1. moss-image.com
  2. Prompts

MOSS-Image prompts that render text right

Official OpenMOSS examples with their published settings, plus templates written to the official prompt guidelines. Copy one, or send it straight to the generator and change the words.

How to write a MOSS-Image prompt

Quote every word you want rendered

Write the text inside quotation marks, exactly as it should appear — including punctuation, prices and line breaks.

Caption the finished image

Describe what the picture looks like — subject, layout, light, style — top to bottom, instead of giving an instruction.

Use English for detail, any language for text

English has the most training data, so it is the safest language for style details. The rendered words can be Chinese, English or both.

The library · 24 prompts

Seven lines of chalk handwriting

Boards & notes · 16:9 · Base
Seven lines of chalk handwriting
A candid classroom photograph, slightly off-center, with a worn dark green chalkboard filling most of the frame. Uneven window light, soft shadows, chalk dust, eraser streaks, subtle camera grain. Visible board frame and chalk tray, a blurred desk corner in the foreground. Natural, slightly irregular white chalk handwriting—not polished typography or a flat blackboard graphic. Keep all text sharp, fully visible, and free of glare.

The chalkboard reads exactly:

"MOSS-Transcribe-Diarize is our flagship SOTA model family
for turning real-world long-form audio into structured,
speaker- aware transcripts in one pass. Instead of stitching
together separate ASR and diarization systems, these
models jointly perform speech transcription and speaker
diarization, producing time-aligned text with precise
timestamps and consistent speaker labels such as [S01], [S02], and beyond."

Official example · Official example (Base, 1360 × 768, 50 steps, seed 2051794908, CFG 3, shift 6, no skip layer, negative “worst quality, bad anatomy, ugly face”). Note the closing instruction to keep all text sharp and fully visible, and the passage given verbatim after “reads exactly”.

Source

Rainy convenience store at night

Photography · 16:9 · Base
Rainy convenience store at night
A cinematic, photorealistic nighttime scene of a small neighborhood convenience store on a quiet, rain-soaked street. The warmly illuminated storefront occupies the right half of the frame, while the street recedes into soft mist on the left. A lone young adult wearing a charcoal coat stands just outside the entrance beneath a transparent umbrella, seen in three-quarter rear view, gazing through the glass. Amber light reveals shelves inside and spills across the sidewalk, mingling with muted teal neon reflections in shallow puddles. Individual raindrops bead on the umbrella and catch the storefront light; fine streaks of rain remain visible against the darker street. A slightly weathered bicycle leans beside the entrance. Distant apartment windows and a few out-of-focus streetlights add depth without distracting from the figure. Eye-level wide shot, 35mm lens aesthetic, natural perspective, balanced negative space, restrained teal-and-amber color grading, realistic wet asphalt and glass textures, soft highlight bloom, subtle film grain, and gentle atmospheric depth. Quiet, intimate, and nostalgic rather than dramatic; believable lighting, with the figure clearly silhouetted against the warm window. No prominent readable lettering, logos, or watermarks.

Official example · Official example (Base, 1680 × 944, 50 steps, seed 42, CFG 3, shift 5, negative-branch skip layer 26, negative “worst quality, bad anatomy”). A long caption that fixes framing, lens, light and color grade — and explicitly asks for no readable lettering.

Source

Pop-up paper storybook (Chinese prompt)

Illustration · 3:2 · Base
Pop-up paper storybook (Chinese prompt)
多层折叠式纸雕立体绘本,放在一张书桌上,背景纯净突出主题,比例 3:2 横版。翻开的跨页呈现「哪吒闹海」的场景:脚踏风火轮的少年、翻涌的层叠海浪与龙宫剪影,所有元素皆可精细折叠组合,呈现逼真细腻的纸张折叠质感与多层景深。构图统一采用正面视角,整体视觉梦幻唯美,色彩缤纷绚丽,充满奇幻生动的故事氛围。

Official example · Official example written entirely in Chinese (Base, 1536 × 1024, 50 steps, seed 2570432421, CFG 5, shift 5, negative-branch skip layer 26, negative “worst quality, bad anatomy, ugly face”). Material, viewpoint and mood are each stated once, plainly.

