Aspect Ratio Midjourney (--ar) Plus FLUX, SDXL, and ChatGPT

Aspect ratio sets the canvas shape before the model draws a single pixel. Searchers who type aspect ratio Midjourney want the --ar flag, the common width:height pairs, and how the same intent maps to FLUX, SDXL, and ChatGPT image tools. You leave with paste-ready Midjourney parameters, pixel size notes for SDXL and FLUX, prose instructions for ChatGPT image, and a pick-ratio-first workflow so you stop regenerating square crops into vertical posts. Soft path when a reference photo already has the frame you want: open PromptMake /image, then append or confirm --ar for Midjourney before you spend a full generation grid.

Why aspect ratio belongs in the prompt, not the crop

Cropping a finished square into a Story or a YouTube thumbnail throws away composition. The model packed the subject, negative space, and horizon into the frame you asked for. A late slice of that frame cuts faces, product edges, and title-safe margins. Set the ratio on the first generate so the subject sits where the platform needs it: centered for a 1:1 feed tile, left-weighted for a 16:9 hero with room for UI chrome, headroom for a 9:16 vertical face crop.

Midjourney treats --ar as a composition control, not a post-process crop. FLUX and SDXL bake width and height into the latent grid. ChatGPT image tools take size presets or a short "make this vertical" instruction in chat. Same creative decision, different dials. Lock the delivery format (Instagram feed, Reel, blog header, print) before you write lighting or style words. Ratio first, adjectives second.

Strong fit for:

• Social and ads teams who ship the same hero across feed, Story, and landscape banners
• Product photographers who need catalog squares and PDP verticals from one brief
• Concept artists who storyboard 16:9 and 21:9 before polish
• Anyone who ports Midjourney --ar prompts into FLUX, SDXL, or GPT Image without guessing pixels

Aspect ratio Midjourney: --ar syntax that sticks

Midjourney defaults to a square. You override that with --ar or --aspect at the end of the prompt, after the prose and before or beside other parameters like --style raw, --stylize, and --v 7. Write whole numbers separated by a colon: --ar 16:9, not --ar 1.78:1. Midjourney rejects decimals. Convert 2.39:1 cinema to --ar 239:100. You can paste raw pixel pairs such as --ar 1920:1080 and Midjourney simplifies them to 16:9. On the web Imagine bar and in Discord, the flag lives at the end of the prompt string. Settings can pin a default ratio for future runs; clear that default when a project needs a different shape. As of mid-2026, Midjourney v7 (and V8 Alpha where you have access) accepts a wide range of ratios, with experimental extremes past about 2:1 in either direction more likely to stretch, repeat, or odd-frame subjects. Pick common platform ratios for client work. Save ultrawide and ultra-tall for tests. The subsections below cover defaults, syntax traps, and how --ar interacts with references and the Editor.

Defaults, whole numbers, and safe ranges

Default without a flag: --ar 1:1. Common production pairs: --ar 16:9 for landscape video and site heroes, --ar 9:16 for Stories and Reels, --ar 4:5 for Instagram portrait feed, --ar 3:2 or --ar 2:3 for photo-print habits, --ar 4:3 for older display and slide decks, --ar 21:9 for cinematic establishing frames. Keep integers. For letter sizes and print oddities, drop decimals: 8.5×11 becomes --ar 85:110 (simplify to 17:22 if you prefer). Extreme ratios near the documented 1:14 to 14:1 band can run, yet faces and products may warp or duplicate along the long axis. Start inside the everyday set above. Upscales may nudge pixels; the ratio intent still holds. If a job needs anamorphic feel, use --ar 239:100 and accept that composition is harder than 16:9.

Where --ar sits with raw, sref, and the Editor

Append parameters in a stable order so you can scan them: subject prose, then --ar, then --style raw when you want photos, then stylize and chaos, then --v 7, then references. Example: "ceramic mug on oak table, soft window light, catalog photo --ar 1:1 --style raw --stylize 120 --v 7". Style and character references (--sref, --cref, --oref) do not replace aspect ratio; they lock look and identity inside the canvas you chose. A vertical cref still needs --ar 9:16 if the delivery is a Story. After a generate, pan, zoom out, and the Midjourney Editor can reframe; that costs extra steps. Prefer the correct --ar on the first grid when you already know the platform. Draft Mode helps you test ratio and composition cheap before a full render. Soft assist when a still already shows the crop you want: upload to PromptMake /image, target Midjourney, then confirm the draft ends with the right --ar.

