🎨

Create Images & Posters

Covers & illustrations — from visual brief to iterable prompts you can reuse

Task

For one product: 16:9 hero → composition iterate → 1:1 social version, ready for title overlay

Time

~25 min

Suggested tool: Doubao

Self-rate (1–5)

How this scenario works

Most image fails are vague prompts (“premium, techy, beautiful”) with no subject, material, composition, light, or use case. This scene runs one product through: hero shot → composition iterate → square social crop → negative cleanup → optional style shift, then overlay Chinese/English titles in a design tool. Good for: PDP hero drafts, social covers, deck art. Check platform ToS before commercial use; don’t fake real people or copy trademarks/IP. Example: product=sugar-free sparkling water; form=clear glass bottle with condensation; use=e-commerce hero; color=misty light blue. Acceptable: 16:9, clear subject, title-safe space, no gibberish text/watermarks/clutter — ready for Canva/Figma title overlay. Tool pick: Midjourney for polish; Flux for controllable/open-weight paths; Doubao-class for easier China access. Overlay critical text later.

Practice steps (with checks)

1Lock four things: subject (with material) + use case + aspect ratio + where title space must stay empty. Done when a teammate can visualize the frame from your brief.
2Generate 2–4 heroes with “Product cover draft”; pick one composition only. Done when title-safe space is obvious.
3Run “Composition iteration” to enlarge subject and clean the background. Done when the product dominates and no new props appear.
4For social, run “Square social version”. Done when the subject sits in the safe area.
5If you see gibberish text, watermarks, or clutter, run “Negative cleanup”. Done when the frame is clean enough to letter.
6(Optional) “Style shift” for a second look to align taste with teammates. Done when it’s the same subject, different style.
7Overlay titles in a design tool — don’t force perfect Chinese glyphs in the image model. Done when text is readable.
8Read the tool’s license before commercial use; if unsure, treat as an internal draft. Done when you know publish vs practice-only.

Practice Workbench

Product nameUse caseProduct formPrimary colorStylePrompt to QA

Product hero cover, 16:9. Subject: sugar-free sparkling water as a clear glass bottle with condensation. Use case: e-commerce hero cover. Background misty light blue + cream, clean modern light texture. Composition: product slightly below center ~40% of frame; title-safe space above and right; generous negative space. Soft studio light, subtle highlights, no harsh shadows. Forbidden: text, logos, QR, watermark, real people, brands/trademarks, cluttered props. One clear cover draft.

Open Midjourney

Outcome checklist

Self rating (1–5)

More prompts (5)

Composition iteration

Iterate the previous cover: same [product]/[form], palette led by [primary_color]. Enlarge product to ~55%; title space top-left; cleaner [style] studio backdrop; stronger rim light. No new text/logos/people/props; don’t recolor the product; no watermark; keep 16:9.

Square social version

Same [product] ([form]) as 1:1 social cover for [use_case]. Subject slightly above center with safe margins; simple [primary_color] backdrop. Clean feed-friendly look. No text, watermarks, or sticker clutter.

Negative cleanup

Regenerate the same [product] cover ([form], [primary_color] backdrop, 16:9). Fix prior issues: bad text, watermarks, extra hands/clutter, warped packaging. Keep subject clear + title space. Explicitly avoid: text, watermark, logo, qr code, extra hands, clutter, deformed label.

Style shift

Same [product]/[form] for [use_case], 16:9, but restyle as [style] (e.g., soft Japanese illustration / minimal vector / cinematic studio). Keep title-safe space; palette may echo [primary_color]. No text/watermark. For taste alignment, not final commercial art.

Prompt QA (chat model)

You are a visual prompt editor. Critique my image prompt: missing observables, vague words, conflicting constraints. Then rewrite a sharper version keeping subject [product], form [form], use [use_case], color [primary_color], style lean [style].

My draft prompt:
[draft]

Recommended tools

Best fitPaid

Midjourney

Top-tier AI image generation tool, renowned for artistic style and high-quality output.

AlternativeFreemium

Flux

Image generation model by the original Stable Diffusion team at Black Forest Labs, with industry-leading image quality and text rendering.

Free-friendlyFree

Doubao

ByteDance's AI assistant with text-to-image, voice chat, web search and other multimodal capabilities, excellent Chinese experience.

After you leave with templates,

ChooseNarrow tools with Choose LabBrowse guides & compares in Lab Related learning pathCreate Images & Short Videos Try another scenario →

Continue your learning path

Practice a scenario Choose Lab

How this scenario works

Practice steps (with checks)

1Lock four things: subject (with material) + use case + aspect ratio + where title space must stay empty. Done when a teammate can visualize the frame from your brief.

2Generate 2–4 heroes with “Product cover draft”; pick one composition only. Done when title-safe space is obvious.

3Run “Composition iteration” to enlarge subject and clean the background. Done when the product dominates and no new props appear.

4For social, run “Square social version”. Done when the subject sits in the safe area.

5If you see gibberish text, watermarks, or clutter, run “Negative cleanup”. Done when the frame is clean enough to letter.

6(Optional) “Style shift” for a second look to align taste with teammates. Done when it’s the same subject, different style.

7Overlay titles in a design tool — don’t force perfect Chinese glyphs in the image model. Done when text is readable.

8Read the tool’s license before commercial use; if unsure, treat as an internal draft. Done when you know publish vs practice-only.

Practice Workbench

Product nameUse caseProduct formPrimary colorStylePrompt to QA

Product hero cover, 16:9. Subject: sugar-free sparkling water as a clear glass bottle with condensation. Use case: e-commerce hero cover. Background misty light blue + cream, clean modern light texture. Composition: product slightly below center ~40% of frame; title-safe space above and right; generous negative space. Soft studio light, subtle highlights, no harsh shadows. Forbidden: text, logos, QR, watermark, real people, brands/trademarks, cluttered props. One clear cover draft.

Outcome checklist

Self rating (1–5)

Iterate the previous cover: same [product]/[form], palette led by [primary_color]. Enlarge product to ~55%; title space top-left; cleaner [style] studio backdrop; stronger rim light. No new text/logos/people/props; don’t recolor the product; no watermark; keep 16:9.

Same [product] ([form]) as 1:1 social cover for [use_case]. Subject slightly above center with safe margins; simple [primary_color] backdrop. Clean feed-friendly look. No text, watermarks, or sticker clutter.

Regenerate the same [product] cover ([form], [primary_color] backdrop, 16:9). Fix prior issues: bad text, watermarks, extra hands/clutter, warped packaging. Keep subject clear + title space. Explicitly avoid: text, watermark, logo, qr code, extra hands, clutter, deformed label.

Same [product]/[form] for [use_case], 16:9, but restyle as [style] (e.g., soft Japanese illustration / minimal vector / cinematic studio). Keep title-safe space; palette may echo [primary_color]. No text/watermark. For taste alignment, not final commercial art.

You are a visual prompt editor. Critique my image prompt: missing observables, vague words, conflicting constraints. Then rewrite a sharper version keeping subject [product], form [form], use [use_case], color [primary_color], style lean [style]. My draft prompt: [draft]