Source

Handwritten sticky note

Boards & notes · 1:1 · Base
A wide view of a refrigerator beside a kitchen counter with a fruit bowl. A small sticky note on the refrigerator door is handwritten in Chinese:“
记得买牛奶和鸡蛋!
牛奶两盒,鸡蛋一打。
顺便带一袋面包。”

Official example · The quick-start prompt from the official README (1456 × 1456, 50 steps, CFG 3). Line breaks inside the quotes set exactly where each handwritten line ends.

Source

Notes on a glass board (Chinese)

Boards & notes · 3:4 · Base
一个穿着"MOSI"标志的T恤的中国美女正拿着黑色的马克笔面向镜头微笑。她身后的玻璃板上手写体写着 "一、MOSS-Image的技术路线:5B 的单流 DiT 模型,基于 Muon 训练。二、MOSS-Image的模型特色:1、复杂文字渲染。支持中英渲染、自动布局; 2、使用 Layout 进行精准图像控制,支持文字编辑和物体增减等自由编辑。三、MOSS-Image的后续发展:探索 R2I 参考生图、助力生成式AI发展。"

Official example · From the official user guide, with negative prompt “worst quality, bad anatomy”. A long Chinese passage the model lays out on its own.

Source

Minimal café poster

Posters · 4:5 · SFT
A minimalist café poster on warm cream paper. At the top, a large bold dark-brown headline reads "SLOW COFFEE". In the middle, a ceramic pour-over set photographed from above in soft morning light. At the bottom, a small line of text reads "Hand-poured · Daily 8:00–20:00".

Top-to-bottom order mirrors the reading order, and each text block has its own size and position.

Indie concert gig poster

Posters · 2:3 · SFT
A screen-printed gig poster in two inks, red and navy, on off-white paper with slight misregistration. A stylized illustration of a cassette tape fills the center. The band name at the top in huge condensed letters reads "NIGHT SWIMMING". At the bottom, three lines:
"LIVE AT THE HARBOR HALL"
"FRI 14 NOV · DOORS 8PM"
"TICKETS $18"

Print technique and ink count give a consistent style; the info block uses one line per fact.

Bilingual sale banner

Posters · 1:1 · SFT
A bold square sale graphic with a vermilion background and a cream frame. Centered huge white Chinese characters read "限时五折". Directly below, smaller cream uppercase text reads "50% OFF THIS WEEKEND". A few paper shopping bags are scattered along the bottom edge. Flat graphic style, no gradients.

Stating the relationship (“directly below”) between the two languages keeps them as one lockup.

Neon bar sign in the rain

Signs · 16:9 · SFT
A rainy night in an old Shanghai alley. Above a small bar door, a pink neon sign reads "OPEN LATE" and beneath it "深夜营业". Reflections on wet cobblestones, cinematic 35mm photograph, shallow depth of field.

The sign is part of a photograph, so the camera language (35mm, depth of field) sets the realism.

Bakery storefront fascia

Signs · 4:3 · SFT
A sunny street-level photo of a small bakery storefront with a painted wooden fascia. The fascia reads "MORNING CRUMB" in cream serif letters, with "BAKERY · SINCE 1998" in smaller letters underneath. A chalkboard A-frame on the sidewalk reads "Fresh sourdough
today".

Two separate text surfaces (fascia and chalkboard), each with its own quoted copy and line break.

Subway wayfinding sign

Signs · 3:2 · SFT
A clean photo of a subway station wayfinding sign mounted on white tiles. Dark grey panel, white text and arrows. Left: an arrow pointing left with "Exit A · 出口A". Right: an arrow pointing right with "Line 2 · 2号线". Even fluorescent light, sharp focus on the sign.

Spatial words (“Left:”, “Right:”) split the sign into zones the model can fill separately.

Literary novel cover

Covers · 2:3 · SFT
A literary novel cover with a deep moss-green background and a single delicate paper moth illustrated in the center. The title in large elegant serif letters reads "The Quiet Season". At the bottom, the author name reads "LIN WEI". Matte paper texture.

One central image leaves clean space for the title and author.

Podcast cover art

Covers · 1:1 · SFT
Square podcast cover art: a halftone portrait-free illustration of two vintage microphones facing each other on a mustard-yellow background. Heavy black geometric sans-serif title across the top: "SIDE CHANNEL". A small black badge in the bottom-right corner reads "EP. 42".

