How to Use Midjourney V8.2: 12 Steps, 90 Min [2026]

Midjourney flipped its default model twice in four months this year, and the latest switch is the one worth stopping for. On July 24, 2026, Midjourney V8.2 replaced V8.1 as the model every new prompt runs on, bringing native 2K output, a rebuilt personalization engine, and prompt tools that finally make sense for people who don’t want to memorize a parameter cheat sheet. If you learned Midjourney on V7 or even the V8.0/V8.1 alpha builds from earlier this year, enough has changed that your old habits will produce worse results than they should. This tutorial walks through account setup, plan selection, prompt structure, the new Omni-Reference and personalization tools, native 2K upscaling, and a full branded-image project you can build in about 90 minutes. Every step below reflects how V8.2 behaves as of August 20, 2026, not the V7 workflow still floating around in older tutorials.

·

Google · Preferred Sources

Don't miss new tech stories on Google

Add Tech Insider once in the Google app and our stories appear in your news suggestions.

Add Now

What Changed in Midjourney V8.2, and Why It Matters Right Now

Midjourney’s V8 line has moved fast in 2026. The V8 Alpha launched March 17, followed by V8.1 Alpha on April 14. V8.1 reached the main web app and Discord on April 30 and became the default model on June 10. Then, on July 24, V8.2 took over as the new default, retiring the V8.0 Alpha branch entirely. That’s three default-model changes inside five months, which is why so many prompts that worked in June now render differently in August.

Midjourney describes V8.2 as an “aesthetics-and-personalization” release rather than a raw capability jump. In practice that means three things change for you as a user. First, default outputs lean bolder and more stylized than V8.1’s flatter, V7-inspired look. Second, the personalization system reads your taste signals more accurately, especially if you’ve already built a profile with a large image pool. Third, the model carries forward the native 2K resolution and roughly 5x faster generation speed that the V8 series introduced over V7, so image quality and iteration speed both improve at once.

V8.2 also restores image prompting, which had been limited during early V8 testing, and ships a Prompt Shortener plus an updated /describe command that produces longer, more detailed reverse-engineered prompts from reference images. If you paused Midjourney experiments after a rough V8 Alpha session in the spring, this is the build worth coming back for.

Notably, V8.2 didn’t jump straight to default status. Midjourney ran it as a preview build behind a --preview flag starting June 25, 2026, giving subscribers about a month to test the aesthetic shift before it became mandatory for everyone. That preview window is why so much commentary and comparison content about V8.2 already existed before its official default rollout — creators had a full month to stress-test it against V8.1 side by side.

Prerequisites: What You Need Before You Start

You don’t need design experience to follow this tutorial, but a few things need to be in place first:

• A Discord account (free) or a browser — Midjourney V8.2 runs on both the web app at midjourney.com and inside Discord, so pick whichever interface you prefer.
• A paid Midjourney subscription. There is no free trial; the minimum spend is the $10/month Basic plan, and V8.2 is included at no extra cost across every tier.
• A credit or debit card, or PayPal, for billing through the Midjourney website.
• Roughly 90 minutes for this full walkthrough, including the branded image-set project in Step 12.
• Optional: 5-10 reference images of your own work or brand style if you want to build a personalization profile in Step 7.
• Optional: basic command-line comfort if you plan to try the batch-automation workflow in Step 11.

That’s it. Midjourney V8.2 doesn’t require a GPU, a local install, or any software beyond a browser or Discord client, which is one of the reasons it remains the default entry point for people learning AI image generation in 2026.

Step 1: Create and Verify Your Midjourney Account

Go to midjourney.com and sign up with an email address or Google account. Midjourney requires phone or ID verification on new accounts to cut down on abuse, so have a phone number ready. Once verified, you land on the web app’s home feed, which now serves as the primary interface for most users — Discord is still fully supported, but Midjourney has spent 2026 pushing feature parity and, in some cases, web-first releases.

If you prefer Discord, join the official Midjourney server from the same signup flow, or add the Midjourney bot to your own server once you’re subscribed. Commands work identically in both places, though the web app’s prompt bar gives you live parameter suggestions that Discord’s slash-command autocomplete doesn’t.