Common ratios in practice: 1:1 through ultrawide

Memorize a short menu instead of inventing a new fraction each brief. Platform docs change crop previews, yet the geometry stays stable: square for avatars and many feed tiles, landscape for video and headers, tall for mobile full-screen, soft portrait for photo apps, print-ish 3:2 and 2:3 for editorial. Write the ratio into the brief as a labeled fact next to subject and light. Midjourney gets --ar W:H. FLUX APIs often take an aspect_ratio string or width and height. SDXL wants explicit pixels near its training sweet spots. ChatGPT image wants a size preset or a sentence that names orientation. Use the pairs below as a shared language across tools so a 9:16 Midjourney winner becomes a 768×1344 SDXL run without mental math each time. The subsections split social and photo ratios from cinema and print habits.

1:1, 4:5, 9:16 for social and mobile

1:1 (square). Midjourney: --ar 1:1. FLUX: aspect_ratio="1:1" or 1024×1024. SDXL: 1024×1024. ChatGPT image: square / 1024×1024 class preset when offered. Use for profile tiles, album art, many product packs, icon-like subjects. Center the hero; leave even margin.

4:5 (portrait feed). Midjourney: --ar 4:5. FLUX: 4:5 or about 1024×1280. SDXL: often 896×1152 or 1024×1280 class sizes. ChatGPT image: ask for a tall portrait feed crop. Use for Instagram portrait posts where 4:5 fills more phone screen than 1:1.

9:16 (Stories, Reels, TikTok, Shorts). Midjourney: --ar 9:16. FLUX: 9:16 or about 816×1456. SDXL: 768×1344. ChatGPT image: vertical / 9:16 language in the ask. Put faces in the upper third, keep text-safe margins if you overlay UI later, and avoid tiny subjects lost in empty sky.

3:4 and 2:3. Midjourney: --ar 3:4 or --ar 2:3. FLUX and SDXL: portrait widths under heights near those ratios (for example SDXL 896×1152 for 3:4-ish). Use for editorial portrait and Pinterest-leaning boards when 9:16 feels too phone-native.

16:9, 3:2, 21:9 for web, video, and print habits

16:9 (HD landscape). Midjourney: --ar 16:9. FLUX: 16:9 or about 1456×816. SDXL: 1344×768. ChatGPT image: widescreen / 16:9. Use for YouTube thumbs and end cards, blog heroes, slide backgrounds, landscape ads. Leave a calm side for title text when the final asset needs type.

3:2 and 4:3. Midjourney: --ar 3:2 or --ar 4:3. FLUX/SDXL: landscape sizes near 1536×1024 (3:2 class) or 1152×896 (4:3 class). Use for classic photo storytelling and older display decks. 3:2 often feels more "camera" than 16:9 for lifestyle stills.

21:9 and anamorphic. Midjourney: --ar 21:9 or --ar 239:100. FLUX: ultrawide width/height or 21:9 when the API lists it. SDXL: tall-wide grids such as 1536×640 class, with more risk of soft detail on the long axis. Use for cinematic establishing shots and site mastheads. Expect odd anatomy on extreme flags; regenerate with a simpler subject if limbs stretch.

Quick paste cheat (Midjourney first, then translate): product pack --ar 1:1; Instagram portrait --ar 4:5; Reel --ar 9:16; YouTube / blog --ar 16:9; editorial photo --ar 3:2; cinema plate --ar 21:9.

FLUX, SDXL, and ChatGPT image: same intent, different dials

Porting a Midjourney winner means translating the ratio, not pasting --ar into tools that ignore it. FLUX.1.x and Flux 2 family builds (name the variant your host shows) take natural-language prompts plus an aspect or width/height control in the UI or API. Many hosts expose presets labeled 1:1, 16:9, 9:16. SDXL (and local UIs like ComfyUI or Automatic1111-class forks) want width and height near training-friendly multiples; odd custom sizes can soften faces. DALL·E / GPT Image inside ChatGPT lean on conversational size choices and follow-up edits ("make it vertical for Stories"). Ideogram and Leonardo also expose ratio or size pickers; treat them like FLUX: set the canvas in the UI, keep the prose clean. PromptMake /image formats paste targets for Midjourney, FLUX, DALL·E, Stable Diffusion, and Leonardo so you can regenerate ratio-aware text from a reference still without rewriting the whole brief.