Small elements get an explicit corner, so they don’t collide with the title.

Three-step infographic

Boards & notes · 3:4 · SFT
A clean flat infographic on a white background titled "How Cold Brew Works". Three numbered steps from top to bottom, each with a simple line icon on the left and a caption on the right:
"1  Grind coarse"
"2  Steep 16 hours in cold water"
"3  Filter and serve over ice"
Navy and light-blue palette, generous spacing.

Numbered lines and a stated icon-left/caption-right pattern keep the steps consistent.

Meeting whiteboard

Boards & notes · 16:9 · SFT
A photo of an office whiteboard after a planning meeting, slightly angled, natural daylight. Blue marker handwriting at the top reads "Q1 LAUNCH PLAN". Below it a simple timeline arrow with three labels: "Beta · Jan", "Pricing · Feb", "Launch · Mar". A red circled note in the corner reads "ship it!".

Marker colors and handwriting make text feel natural rather than typeset.

Honey jar label

Products · 1:1 · SFT
A studio product photo of a glass jar of honey with a kraft-paper label, on a beige background with soft shadows. The label reads "WILDFLOWER HONEY" and below it "槐花蜂蜜 · 500g".

Short label copy on a curved surface — keep it to two lines for the cleanest result.

Tea box packaging

Products · 4:5 · SFT
Packaging design photo of a rectangular tea box in matte sage green, standing next to a small celadon cup. Front of the box: a thin gold line drawing of mountains, the brand "山雾" in large brush-style characters, and "MOUNTAIN MIST · OOLONG · 100g" in small spaced capitals at the bottom.

Describe the box face as its own canvas: graphic first, then brand, then small print.

Morning market street

Photography · 3:2 · Base
A documentary photograph of a morning wet market in Chengdu, eye-level, 35mm. A vendor in a blue apron weighs green peppers on an old scale; steam rises from a breakfast stall behind her. Soft overcast light, natural colors, slight film grain. Handwritten cardboard price tags on the produce read "青椒 4元/斤" and "土豆 2元/斤".

Base tends to look more photographic; small in-scene text (price tags) adds authenticity.

Window-light portrait

Photography · 4:5 · Base
A quiet portrait of an elderly watchmaker at his workbench by a large window, looking down at a pocket watch through a loupe. Soft directional window light from the left, dark warm background, shallow depth of field, 85mm lens, muted earthy colors, fine detail on hands and tools.

Light direction plus lens choice is usually enough to steer a portrait.

Anime tag-style prompt (Base)

Illustration · 2:3 · Base
best quality, highres, 1girl, solo, school uniform, holding umbrella, rain, puddle reflection, city street at dusk, neon lights, looking back, wind, cinematic lighting

Base reads booru-style tags. The official order is quality → score → meta → year → character → series → artist → general.

Watercolor travel postcard

Illustration · 3:4 · SFT
A loose watercolor illustration of the West Lake in Hangzhou with a stone bridge, willow trees and a small boat, white paper showing through. Hand-lettered text in the upper-left corner reads "Greetings from Hangzhou" and a small stamp in the lower-right corner reads "杭州 2026".

Corner placement for text keeps it out of the painted subject.

Official prompts are quoted verbatim from OpenMOSS and link to their source. Templates are starting points: results vary with seed and model, so check text before you publish.

Questions

What makes a good MOSS-Image prompt?

A caption of the finished image, written top to bottom, with every word to render inside quotation marks and explicit line breaks. MOSS-Image accepts up to 4,096 tokens, so long, specific prompts are fine — the official examples run to several hundred words.

Should I write prompts in English or Chinese?

Both work. OpenMOSS recommends English because it has the most training data, but Chinese prompts produce good results too. The text you want rendered can be in either language regardless of the prompt language.

What negative prompt should I use?

The official guide recommends: nsfw, worst quality, bad anatomy, watermark, jpeg artifact. Our generator fills this in by default under Advanced settings.

Why does the text come out slightly wrong?

Long passages, rare characters and very small text are the hardest cases. Shorten each line, add line breaks yourself, try SFT, or generate again with a different seed. Always proofread before publishing.