Step 2: Choose the Right Subscription Plan

Midjourney sells GPU compute time, not a fixed image count, which trips up a lot of new users. Every plan includes a monthly allotment of “Fast” GPU hours for quick generations, plus unlimited slower “Relax” mode generations on Standard tier and above. There’s no separate V8.2 pricing tier or surcharge — the new model runs on the same four plans Midjourney has offered all year.

Plan	Monthly Price	Annual Price (20% off)	Fast GPU Hours	Best For
Basic	$10/mo	$8/mo	3.3 hours	Casual use, testing V8.2
Standard	$30/mo	$24/mo	15 hours	Unlimited Relax mode, regular use
Pro	$60/mo	$48/mo	30 hours	Stealth Mode, private generations
Mega	$120/mo	$96/mo	60 hours	Studios, high-volume video and image work

For this tutorial, Standard is the practical minimum — Basic’s 3.3 Fast hours burn through quickly once you start iterating on prompts, personalization profiles, and upscales. If you plan to keep your work private (client projects, unreleased products), Pro’s Stealth Mode hides your generations from Midjourney’s public community feed.

Step 3: Access V8.2 on the Web App or Discord

Once subscribed, V8.2 is already your default model — you don’t need to flag it manually. On the web app, click the prompt bar at the top of the Create page and type your prompt directly. In Discord, use the /imagine command in any channel where the Midjourney bot is active.

/imagine prompt: a lighthouse on a rocky coastline at dawn, cinematic light, --ar 16:9 --v 8.2 --stylize 250

Notice the explicit --v 8.2 flag above. You don’t need it since V8.2 is default, but including it is good practice when you’re troubleshooting or sharing prompts with collaborators who might be on an older personalization profile or pinned version.

Step 4: Learn the New Interface and Draft Mode

V8.2 carries forward Draft Mode from earlier V8 builds, which generates lower-cost, faster preview grids so you can rough out composition and lighting ideas before committing full Fast GPU hours to a final render. Draft Mode outputs are visibly softer than a full-resolution V8.2 render, but they’re accurate enough to judge composition, framing, and color palette.

The workflow most experienced users settle into: draft five to ten quick variations, pick the one or two with the strongest composition, then run those specific prompts at full V8.2 quality. This alone can cut your Fast GPU hour usage by half compared with generating everything at full resolution from the start.

Also new in the V8.2-era interface: Conversational Mode, which lets you refine an image through natural-language follow-up instructions (“make the sky more orange,” “move the subject left”) instead of rewriting the entire prompt string each time. It’s slower per-iteration than raw prompt editing but far more approachable if you’re new to Midjourney’s parameter syntax.

Step 5: Write and Structure Your First V8.2 Prompt

Midjourney prompts follow a rough hierarchy: subject first, then setting and action, then style and medium, then technical/lighting descriptors, then parameters last. V8.2 rewards more detail than V7 did, partly because the model’s improved prompt adherence means it actually uses the extra description instead of ignoring it.

/imagine prompt: a senior software engineer debugging code at a standing desk, dual monitors showing terminal windows, moody blue office lighting, shallow depth of field, shot on 35mm film, editorial photography style --ar 3:2 --stylize 300 --chaos 10

Two parameters to understand immediately: --stylize (0-1000) controls how much Midjourney’s aesthetic bias overrides literal prompt interpretation, and --chaos (0-100) controls how varied your four grid results are from each other. Low stylize plus low chaos gives you literal, predictable, similar-looking results. High stylize plus high chaos gives you dramatically different, more “designed” interpretations of the same prompt — useful for exploring options, risky when you need a precise, repeatable output.

Step 6: Master Core Parameters (Aspect Ratio, Stylize, Chaos, Weird)

Beyond stylize and chaos, a handful of parameters cover most day-to-day needs. --ar sets aspect ratio (16:9 for hero images, 1:1 for social, 9:16 for vertical/mobile). --weird (0-3000) pushes toward unconventional, surreal compositions — useful for concept art, distracting for product photography. --no lets you exclude elements the model keeps adding uninvited.

