Midjourney V8.2 Image Generator

Midjourney V8.2 has been the bolder default since July 2026. We re-ran our V8.1 test prompts on it here — four images per run, no Discord.

0 / 2,000

~60 cr · 4 imgs

What is Midjourney V8.2?

Midjourney V8.2 is the model Midjourney, Inc. made its default on 24 July 2026, promising bolder images and far fewer random low-quality results. This workspace runs it through a third-party API gateway, with no affiliation to Midjourney: four images per run, no Discord, no monthly plan. Your older V8.1 results stay usable — variations of them still run on V8.1.

Our tests found more drama and less obedience. V8.2 added a lighthouse nobody asked for, staged a menu board as a whole café, and still misspelled the menu. Runs took 60–84 seconds; a V8.1 control started at the same moment took 52. Character and omni references are gone. When the words must be exact, use GPT Image 2.

Release date and headline claims are from the official update note: Midjourney — Version 8.2

What reviewers say about V8.2

“Realism sits a clear step up from before, particularly in how the light and mood behave together.”

— Geeky Animals — Midjourney V8.2 is a quiet win for photographic images

“If it thinks adding an element will improve the composition, it adds it.”

— APIYI — Midjourney 8.2 vs GPT-image-2 comparative review

“Once you move to longer sentences, it starts producing scrambled letters, spelling mistakes, and merged strokes.”

— APIYI — Midjourney 8.2 vs GPT-image-2 comparative review

Every V8.2 image below was generated here on 2 October 2026 — ten runs, all first takes, prompts shown in full. Each V8.1 image is from our 19 August run of the identical prompt and ratio.

Key features of Midjourney V8.2

Same prompt, V8.1 and V8.2 side by side: Our August V8.1 prompt, re-run word for word: V8.2 paints heavier, darker and more dramatic.
Prints and portraits: the look holds, the brief bends: Woodblock prints and editorial portraits keep their finish; the fox leaves the bridge in two of four takes.
In-image text: closer, still not usable: The menu-board test V8.1 failed, re-run: V8.2 gets five of six prices right and still misses.
--q 1 to --q 4 at the same price: V8.2 accepts every --q level from 1 to 4; with the seed fixed, --q 4 resolves finer engraving at no extra cost.
Style reference: borrow a look, not a picture: One --sref image carried a woodblock palette and paper margin onto a new subject in all four takes.

Same prompt, V8.1 and V8.2 side by side

We kept the prompt and the 3:2 ratio identical and changed only the version. V8.1 gave a stylised, brushy cottage; V8.2 rendered heavier surf, a deeper grade and finer detail — the "bolder" the official note promises, and a step away from "thick impasto". One V8.2 take also put a working lighthouse beside the cottage. The prompt asked for a keeper's cottage, not a lighthouse.

Midjourney V8.1 take: a cliff cottage in cold blue light with thick brushwork waves — V8.1 · 19 August

Midjourney V8.1 take: the cliff cottage under a violet sunset — V8.1 · 19 August

Midjourney V8.2 take: the cottage above towering white surf under storm clouds — V8.2 · 2 October

Midjourney V8.2 take that added a lit lighthouse tower beside the cottage — V8.2 · added a lighthouse

Two of the four V8.1 takes from 19 August beside two of the four V8.2 takes from one 62-second run on 2 October. Same prompt, same ratio, no reroll on either side.

Prints and portraits: the look holds, the brief bends

On the ukiyo-e brief, V8.2 kept the indigo-and-vermilion palette and added a printed paper margin and a red seal nobody asked for — the seal's characters are invented. The prompt said the fox crosses the bridge; in two of the four V8.2 takes it walks through the snow beside it instead, as one of our kept V8.1 takes also did. The fisherman is where V8.2 reads best: four different, believable faces, and hands that hold up at full size.

Midjourney V8.1 ukiyo-e print of a red fox on an arched bridge under paper lanterns — V8.1 · on the bridge

Midjourney V8.1 print: the fox walking through snow below the bridge — V8.1 · beside the bridge

Midjourney V8.2 print: the fox crossing the snowy bridge beside a street lantern — V8.2 · on the bridge

Midjourney V8.2 print: the fox walking through snow below the bridge, with an invented red seal — V8.2 · beside the bridge

Midjourney V8.1 portrait of an elderly fisherman mending a net — V8.1

Midjourney V8.1 portrait of the fisherman in a hood and glasses — V8.1

Midjourney V8.2 portrait of the fisherman in a hood, hands working the net — V8.2

Midjourney V8.2 portrait of the fisherman in a knit cap with rigging behind him — V8.2

V8.1 images are from 19 August; each pair of V8.2 images comes from one 71-second run on 2 October.

