Xiaohongshu and YouTube are visually opposite. Xiaohongshu wants minimal, premium, spacious. YouTube wants exaggerated, bold, high-contrast. The 10 prompts below follow the viral-note pattern that reliably reads as Xiaohongshu: natural light, a light backdrop, one subject, and a large top space reserved for the Chinese title.
TL;DR
- Lock five things in every prompt:
--ar 3:4, natural light, a light single-color backdrop, one subject with 1–2 props, and 25–35% empty space at the top. - Pick the model by job: Nano Banana Pro (Gemini 3 Pro Image) or GPT Image 2 if you want the Chinese title baked into the image; Midjourney V8 if you want the most magazine-like flat-lay and will add the title in post.
- As of June 2026, GPT Image 2 and Nano Banana Pro render Chinese characters legibly (~99% accuracy on GPT Image 2), so the old “never let the model type text” rule only applies to Midjourney.
- The two highest-leverage switches are a beige-linen surface and soft window light. Add
--ar 3:4or you get a square that the feed crops badly.
Which model for Xiaohongshu covers (June 2026)
All three top models nail soft, photographic, light-backdrop scenes. They differ mainly on aspect-ratio syntax and whether you can trust them to type Chinese.
| Model | Best for | 3:4 syntax | Chinese title in-image? |
|---|---|---|---|
| Midjourney V8 (alpha) | Most editorial flat-lays and atmosphere | append --ar 3:4 | Avoid; compose title in post |
| GPT Image 2 (ChatGPT Images 2.0) | Title baked in, layout reasoning | ask for “vertical 3:4” / “portrait” | Yes, ~99% character accuracy |
| Nano Banana Pro (Gemini 3 Pro Image) | Multilingual text, up to 4K | request “4:5” or “3:4 portrait” | Yes, industry-leading multilingual |
Midjourney V8 (alpha, 2026) improved text rendering, but Latin glyphs benefit far more than Chinese, so for CJK titles still composite in post. GPT Image 2 launched April 21 2026 with 2K output and reasoning over layout; Nano Banana Pro (Gemini 3 Pro Image) reached general availability in June 2026 with up to 4K and the strongest multilingual text. The --ar 3:4 flag in the prompts below is Midjourney syntax; for GPT Image 2 and Nano Banana Pro, drop the flag and say “vertical 3:4 portrait” instead.
What a high-quality prompt should contain
Six required elements for a Xiaohongshu cover:
- Strict 3:4 ratio: the Xiaohongshu default; without it you get a 1:1 the feed crops badly
- Top-down or eye-level: top-down (flat-lay) is the signature; half-body eye-level for people
- Natural light:
soft natural window light,soft overcast daylight, never dramatic - Light backdrop: beige linen, light wood, pink, or cream
- Top negative space: leave 25–35% for the Chinese title
- Single subject plus 1–2 props: minimalist but layered
10 copy-ready prompt templates
The --ar 3:4 flag is for Midjourney. On GPT Image 2 or Nano Banana Pro, remove it and append “vertical 3:4 portrait” instead.