/imagine prompt: minimalist product photo of a wireless earbuds case on a marble surface, soft studio lighting, --ar 1:1 --stylize 100 --no text, watermark, hands, reflection

Keep a running note of parameter combinations that work for your specific use case (product shots, character concepts, landscapes) — V8.2’s aesthetic shift means combinations that were reliable on V7 or V8.1 may now produce more stylized results than you want, and dialing stylize down 50-100 points from your old default is a common first fix.

Step 7: Build a Personalization Profile

Personalization is the headline feature of V8.2. Run /tune (or find the Personalization tab in the web app) and Midjourney generates paired image sets, asking you to pick your preference from each pair. The more pairs you rank, the larger your “taste pool,” and V8.2 reads that pool more precisely than earlier versions — Midjourney’s own release notes describe this as the model “reading your taste” across prompts and past choices rather than applying a fixed house style.

Once you have a profile, append --p to any prompt to apply it, or add a specific profile code if you’ve saved multiple personalization sets for different projects (one for photography, one for illustration, for example).

/imagine prompt: a cozy reading nook by a rain-streaked window, warm lamp light --ar 4:5 --p --stylize 200

If you built a personalization profile earlier in 2026 on V7 or V8.1, don’t assume it carries over identically — Midjourney’s guidance for the V8.2 migration is to re-rank at least 20-30 fresh pairs so the profile reflects the new model’s aesthetic range instead of skewing toward outdated preferences.

Step 8: Use Omni-Reference and Style Reference for Consistency

V8.2 restores full image-prompting support, which early V8 alpha builds had limited. Two reference tools matter here: Omni-Reference (--oref), which locks a specific subject’s appearance (a character, a product, a face) across multiple generations, and Style Reference (--sref), which borrows the visual style — color palette, texture, rendering technique — from a reference image without copying its content.

/imagine prompt: the same character now standing in a neon-lit alley at night --oref https://example.com/character-ref.png --oref_weight 350 --sref https://example.com/style-ref.png --ar 2:3

--oref_weight (0-1000) controls how strictly the model sticks to the reference subject. Push it high (600+) for strict character consistency across a series of images — useful for comic panels, brand mascots, or a recurring product across a marketing set. Keep it lower (100-300) when you want the reference as inspiration rather than a strict template.

Step 9: Generate Multi-Subject Scenes and Render Text

Two areas where AI image generators have historically struggled — coherent multi-subject composition and legible in-image text — are meaningfully better in the current generation of models, V8.2 included. For multi-subject scenes, describe each subject’s position and relationship explicitly rather than listing them as a flat list (“a woman on the left pointing toward a man on the right” beats “a woman and a man”).

For text rendering, quote the exact text you want in the image and keep it short — one to four words performs far more reliably than full sentences or paragraphs, even on the newest models.

/imagine prompt: a hand-painted wooden cafe sign hanging outside a brick storefront, the sign reads "OPEN DAILY", warm afternoon light --ar 4:5 --stylize 150

Expect roughly 70-85% legibility on short text strings — still worth a second or third grid if the first attempt garbles a letter, and still worth a manual touch-up pass in an editor for anything client-facing.

Step 10: Upscale, Export, and License Your Native 2K Images

The V8 series’ headline technical upgrade is native 2K output, carried forward into V8.2, versus V7’s lower default resolution. Once you generate a grid, click any individual image to open it, then use the upscale option to render the full-resolution version. Because 2K is now native rather than an upscale-only add-on, quality loss between the preview and the final export is smaller than it was on V7.

Download from the web app’s asset library or right-click and save from Discord. For commercial use, remember that Midjourney’s standard terms grant paid subscribers commercial rights to images they generate, but Pro and Mega plans add Stealth Mode if you need generations kept off the public feed before a product launch — a common requirement for agencies and in-house marketing teams.

Step 11: Automate Prompts With Batch Mode and Third-Party Tooling

Midjourney doesn’t publish an official public REST API, so batch and automated workflows typically run through Discord automation libraries that queue /imagine commands programmatically, or through the web app’s built-in batch tools for logged-in subscribers. If you’re generating dozens of variations from a spreadsheet of prompts — common for e-commerce catalogs or A/B testing ad creative — a simple queue script keeps you from manually re-typing each command.

import csv
import time

# Read prompts from a CSV and queue them one at a time,
# respecting Midjourney's rate limits and your Fast GPU budget.
with open("prompts.csv", newline="") as f:
    reader = csv.DictReader(f)
    for row in reader:
        prompt = row["prompt"]
        params = row.get("params", "--v 8.2 --ar 1:1")
        full_command = f"/imagine prompt: {prompt} {params}"
        print(f"Queuing: {full_command}")
        # send full_command via your Discord automation client here
        time.sleep(20)  # stay well under rate limits

Treat any third-party “Midjourney API” service as an unofficial wrapper around Discord automation, not a first-party integration — read its terms carefully, since Midjourney’s own terms of service restrict certain forms of automated or bulk access outside its supported tooling.

