Traditional 水墨山水 (Chinese ink landscape) is defined by two things diffusion models forget: generous negative space and calligraphic brushwork. Without them, a “Chinese landscape” prompt produces colored chinoiserie that misses the genre entirely. The 12 prompts below force vertical composition, rice-paper texture and explicit “mostly negative space” so the result reads as ink rather than stylized color. They are written for the two models most readers reach for as of June 2026 — Midjourney V8.1 (released April 30, 2026) for aesthetic control, and GPT Image 2 (ChatGPT’s image model since April 2026) for prompt-following and in-chat editing.
TL;DR
- Always say “monochrome ink, no color” and reserve 60–80% of the frame as empty space — that single instruction does more for authenticity than any style word.
- Use a vertical ratio:
--ar 2:3or--ar 9:16in Midjourney; pick a 2:3 or 9:16 size in GPT Image 2’s aspect-ratio control (it has no--arflag). - In Midjourney,
--noreads each word independently, so write--no color, saturation, chinoiserierather than long phrases. - A single red seal stamp is the strongest “real ink painting” signal — keep it in the prompt.
- Reference the Northern Song “three distances” (三远) to get believable far/middle/near depth instead of a flat backdrop.
Best for
- Book covers and editorial illustration for Chinese cultural topics
- Tea, incense and heritage-brand visual identity
- Scroll-format social posts where vertical composition is native
- Cultural-exhibition collateral that needs an authentic ink language
- Wuxia / xianxia fiction covers that need more than fantasy painting
- Travel and slow-living brands referencing classical landscape
Which model to use
| Model (June 2026) | Strength for ink landscape | Aspect control | Negative prompts |
|---|---|---|---|
| Midjourney V8.1 | Best aesthetic; reads “ink wash”/“sumi-e” well | --ar 2:3, --ar 9:16 | --no color, saturation, chinoiserie |
| GPT Image 2 (ChatGPT) | Best prompt-following + iterate in same chat | Aspect-ratio picker (2:3, 9:16, 1:1) | Plain language: “monochrome, no warm tints, no extra text” |
| Flux 2 / Ideogram v3 | Good when you need crisp title text on the canvas | Width/height fields | Built-in negative field |
Midjourney still gives the most convincing brushwork; GPT Image 2 wins when you want to nudge “make the sky emptier” or “remove the color tint” conversationally without re-rolling the whole prompt.
1. Misty mountains (Song-dynasty influence)
traditional Chinese ink landscape, misty distant mountains in pale wash, foreground pine tree with calligraphic brush strokes, middle-ground hidden, rice paper texture, generous negative space (top half mostly empty), vertical scroll composition, Song dynasty landscape influence, monochrome ink, no color2. Lone boat on river
minimal Chinese ink wash, lone boat on a wide river, single fisherman silhouette, soft fog dissolving the far shore, mountains barely visible in distance, sumi-e style, mostly negative space (80% empty), vertical composition, monochrome3. Bamboo grove with figure
Chinese ink painting, dense bamboo grove on the right third, varied stroke thickness, subtle gradient wash, single small scholar figure for scale at the bottom, vertical scroll feel, generous empty sky on the left, monochrome ink only4. Solitary plum branch (花鸟 crossover)
ink painting of a single plum branch entering from the upper right corner, sparse blossoms in pale gray wash, large empty space below, calligraphic seal and red signature stamp in the lower left, rice paper texture, monochrome with one accent of red only on the seal5. Far-middle-near depth layering (三远)
This is where most “Chinese landscape” outputs fall apart: the model paints one flat plane. The fix is to name the depth explicitly. Northern Song painter Guo Xi codified three viewpoints in his treatise Linquan Gaozhi (林泉高致): 高远 (high distance, looking up at a towering peak), 深远 (deep distance, peering past near mountains into receding ones), and 平远 (level distance, gazing flat toward a far shore). Telling the model which one you want — and forcing three named planes — is what produces real depth.
Chinese ink landscape with explicit three-plane composition:
- Far: pale mountain silhouettes barely visible
- Middle: a single mist band dissolving form
- Near: detailed pine and rock in confident strokes
70% negative space, vertical scroll, monochrome ink, rice paper texture, no colorFor a sweeping high-distance feel, add low viewpoint looking up at a towering peak. For a quiet 平远 mood, add flat far horizon, wide calm water, low mist.