Translation habit: keep a one-line map in your project notes. Midjourney --ar 9:16 → FLUX preset 9:16 → SDXL 768×1344 → ChatGPT "vertical 9:16 Story frame". Midjourney --ar 16:9 → FLUX 16:9 → SDXL 1344×768 → ChatGPT widescreen. Midjourney --ar 1:1 → 1024×1024 everywhere that offers it. Hosts with only a few presets need the closest box; compose for that box instead of forcing a custom size that softens the render.

Language models can draft the prose while you lock the canvas. As of mid-2026, GPT-5.6 Sol (API gpt-5.6), Terra, and Luna handle hard brief writing; ChatGPT often defaults to GPT-5.5 Instant for fast chat. Claude Fable 5 and Claude Opus 5 tidy parameter lists. Gemini 3.5 Flash is quick for ratio tables; Gemini 3.1 Pro helps when you paste a long brand size bible. Ask for "one prompt only" plus a separate line that states the target --ar or pixel size so the generator paste stays clean.

Step-by-step: pick the ratio before you generate

A fixed order beats lucky crops. Decide the platform slot, write the ratio into the brief, format for the model, generate once, then change one layer (light, lens, or subject pose) without touching the ratio unless the platform changed. Two to five rounds with a locked canvas beat ten rounds that hop from 1:1 to 9:16 and destroy learning. If you already hold a photo or a prior AI still with the correct crop, run image-to-prompt so the first draft inherits framing cues, then enforce the matching --ar or pixel size on paste. Guests get about 3 image runs per day on PromptMake; a free account raises that to about 5. Soft start: https://promptmake.net/image. The steps below cover blank briefs and photo-first jobs.

From a blank brief to first paste

1. Write the delivery slot in plain words: "Instagram Reel cover," "YouTube thumb," "Shopify square," "blog hero."
2. Map the slot to a ratio: Reel → 9:16, thumb → 16:9, Shopify → 1:1, many IG feed portraits → 4:5.
3. Write subject, setting, light, and medium in one short stack. Keep the ratio out of the adjective pile; put it in the parameter or size field.
4. Format: Midjourney ends with --ar W:H (add --style raw for photos). FLUX gets prose plus the UI/API ratio. SDXL gets width×height. ChatGPT image gets a size preset or an orientation sentence.
5. Generate once. Check subject placement for that platform (headroom on 9:16, side room on 16:9). Change light or pose next. Leave --ar alone unless the deliverable changed.

Save the winning prompt beside the PNG with the ratio in the filename (hero_16x9_v3.png) so the next teammate does not reopen a square and wonder why the crop feels wrong.

From a reference photo (Recreate, Variation, Change Style)

Upload a still that already shows the crop you need: a vertical phone frame for Stories, a landscape plate for a header. On PromptMake /image, pick Midjourney, FLUX, DALL·E, Stable Diffusion, or Leonardo as the paste target. Recreate Exactly when you want near-match language and will append the matching --ar. Create Variation when you need sibling moods inside the same box. Change Style when the crop stays but the look moves. Adjust Lighting when framing is fine and the key is wrong. After copy-out, verify Midjourney lines end with the correct --ar; verify FLUX/SDXL hosts show the matching preset or pixels. Do not invent a new ratio mid-series if the campaign board already locked 4:5. Split style and lighting into separate runs when both need work; one overloaded rewrite muddies the draft.

Common mistakes with aspect ratio Midjourney and ports

Generating square, cropping later. You lose composition. Set --ar or pixels on round one.

Decimals in --ar. --ar 1.78:1 fails. Use --ar 16:9 or --ar 178:100.

Pasting --ar into SDXL or a FLUX field that ignores flags. Translate to width/height or the host preset.