Step 12: Build a Complete Project — A Branded Product Image Set

Put everything together with a real deliverable: a five-image branded product set for a hypothetical direct-to-consumer coffee brand. This is the kind of task V8.2’s personalization and Omni-Reference tools were built for.

• Build the profile: Rank 25-30 personalization pairs favoring warm, editorial, natural-light photography to match the brand’s aesthetic.
• Establish the hero shot: Generate the primary product photo (bag of coffee on a wooden counter) at Draft Mode first, refine composition, then render full V8.2 quality with --p --stylize 200 --ar 4:5.
• Lock consistency: Use --oref pointing at the approved hero shot to generate four supporting images (lifestyle shot, ingredient close-up, packaging flat-lay, in-use scene) that all match the same product appearance.
• Add the text variant: Generate one social-ready image with a short brand tagline rendered directly in-scene, then verify legibility and touch up if needed.
• Upscale and export: Render all five images at native 2K, export as PNG, and organize into a single asset folder labeled by use case (hero, social, packaging, web).

Budget roughly 1.5-2 Fast GPU hours for this project on Standard tier, including drafts and re-rolls, well within a single month’s Standard allotment.

Turning a V8.2 Image Into Video

V8.2 itself is an image model, not a video model, but Midjourney’s platform lets you take any still image you generate and extend it into motion. Open a finished image, select the image-to-video option, and Midjourney produces a 5-second clip animating elements of the scene — camera drift, subtle environmental motion, or subject movement, depending on the source image. From there you can extend the clip in additional increments up to roughly 21 seconds total.

This matters for the branded-image project in Step 12: once your five hero images are locked, the same assets can become short looping social clips without generating anything new from scratch, which is a useful way to stretch a single Fast GPU hour budget across both static and motion deliverables.

Midjourney V8.2 vs V8.1 vs V7: What Actually Changed

Feature	V7	V8.1	V8.2
Default status	Retired	Default June 10 – July 24, 2026	Default since July 24, 2026
Native resolution	Standard	Native 2K	Native 2K
Generation speed	Baseline	~5x faster than V7	~5x faster than V7
Aesthetic style	Photorealistic default	Familiar, V7-inspired look	Bolder, more stylized, edgier default
Image prompting	Supported	Limited	Restored, full support
Personalization	Standard	Improved	Reads taste more precisely, larger image pool
Prompt tools	Standard /describe	Standard /describe	Prompt Shortener, updated /describe

The practical takeaway: if you were happy with V8.1’s more restrained, photorealistic look, expect V8.2 to push your results toward bolder color and composition choices by default. Lowering --stylize by 50-100 points from your old baseline is the fastest way to pull results back toward a V8.1-style output while still benefiting from V8.2’s speed and personalization improvements.

How Midjourney V8.2 Compares to Other 2026 Image Models

Midjourney isn’t operating in a vacuum in 2026. Black Forest Labs’ FLUX.2 line ships as an open-weight alternative with a 32-billion-parameter flagship checkpoint, and the company announced FLUX 3, a new multimodal frontier model, on July 23, 2026. Alibaba’s Qwen-Image 2.0, released February 10, 2026, condensed what previously took two 20-billion-parameter models into a single 7-billion-parameter architecture that handles both generation and editing. Krea released its first in-house foundation model, Krea 2, on June 22, 2026, built for illustration, painterly styles, and photorealism in one model. Luma’s Photon Uni-1.1 targets a similar native-2K tier at roughly $0.0404 per image on its API.