In-image text: closer, still not usable

This is the test V8.1 failed in all four takes. V8.2's best take spells the heading and five of the six items with the right prices, then drops "Filter 3.20" and adds a garbled copy of the croissant line. Another take invented a "Toul B Coiggant". Every take also ignored "photographed straight on" and built a whole café around the board. Closer than V8.1, but not something you could print — for exact wording, use GPT Image 2.

Midjourney V8.1 best take of the chalkboard menu, with one scribbled line and a dropped word — V8.1 · best of 4

Midjourney V8.2 best take: five of six menu items correct, Filter missing, croissant line duplicated — V8.2 · best of 4

Midjourney V8.2 take with invented menu items and wrong prices on the food board — V8.2 · invented items

Midjourney V8.2 take with two BREW boards and the croissant listed twice at different prices — V8.2 · duplicated lines

Same prompt and 2:3 ratio on both versions, first takes only — the V8.1 image is its best of four, the three V8.2 images come from one 71-second run.

--q 1 to --q 4 at the same price

V8.1 only took --q 1 or 4; V8.2 accepts any level in between. We fixed the seed and changed only --q. The composition stays identical, but at --q 4 the astrolabe's plate resolves into engraved rings and star lines where --q 1 leaves an even, noisy texture. Both runs took about a minute and cost the same credits here, so when fine detail matters there is no reason not to add --q 4.

Midjourney V8.2 astrolabe at --q 1: the plate engraving reads as an even texture — --q 1

Midjourney V8.2 astrolabe at --q 4: the same composition with engraved rings and star lines resolved — --q 4

Midjourney V8.2 second astrolabe take at --q 1, gears in front of a flat engraved plate — --q 1

Midjourney V8.2 second astrolabe take at --q 4 with finer engraving on the plate — --q 4

Same prompt, same --seed 4417, same 3:2 ratio — only --q changed. Each --q 1 image sits next to its --q 4 counterpart.

Style reference: borrow a look, not a picture

We pointed --sref at our V8.1 fox print and asked for something unrelated: a grey heron fishing in a lotus pond at dawn. All four takes came back as woodblock prints in the same vermilion-and-indigo palette with the paper margin, three of them with a seal stamp, and none copied the fox, the bridge or the lanterns. The seal glyphs are invented, not real characters. Native flags like --sref pass through in your prompt untouched.

The style reference: a V8.1 ukiyo-e print of a fox on a lantern-lit bridge — Style source

Midjourney V8.2 heron among red lotus flowers under a vermilion sky, in the reference print style — V8.2 · --sref

Midjourney V8.2 heron standing in a pond of lotus blossoms, indigo and vermilion woodblock style — V8.2 · --sref

Midjourney V8.2 heron wading among lotus flowers at sunrise, in the reference print palette — V8.2 · --sref

All three herons are takes from one 85-second run. The prompt text named only the heron, the pond and the mist — the palette, border and print texture all came from the reference.

More Midjourney V8.2 outputs from this workspace

The rest of the 2 October session, first takes: an HD product shot at 2688×1792 and four of the 24 sketches from one Draft run.

Midjourney V8.2 HD product photograph of a speckled ceramic pour-over coffee set on travertine — PromptHD · 2688×1792 · 141-second run
Studio product photograph of a hand-thrown ceramic pour-over coffee set in speckled oatmeal glaze on a travertine slab, morning side light, soft shadow, generous negative space, editorial catalogue style

Midjourney V8.2 HD take of the ceramic coffee set with a jug and a linen cloth — PromptHD · same run
Studio product photograph of a hand-thrown ceramic pour-over coffee set in speckled oatmeal glaze on a travertine slab, morning side light, soft shadow, generous negative space, editorial catalogue style

Midjourney V8.2 Draft sketch of a floating island night market lit by lanterns — PromptDraft · 512px · one of 24 from a 21-second run
Isometric concept art of a floating island night market, paper lanterns, tiny stalls and bridges, warm amber and deep teal

Midjourney V8.2 Draft sketch of the floating market with sky lanterns rising — PromptDraft · same run
Isometric concept art of a floating island night market, paper lanterns, tiny stalls and bridges, warm amber and deep teal

Midjourney V8.2 Draft sketch of the island market with strings of lights — PromptDraft · same run
Isometric concept art of a floating island night market, paper lanterns, tiny stalls and bridges, warm amber and deep teal

Midjourney V8.2 Draft sketch of the floating market framed by red autumn trees — PromptDraft · same run
Isometric concept art of a floating island night market, paper lanterns, tiny stalls and bridges, warm amber and deep teal

Midjourney V8.2 vs V8.1 vs GPT Image 2

V8.1 is the previous default and GPT Image 2 the model to use when words must be exact — what each one does in this workspace.

