3 different image prompts, each as one rich paragraph, using a clear structure: subject, concrete details, style or medium, lighting, composition, mood. 2. Briefly say how the 3 differ, so I know which to try first. 3. Remind me what to send back to you after I test them, so you can adjust the wording.

Here are three image prompts for a candle-brand hero (paste into your image tool): **Version A (warm photo):** "A wide overhead photo of three matte-clay candles on a cream linen surface, one lit with a soft small flame, scattered dried herbs, warm late-afternoon light from the left, shallow focus, calm and unhurried mood." **Version B (flat illustration):** "A wide flat vector illustration of three simple candles in earthy tones on a cream background, minimal shapes, one warm glow, lots of negative space, modern and gentle." **Version C (moody close-up):** "A close-up photo of a single candle flame in a dark clay vessel, deep shadows, one warm highlight, rich and quiet mood, cinematic." **How they differ:** A is homey and real, B is clean and brand-flexible, C is dramatic. Try A first for a homepage. **Send me back:** which one came closest and what felt off (too busy, wrong light, etc.), and I will rewrite the words to fix it.