Model	Latest Version	Released	Access Model	Standout Feature
Midjourney	V8.2	July 24, 2026	Subscription ($10-$120/mo)	Personalization, native 2K, Discord/web workflow
Black Forest Labs	FLUX.2 / FLUX 3 (announced)	Nov 2025 / July 23, 2026	Open weights + API	32B open checkpoint, multimodal roadmap
Krea	Krea 2 / Krea 2 Turbo	June 22, 2026	Subscription + API	Illustration and painterly range
Alibaba	Qwen-Image 2.0	Feb 10, 2026	Open weights	7B unified generation + editing
Luma AI	Photon Uni-1.1	Mar-May 2026	Pay-per-image API	~$0.0404/image, native 2K

If your workflow needs an API-first, pay-per-generation model for app integration, Luma Photon or the FLUX family fit better than Midjourney’s Discord/web-first, subscription-based design. If you want a guided creative tool with strong aesthetic defaults and a genuinely useful personalization system, Midjourney V8.2 remains the more approachable option for non-developers. Aggregators like Freepik now bundle 39-plus image models, including several listed here, if you want to compare outputs side by side before committing to a subscription.

Prompt Recipes for Common Use Cases

Rather than starting from a blank prompt bar every time, it helps to keep a small library of tested starting points for common categories of work. The four recipes below reflect parameter combinations that behave predictably on V8.2 as of August 2026 — treat them as a starting draft, not a fixed formula, since your specific subject and personalization profile will still shift the output.

Portrait Photography

Portraits are one of the categories most affected by V8.2’s bolder default aesthetic — left unchecked, stylize values that produced natural skin tones and lighting on V8.1 can push toward an overly polished, almost airbrushed look on V8.2. Dropping stylize into the 100-200 range and keeping chaos low usually keeps portraits closer to a documentary or editorial feel rather than a beauty-retouch look.

/imagine prompt: candid portrait of a mid-30s architect reviewing blueprints in a sunlit studio, natural window light, slight film grain, 50mm lens look --ar 4:5 --stylize 150 --chaos 5

Product Photography

Product shots need consistency and low chaos above almost everything else — you’re rarely looking for four wildly different interpretations of a bottle of shampoo. Keep stylize low, specify the surface and lighting explicitly, and use --no aggressively to strip out props Midjourney tends to add unprompted (hands, extra reflections, invented logos).

/imagine prompt: matte black skincare bottle on a light gray concrete surface, single soft key light from upper left, subtle shadow, e-commerce catalog style --ar 1:1 --stylize 80 --chaos 0 --no props, text, reflection, hands

Architecture and Interior Concepts

Architectural visualization benefits from V8.2’s improved multi-element composition, since interiors and building exteriors typically involve dozens of secondary details (furniture, fixtures, materials) that need to stay coherent together. Mid-range stylize with a moderate aspect ratio tuned to the structure (16:9 for exteriors, 4:5 or 3:2 for interiors) tends to produce the cleanest results.

/imagine prompt: minimalist Scandinavian living room, oak floors, floor-to-ceiling windows, midday natural light, muted sage and cream palette, architectural photography --ar 3:2 --stylize 250

Character Concept Art

Concept art is where V8.2’s higher default stylize actually works in your favor — you generally want bold, designed interpretations rather than literal ones. Push stylize and chaos higher for early exploration, then lock in a favorite direction with Omni-Reference once you’ve picked a design to refine.

/imagine prompt: a weathered bounty hunter in patchwork armor, cybernetic left arm, standing in a dust storm on a desert planet, dramatic rim lighting, concept art --ar 2:3 --stylize 600 --chaos 30 --weird 200

Common Pitfalls When Moving to V8.2

• Assuming your old prompts still calibrate correctly. V8.2’s bolder default aesthetic means a prompt tuned for V7 or V8.1 often over-stylizes now. Start new projects by re-testing your stylize value rather than reusing an old default blindly.
• Skipping the personalization re-tune. A stale personalization profile built on an earlier model can actively pull V8.2 away from the taste signals you originally wanted. Re-rank fresh pairs after a version jump instead of assuming the profile transfers cleanly.
• Burning Fast GPU hours on full-resolution exploration. Skipping Draft Mode and rendering every experimental variation at full quality is the single fastest way to exhaust a Basic or Standard plan’s monthly allotment.
• Overloading a single prompt with conflicting descriptors. Piling on contradictory style cues (“photorealistic” plus “cartoon” plus “oil painting”) confuses even a strong prompt-adherence model like V8.2 and produces muddy, inconsistent results.
• Ignoring Omni-Reference weight tuning. Leaving --oref_weight at a default value for both loose-inspiration and strict-consistency use cases produces inconsistent character or product appearance across a set — tune it per project, not once globally.
• Expecting perfect text rendering on long strings. Even with V8.2’s improvements, full sentences of in-image text remain unreliable. Keep rendered text short and plan on a manual edit pass for anything client-facing.

