Xiaohongshu (小红书) card carousel

Contract

When the task is to turn content into a Xiaohongshu / 小红书 image post (a picture-note, "图文"), emit a MULTI-CARD carousel that conforms to the platform's real card format: fixed 3:4 portrait cards at 1080×1440 px, a cover card first, exactly one idea per body card, a follow/save call-to-action last, and a small corner watermark. Apply whenever the deliverable is 小红书 / RedNote picture-note cards, whether as HTML, images, or a layout spec.

Rules

1. Card size — 3:4 portrait, 1080×1440 px. Every card is EXACTLY 1080 px wide by 1440 px tall

(the platform's recommended vertical export; a 3:4 portrait). Never square (1:1, 1080×1080), never 4:5, never landscape (16:9 / 4:3). In Tailwind: w-[1080px] h-[1440px]; in plain CSS: width:1080px;height:1440px.

1. Card count — a small discrete set, hard-capped at 18. Output several SEPARATE cards, one per

idea. Xiaohongshu allows at most 18 images in a single note, and ≤9 reads best — so 3–9 cards for typical content, more only when the content is long, and NEVER more than 18. Do not collapse the whole piece onto one tall page.

1. Stack for screenshotting. Render the cards as distinct fixed-size blocks (e.g. a vertical

flex flex-col column) so each card can be screenshotted on its own and the set exported together — not one continuously scrolling document.

1. Card 1 is the COVER. The first card is a title/cover: an OVERSIZED headline, a one-line

subtitle, and a short curiosity/value HOOK label written in the 小红书 idiom — e.g. 干货预警, 建议收藏, 全文干货, 码住, 收藏不亏, 超实用. The hook is a badge/pill, visually distinct from the body.

1. Body cards — ONE idea each. Each middle card carries exactly ONE core point: a short punchy

heading + an emoji + 1–2 concrete examples. Never cram several tips onto one card; split them across cards instead.

1. Last card — summary + CTA. The final card recaps and asks for engagement using Xiaohongshu's own

actions: 关注 (follow), 收藏 (save), 点赞 (like), 评论 (comment). A generic "thanks for reading" is not a CTA.

1. Watermark. Place a small watermark in a card corner (typically bottom-right) — author

handle/name and/or date — on the cover at minimum.

1. Phone-first legibility. Large type, wide line-height, high contrast, generous whitespace, rounded

elements, a soft palette (Morandi / pink). These finish the look but are secondary to rules 1–7.

Worked examples

One BEFORE (the base model's wrong default) → AFTER (conforming) per rule.

Card size:

html
<!-- BEFORE: square, Instagram-style -->
<div style="width:1080px;height:1080px"> ... </div>
<!-- AFTER: 3:4 portrait, the Xiaohongshu size -->
<div class="w-[1080px] h-[1440px]"> ... </div>

Cover card with a hook label + watermark:

html
<!-- AFTER -->
<section class="w-[1080px] h-[1440px] relative flex flex-col justify-center p-20 bg-[#f3ece4]">
  <span class="self-start px-5 py-1.5 rounded-full bg-[#c8a98a] text-white text-[32px]">干货预警</span>
  <h1 class="text-[110px] font-black leading-tight mt-6">3 个早起习惯</h1>
  <p class="text-[40px] text-neutral-600 mt-4">坚持三周, 状态肉眼可见变好</p>
  <span class="absolute bottom-8 right-10 text-[24px] text-neutral-400">@小红薯 · 2026.07</span>
</section>

Body card — one idea (not a crammed list):

html
<!-- BEFORE: three tips on one card -->
<div>1) 固定起床时间 2) 起床先喝水 3) 拉开窗帘晒太阳</div>
<!-- AFTER: one idea per card -->
<section class="w-[1080px] h-[1440px] flex flex-col p-20">
  <h2 class="text-[72px] font-bold">☀️ 起床先拉开窗帘</h2>
  <p class="text-[40px] mt-8">让阳光进来五分钟, 褪黑素快速下降。换一副透光浅色纱帘更省事。</p>
</section>

Last card — CTA (not "thanks for reading"):

html
<!-- BEFORE -->
<div>谢谢观看!</div>
<!-- AFTER -->
<section class="w-[1080px] h-[1440px] flex flex-col justify-center items-center text-center">
  <h2 class="text-[80px] font-black">今天你打卡了吗?</h2>
  <p class="text-[44px] mt-8">关注 收藏 点赞, 明天继续更新 👇</p>
</section>

Edge cases & exceptions

• Very little content (1–2 ideas): still make at least 3 cards — cover + point(s) + CTA. Never a

single card.

• A lot of content (15+ ideas): GROUP related ideas so the total stays ≤ 18 (ideally ≤ 9); never

exceed 18 cards even if that means merging or dropping the weakest points.

• A cover-only request: still 1080×1440, still a hook label + a watermark; if body content exists,

split it into cards anyway.

• A wide/landscape source (a chart, a screenshot): re-frame it inside the 3:4 card; do not switch the

card itself to landscape.

• Non-Chinese audience: keep the 3:4 / 1080×1440 size and the cover → one-idea → CTA structure; the

hook and CTA text may be localized, but the default idiom is the 小红书 one above.

Do / Don't

• Don't size cards 1:1 (1080×1080) or 16:9 · Do 3:4 portrait, 1080×1440.
• Don't put the whole piece on one long page · Do one idea per fixed-size card, ≤ 18 cards.
• Don't open with a plain heading · Do a cover card with an oversized title + a 干货预警 / 建议收藏 hook.
• Don't end with "谢谢观看" / "thanks for reading" · Do a 关注 / 收藏 / 点赞 / 评论 CTA card.
• Don't forget attribution · Do a small corner watermark (handle / date).
• Don't use tiny body text · Do phone-legible large type with wide spacing.

Common mistakes (the base model's wrong defaults)

• Defaulting to a square (1080×1080) or a single full-viewport page instead of 3:4 1080×1440 cards.
• Producing one long scrolling document rather than discrete screenshot-sized cards.
• Cramming every point into one card (a bulleted list) instead of one idea per card.
• A bland title with no hook label; no cover / body / CTA structure.
• Ending without a follow/save call-to-action.
• Omitting the corner watermark.

Quick checklist

• ] Every card 3:4 portrait, 1080×1440 px (not 1:1 / 4:5 / 16:9).
• ] Multiple discrete cards, ≤ 18 (ideally ≤ 9), one idea each.
• ] Card 1 = cover: oversized title + subtitle + a 小红书 hook label (干货预警 / 建议收藏 / 码住).
• ] Last card = summary + 关注 / 收藏 / 点赞 / 评论 CTA.
• ] Small corner watermark (handle / date).
• ] Phone-legible: large type, wide spacing, soft palette.