6. Storm and crag
dramatic Chinese ink landscape, sheer cliff face on the left, dark heavy ink with wet-on-wet bleeding, wind-bent pine clinging to rock, swirling mist below, distant peaks lost in storm, vertical composition, mostly dark ink with negative space cutting through, no color7. Tea pavilion in the mountains
Chinese ink landscape, small thatched pavilion nestled among trees mid-canvas, narrow path winding down to a stream, far peaks in pale wash, two tiny figures at the pavilion for scale, restrained brushwork, monochrome ink, vertical scroll, generous sky8. Wuxia film-still palette
Chinese ink landscape with cinematic wuxia framing, lone swordsman silhouette on a high cliff, dramatic backlight through mist, jagged peaks fading into distance, calligraphic brush, monochrome ink with subtle dark-teal undertone only, vertical composition, room for vertical title text on the right9. Four-season set
Set of 4 Chinese ink landscapes for [spring / summer / autumn / winter]. Each:
- Same vertical scroll composition
- Same brush language and rice paper texture
- Season cue: spring = plum bloom; summer = lotus pond; autumn = bare branches + crane; winter = snow-cloaked pine
- Monochrome ink, one consistent red seal in lower-left of eachTo keep all four panels visually consistent in Midjourney V8.1, generate the first one you like, then add --sref <code> (or the image URL) to the next three so they share one style signature.
10. Modern editorial crop (horizontal)
Chinese ink landscape adapted for horizontal editorial use, wide aspect ratio, mountains stretching left to right, mist dissolving the middle, single boat on the lower right, calligraphic strokes, mostly negative space in the upper half for headline placement, monochrome, rice paper texture, designed for magazine spread11. Square social composition
Chinese ink landscape, 1:1 square composition, single distant peak centered with mist below, generous negative space, calligraphic seal stamp in lower-right corner only, monochrome, rice paper texture, designed for social-post share — title overlays cleanly on the upper third12. Title-overlay scroll for book cover
Chinese ink landscape book cover, vertical 2:3 ratio, scroll composition, single tall mountain silhouette on the right, mist and empty sky on the upper-left two thirds reserved for vertical title in classical font, foreground pine + small figure for scale on the lower right, monochrome ink, one red seal accent onlyHow to refine
- Force the ratio. A vertical aspect ratio (
--ar 2:3or--ar 9:16in Midjourney; the 2:3 or 9:16 size in GPT Image 2’s aspect-ratio control) matches the scroll. Horizontal ratios fight the genre. - Negate the right way. In Midjourney add
--no color, saturation, photoreal, chinoiserie. Note the V8.1 quirk:--noreads each word on its own, so a phrase like--no modern buildingsbecomes “no modern” + “no buildings” — keep negatives single-concept. In GPT Image 2 there is no--noflag; say it in plain prose: “monochrome only, no warm tints, no extra text or lettering.” - If color still leaks, lower the word “painting” and raise “ink wash” and “sumi-e”; the model defaults to subtle warm tinting unless you suppress it.
- Iterate in chat (GPT Image 2). Instead of re-rolling, follow up with “make the upper half emptier” or “remove the color tint, keep it pure ink” — it edits the existing image in place.
Common mistakes
- Color spilling into the ink — the model defaults to subtle warm tinting if you don’t negate it.
- No negative space — the landscape feels crowded; the genre depends on emptiness as a compositional element.
- Photorealistic textures sneaking in (grass blades, glossy stone, individual leaves).
- Horizontal composition where the canonical form is a vertical scroll.
- Decorative chinoiserie patterns (cloud scrolls, dragons) replacing real landscape language.
- Forgetting the seal — a single red stamp is what signals “real ink painting” most strongly.
- Writing one flat plane instead of named far/middle/near depth (see section 5).
FAQ
Why does my “Chinese landscape” prompt come out in color?
Diffusion models default to warm tinting on cultural prompts. Add “monochrome ink, no color” inside the prompt and negate it again in the model’s negative field (--no color, saturation in Midjourney). If it persists, weaken “painting” and strengthen “ink wash, sumi-e”.
Midjourney V8.1 or GPT Image 2 for ink landscapes?
Midjourney V8.1 produces the most convincing brushwork and rice-paper texture, and --sref keeps a series consistent. GPT Image 2 follows complex composition instructions more literally and lets you refine conversationally (“make the sky emptier”) in the same chat. Many designers draft in GPT Image 2, then re-render the winner in Midjourney for final texture.
What aspect ratio should I use?
Vertical: --ar 2:3 for book covers and --ar 9:16 for phone-screen and story formats. Midjourney added more ratios (including 4:5 and 5:4) in early 2026 if you need a gentler crop. GPT Image 2 has no --ar flag — choose the size in its aspect-ratio control before generating.
How do I get believable far/middle/near depth? Name three planes explicitly and assign each a different ink weight (pale far, dissolving middle, confident near), as in section 5. This mirrors Guo Xi’s three-distances (三远) framework and reads as real depth rather than a flat backdrop.
How do I keep a four-panel seasonal set consistent?
Generate the first panel, then pass its --sref <code> or image URL to the remaining three in Midjourney so they share one style signature. In GPT Image 2, ask it to “match the brush style and paper texture of the previous image” in the same chat.
Related
- Hanfu character prompts — companion Chinese-aesthetic library
- Chinese aesthetic character prompts — character-side prompts in the same world
- Chinese-style lyric prompts — audio language for the same audience
- Suno Chinese-style workflow — production tutorial
Tags: #Chinese-style