	Midjourney V8.2	Midjourney V8.1	GPT Image 2
New generations here	Yes — the current default	No — only variations and remixes of existing V8.1 results	Yes
Credits per run	60 flat for four images (90 in HD)	60 flat for four images (90 in HD)	5–560 per image by resolution × quality, plus 10 per reference image
Measured time, Fast run	60–84 s (2 October 2026)	13–23 s (19 August 2026); 52 s in a same-moment control on 2 October	32–166 s per image at 2K, by quality (its page)
Menu-board text test	Best of four: heading and five of six prices right, one item dropped	Best of four: one line scribbled, one word dropped	The most reliable of the three (tested on its page)
--q detail levels	1, 2, 3 or 4	1 or 4	Not applicable — low / medium / high quality tiers
Reference images	Up to 5, plus --sref style references	Up to 5	Up to 16

Credits and reference limits are read from this workspace's model registry on 2 October 2026. Times are our own runs through the same API gateway on the dates shown; gateway load moves them, which is why we ran a V8.1 control at the same moment as V8.2. GPT Image 2 figures come from the tests on its own page.

More Image Models Like Midjourney V8.2

Compare Midjourney V8.2 with other image models — same workspace, one click to switch.

Midjourney V8.1

Four artistic variations per prompt, no Discord.

GPT Image 2

Precise prompts, clean in-image text, product shots.

Seedream 5 Pro

Flagship fidelity — ultra-detailed 1K/2K rendering.

Free

Z Image Turbo

Fast free drafts for everyday ideas.

Midjourney V8.2 specs on CreateVision AI

Model	mj-v8.2 through a third-party API gateway — Midjourney's default model since 24 July 2026 (no affiliation)
Modes	Text-to-image and image-to-image — one reference image plus a prompt, or two or more images
Post-generation edits	Variation, remix, background removal, retexture — variations and remixes of V8.1 results keep running on V8.1
Speed modes	Fast (4 images) · Draft (24 sketches at 512px, Standard quality only)
Quality	Standard 1344×896 at 3:2 · HD 2688×1792 — HD requires Fast mode
Aspect ratios	7 presets — 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16 — sent as a native --ar flag
Prompt	Native flags pass through (--q 1–4, --s, --sref, --seed and more); --ar is set for you. About 2,000 characters — longer prompts are trimmed with the flags kept
Not available	--cref and --oref (removed in V8.2); personalization profiles and moodboards (they need an official Midjourney account)
Reference images	Up to 5 per generation
Credits	60 per run (Fast/Standard or Draft) · 90 (Fast/HD) — one price covers every image in the run
Measured speed	Fast 60–84 s, HD 112–141 s, Draft 21 s for 24 sketches — 11 first-take runs on 2 October 2026

How to use Midjourney V8.2 here

Start Generating Images

Sign in — the generator is already set to V8.2
No Discord, no Midjourney subscription, no API key. The generator at the top of this page runs Midjourney V8.2.
Write the prompt, flags included
One confident line is often enough. Add native flags such as --q 4, --s or --sref with a public image URL; the ratio picker sets --ar for you.
Choose Draft, Fast or HD
Draft for 24 quick sketches, Fast for four finished images, HD for print-size pixels. The credit cost shows before you run, and results stay in your history.

Midjourney V8.2 Pricing

Try Midjourney V8.2 with your sign-up credits, or subscribe for a monthly credit pool that covers every model in this workspace.

Free

Perfect for getting started

200 welcome credits, up to 80/day
Z Image Turbo drafts at 0 credits
Standard generation speed
Generation history included

Get Started

Premium

Ultimate

$59/month

Billed annually · $468/year

18,000 credits/month, no daily limits
HD generation up to 4K resolution
Priority support
Fastest queue and early access to new models

Upgrade to Ultimate

Midjourney V8.2 — Frequently Asked Questions

What is new in Midjourney V8.2 compared with V8.1?

Midjourney says V8.2 makes images more creative, bold and edgy, sharply reduces random low-quality results, and improves personalization. In our same-prompt tests the grade got deeper and the drama heavier, --q now accepts every level from 1 to 4, and --cref and --oref were removed. It also took more liberties with the brief — adding a lighthouse and staging a menu board as a café.

Is Midjourney V8.2 slower than V8.1?

In our runs, yes — partly. On 2 October 2026 our V8.2 Fast runs took 60–84 seconds. A V8.1 run started at the same moment took 52 seconds against V8.2's 82, so V8.2 was slower, but gateway load also slowed both: V8.1 had taken 13–23 seconds in August. Draft is still quick — 24 sketches in 21 seconds.

Can Midjourney V8.2 render text inside images?

Better than V8.1, still not reliably. On our chalkboard-menu test the best of four takes got the heading and five of six prices right, then dropped one item and repeated another in garbled letters; a second take invented items. If the image must contain exact words, use GPT Image 2.