1. Minimal lifestyle flat-lay
Best for: Lifestyle / food notes
Xiaohongshu cover, flat-lay top-down photo of a notebook, latte, croissant on a beige linen surface, soft natural window light, warm minimal aesthetic, large empty space at top for title text, --ar 3:42. Travel atmospheric back-view
Best for: Travel / city exploration notes
Xiaohongshu travel cover, a vertical scene of a person from behind looking out at a misty Kyoto bamboo forest, soft overcast light, slight rim from morning sun, calm tasteful Japanese vibe, large empty area at top for title, --ar 3:43. Skincare pink top-down
Best for: Skincare / beauty rec notes
Xiaohongshu beauty cover, top-down on a soft pink linen, three small skincare bottles, dried petals around, soft natural light, clean refined editorial, large empty space at top, --ar 3:44. Home cozy-corner
Best for: Home / interior decor notes
Xiaohongshu home cover, soft cream interior with a single armchair, plant, golden afternoon light from window, calm cozy living-aesthetic, large empty space at top for title text, --ar 3:45. Fashion street candid
Best for: OOTD / outfit notes
Xiaohongshu fashion cover, half-body of a young East Asian woman in oversized cream blazer, walking on a sunny side-street, candid mid-step, soft warm afternoon light, large empty space at top, 35mm f/2.0 documentary, --ar 3:46. Healthy bowl artful
Best for: Healthy meal / diet notes
Xiaohongshu food cover, top-down of an artful mango smoothie bowl on a pastel green linen, lifestyle health-food editorial, large empty space at top for "I tried this" style title, --ar 3:47. Study desk flat-lay
Best for: Study / exam-prep notes
Xiaohongshu studying cover, top-down workspace flat-lay: open notebook with handwriting, MacBook, coffee, a single yellow flower, beige linen surface, soft natural daylight, large empty title area, --ar 3:48. Lipstick single-hand
Best for: Makeup single-product review
Xiaohongshu makeup cover, close-up of a hand holding a single lipstick over a clean cream background, soft window light, refined minimal beauty, large empty area on top, --ar 3:49. Pet warm daily
Best for: Pet / cute-pet notes
Xiaohongshu pet cover, top-down of a sleeping golden retriever puppy on a warm beige rug with a single small toy, soft natural light, calm tasteful pet content, large empty area at top, --ar 3:410. Wellness top-down
Best for: Fitness / wellness notes
Xiaohongshu wellness cover, top-down of a yoga mat with a water bottle, a folded towel, a single white flower, soft morning daylight on a warm wood floor, calm wellness aesthetic, large empty title area, --ar 3:4Adding the Chinese title
Two routes as of June 2026:
- Bake it in (GPT Image 2 or Nano Banana Pro): add a literal instruction such as
with the Chinese title "三天搞定旅行计划" in the top-left, clean sans-serif, dark text. GPT Image 2 advertises ~99% character-level accuracy across Latin, Chinese, Japanese, Korean; Nano Banana Pro leads on multilingual layouts. Still proofread, and keep titles short (4–10 characters) so the model does not mangle a stroke. - Composite in post (any model, required for Midjourney): generate the clean cover, then set type in Canva, Snapseed, or Picsart. This gives you full control over kerning and font. Reliable Chinese fonts: Source Han Serif (premium), PingFang (modern), Zhanku Wenyi (literary).
For high-stakes covers, post-compositing still wins on typographic control even when the model can type.
Common mistakes
vibrant colorful— Xiaohongshu wants soft; saturated reads cheap- Stacking elements (laptop + coffee + book + flowers + plant + watch)
- Exaggerated expression — Xiaohongshu wants ease, not shock
- Dramatic single spotlight — pulls toward editorial-product, not Xiaohongshu
- No aspect ratio — defaults to 1:1, wrong for the feed
- Letting Midjourney type the Chinese title — it garbles CJK; composite instead
How to push results further
- Beige linen plus natural light are the two most reliable Xiaohongshu switches
- Viral fits: templates 1 / 6 / 7, the top-down flat-lays
- Half-body people: 35mm documentary style for outfit / beauty / travel
- A single palette (pink, beige, or cream) beats mixed colors
- If you want the model to type the title, use GPT Image 2 or Nano Banana Pro and keep it to 4–10 characters
FAQ
Q: Which model is best for Xiaohongshu covers in 2026?
A: For the most magazine-like flat-lay, Midjourney V8, then add the title in post. If you want the Chinese title baked into the image, GPT Image 2 or Nano Banana Pro (Gemini 3 Pro Image), which both render Chinese legibly as of June 2026.
Q: Can the AI write the Chinese title now, or do I still add it later?
A: GPT Image 2 (around 99% character accuracy) and Nano Banana Pro can type short Chinese titles directly. Midjourney still garbles Chinese, so composite the title in post for it. Either way, keep titles to 4–10 characters and proofread the strokes.
Q: Xiaohongshu vs YouTube biggest difference?
A: Xiaohongshu is “I want to live this way” minimal aspiration. YouTube is “I must click this” attention spike. Completely different motivation, so the prompts share almost nothing.
Q: How do I get a stable flat-lay top-down?
A: top-down, flat-lay, and overhead view all trigger it. Pair with a same-color backdrop, natural light, and about three items.
Q: Why does my Xiaohongshu cover look like just a life photo?
A: Usually no top negative space and no color discipline. Reserve 25–35% at the top and lock one palette, and it reads Xiaohongshu-grade instantly.