Fighting a pinned Midjourney default. Check Settings if every prompt lands in 16:9 when you forgot to type --ar.

Extreme ratios for client faces. 21:9 and ultra-tall tests stretch anatomy. Stay in 1:1, 4:5, 9:16, 16:9, 3:2 for delivery.

Text overlays planned after a tight crop. Leave title-safe margin in 16:9 and 9:16 when type lands in post.

One-and-done with the wrong box. Plan two to five rounds inside one locked ratio before you open a second canvas.

When PromptMake /image helps ratio-aware prompts

Blank-page prompts work when you already know the slot and the stack. Photo-first workflows win when a still defines the crop: a vertical selfie for Reels, a landscape plate for a masthead, a square pack shot for catalog. PromptMake /image runs vision with model-specific formatting and goal modes: Recreate Exactly, Change Style, Adjust Lighting, Create Variation. Same upload, Midjourney dialect one run, FLUX the next. Soft path: https://promptmake.net/image. After the draft lands, you still own the canvas dial: append --ar, pick the FLUX preset, or set SDXL pixels before you spend generator credits.

Use chat when you explore GPT Image wording and need a fast orientation tweak in one thread. Use /image when a still exists and you need calibrated recreate or variation text aimed at Midjourney, FLUX, DALL·E, Stable Diffusion, or Leonardo. Keep this article on aspect ratio Midjourney syntax and cross-platform size maps so your first grid matches the platform that ships.

FAQ

What is aspect ratio Midjourney syntax?

Aspect ratio Midjourney syntax is the --ar or --aspect parameter plus two whole numbers, such as --ar 16:9. You place it at the end of the prompt with your other parameters. The default without a flag is 1:1. Midjourney rejects decimals, so convert 2.39:1 to --ar 239:100. Common delivery pairs are 1:1, 4:5, 9:16, 16:9, and 3:2.

Which --ar should I use for Instagram, TikTok, and YouTube?

For many Instagram portrait feed posts, start with --ar 4:5. For Stories, Reels, TikTok, and Shorts, use --ar 9:16. For YouTube thumbnails and landscape site heroes, use --ar 16:9. Squares (--ar 1:1) still work for profile tiles and many product packs. Match the slot before you write style words so composition fits the app chrome.

How do I set the same ratio in FLUX and SDXL?

On FLUX hosts, pick the matching aspect preset (1:1, 16:9, 9:16) or pass width and height in the API. On SDXL, set pixels near training-friendly sizes such as 1024×1024, 1344×768, or 768×1344. Do not paste Midjourney --ar into UIs that ignore it. Keep a one-line translation in your project notes so ports stay consistent.

Does ChatGPT image support Midjourney --ar flags?

ChatGPT image tools do not read Midjourney --ar as a Discord-style flag inside ordinary chat. Choose a size preset when the UI offers one, or ask in prose for a vertical Story frame or a widescreen hero. If you need Midjourney parameters, export or rewrite the prompt for Midjourney and append --ar there. PromptMake /image can format a photo toward Midjourney or DALL·E targets in separate runs.

Can I change aspect ratio after I generate in Midjourney?

Yes. Pan, zoom out, and the Editor on midjourney.com can reframe a finished image. Those tools cost extra steps and can alter composition. Prefer the correct --ar on the first generate when you already know the platform. Use post tools for small adjustments, not as your default sizing method.

Why does my ultrawide or ultra-tall Midjourney image look wrong?

Extreme ratios push Midjourney outside the shapes it handles with the most stable composition. Faces stretch, products repeat, and empty bands grow. Move back to 16:9, 9:16, or 21:9 only when the brief needs cinema width, and simplify the subject. Client delivery work belongs in the common social and photo ratios listed above.

How do I start for free today?

Pick one delivery slot, map it to a ratio, and paste a Midjourney prompt that ends with the matching --ar (for example a product on --ar 1:1 or a Reel on --ar 9:16). If you hold a reference photo with the right crop, open https://promptmake.net/image, choose your model, run Recreate Exactly or Create Variation on the free daily quota (about 3 guest runs, about 5 registered), then confirm the ratio on paste before you generate in Midjourney, FLUX, or SDXL. Keep the winning prompt beside the PNG so the next round stays in the same box.