Can I still use --cref or --oref?

No — V8.2 does not support character or omni references, and Midjourney's new edit model is not part of this integration. To steer from your own pictures, attach up to 5 reference images, or add --sref with a public image URL to borrow a style.

What happens to images I made with Midjourney V8.1?

They stay in your history. A variation or remix of a V8.1 result runs on V8.1 automatically — the V8.2 endpoints refuse V8.1 sources — while background removal and retexture run on V8.2. Every new generation uses V8.2.

Do I need Discord or a Midjourney subscription — and is this affiliated with Midjourney?

No and no. This workspace runs Midjourney V8.2 through a third-party API gateway and charges per run in credits. CreateVision AI is independent and is not affiliated with, endorsed by, or sponsored by Midjourney, Inc.; Midjourney is a trademark of Midjourney, Inc. Account-only features such as personalization profiles and moodboards are not available here.

Can I use Midjourney V8.2 images commercially?

Images you generate here can be used in personal and commercial projects under our terms of service. You remain responsible for prompt content, trademarks, likeness rights, and the rights to any reference images you attach.

Run your own V8.1 prompts on V8.2 and compare

Start Generating Images

Discover our one-click AI image tools

One-click tools for everything around generation — polish, restyle, and share.

AI Background Remover AI Background Changer AI Image Upscaler AI Old Photo Restoration AI Object Remover AI Watermark Remover AI Headshot Generator AI Anime Avatar AI Cartoon Portrait AI Hair Color Changer AI Hairstyle Changer AI Bald Filter AI Gender Swap GIF Face Swap AI Aging Filter AI Face Swap AI Photo to Anime AI Photo to Ghibli AI Photo to Sketch AI Art Style Transfer AI Time Travel Photo AI Couple Photo AI Family Photo Generator Past & Present Photo AI Polaroid Generator AI Coloring Page Generator AI Action Figure Generator AI Doll Generator AI Photo Booth AI Celebrity Photo AI Pet Portrait AI Halloween Photo AI Storyboard Generator AI Wedding Photo AI Christmas Cards AI Christmas Photo AI Birthday Card AI Diwali Card Maker AI Valentine Card Maker AI Mother’s Day Card Maker AI Father’s Day Card Maker AI Group Photo Mixer FIFA Photo Maker AI ID Photo Maker AI Clothes Changer AI Virtual Try-On AI Logo Generator AI Album Cover Generator AI Rap Generator Rap Lyrics Generator AI Lyrics Generator Diss Track Generator AI Logo Remover AI Text Remover AI Image Translator AI Room Designer Virtual Staging AI AI Product Mockup View all tools →

Pricing & Access — The Facts — Midjourney V8.2

Model: Midjourney V8.2
Developer: Midjourney
Access: Runs in the CreateVision AI workspace via API — sign in and generate. No API key, no cloud project, no software install.
Price per generation: From 60 credits ≈ $0.22 per image on the Premium plan ($29/month for 8,000 credits). The free plan includes 80 credits every day.
Privacy: Private by default — no other user can see your generations, on any plan. We don't sell or misuse member data. Privacy Policy

Midjourney V8.2 Image Generator

What is Midjourney V8.2?

What reviewers say about V8.2

Key features of Midjourney V8.2

Same prompt, V8.1 and V8.2 side by side

Prints and portraits: the look holds, the brief bends

In-image text: closer, still not usable

--q 1 to --q 4 at the same price

Style reference: borrow a look, not a picture

More Midjourney V8.2 outputs from this workspace

Midjourney V8.2 vs V8.1 vs GPT Image 2

More Image Models Like Midjourney V8.2

Midjourney V8.1

GPT Image 2

Seedream 5 Pro

Z Image Turbo

Midjourney V8.2 specs on CreateVision AI

How to use Midjourney V8.2 here

Sign in — the generator is already set to V8.2

Write the prompt, flags included

Choose Draft, Fast or HD

Related reading

Best AI Image Generators 2026

GPT Image 2: Complete Guide

Seedream 5: Complete Guide

How Everyone Can Use AI to Generate Images in 2026

Midjourney V8.2 Pricing

Free

Premium

Ultimate

Midjourney V8.2 — Frequently Asked Questions

What is new in Midjourney V8.2 compared with V8.1?

Is Midjourney V8.2 slower than V8.1?

Can Midjourney V8.2 render text inside images?

Can I still use --cref or --oref?

What happens to images I made with Midjourney V8.1?

Do I need Discord or a Midjourney subscription — and is this affiliated with Midjourney?

Can I use Midjourney V8.2 images commercially?

Run your own V8.1 prompts on V8.2 and compare

Discover our one-click AI image tools

Pricing & Access — The Facts — Midjourney V8.2