Troubleshooting Midjourney V8.2

• Images look more stylized/exaggerated than expected. Lower --stylize by 50-100 points from your previous baseline; V8.2’s default aesthetic runs bolder than V8.1.
• Personalization results feel off or generic. Re-run /tune and rank at least 20-30 fresh pairs specifically on V8.2 rather than relying on an older profile.
• --oref isn’t keeping the subject consistent. Increase --oref_weight toward 500-700, and make sure your reference image is high-resolution and shows the subject clearly without heavy occlusion.
• Fast GPU hours running out mid-month. Switch exploratory work to Draft Mode, move finished-but-non-urgent renders to Relax mode (Standard tier and above), and reserve Fast hours for final, client-ready output.
• In-image text is garbled or misspelled. Shorten the requested text to one to four words, increase --stylize moderately (text renders more cleanly on some compositions with lighter stylization), and re-roll rather than editing a garbled result.
• Web app and Discord show different results for the “same” prompt. Check for hidden parameter differences — the web app’s UI sometimes appends default parameters (aspect ratio, stylize) that aren’t visible unless you open the prompt details panel.
• Verification loop blocking account access. Midjourney’s anti-abuse verification occasionally re-triggers on new devices or after long absences; complete the phone/ID check fully rather than retrying repeatedly, which can extend the lockout.
• Batch/automation script gets rate-limited or blocked. Slow down request intervals significantly (20+ seconds between commands), and confirm your automation approach doesn’t violate Midjourney’s terms of service for bulk or bot access.
• Upscaled export looks softer than the preview. Confirm you’re using the native 2K render path rather than an older upscale-only route; some cached or shared links may point to a lower-resolution version.
• Stealth Mode generations still appearing in public search. Confirm your plan is Pro or Mega (Stealth Mode isn’t available on Basic or Standard) and that Stealth is toggled on before generating, not applied retroactively.

Advanced Tips for Power Users

Once the basics click, a few habits separate efficient V8.2 workflows from wasteful ones. Keep a personal prompt library organized by use case (product photography, character design, marketing hero images) with your best-performing parameter combinations documented next to each — this cuts iteration time dramatically on repeat projects.

Combine Omni-Reference with multiple personalization profiles for agency or freelance work: maintain a separate --p profile per client brand rather than one blended profile that averages everyone’s aesthetic together. Use Conversational Mode for client-facing revision sessions where you’re iterating live on a call — it’s slower than direct prompt editing but far easier to explain and demo in real time.

For teams generating high volumes of similar assets (e-commerce catalogs, ad variant testing), build a prompt template with variable placeholders for subject, setting, and parameters, then swap in the running values from a spreadsheet — this keeps output style consistent across dozens of images without manually re-tuning stylize and chaos each time. Finally, budget Fast GPU hours deliberately: track actual usage for two weeks after switching to V8.2, since the model’s speed improvements mean your historical Fast-hour consumption from V7 or early V8 builds is no longer an accurate baseline.

Licensing is worth a deliberate check before you ship anything client-facing. A paid Midjourney subscription generally grants you commercial usage rights to what you generate, but the exact terms have shifted before and vary by plan tier, so re-read the current terms of service rather than relying on what a plan covered a year ago. If your work touches a regulated industry (advertising claims, pharma, finance), route AI-generated imagery through the same legal review your team already applies to stock photography — the fact that an image was AI-generated doesn’t exempt it from the same trademark, likeness, or brand-safety checks.

It’s also worth setting expectations with clients or stakeholders early: V8.2 is a moving target. Midjourney shipped three default-model changes between March and July 2026, and there’s no indication that pace is slowing. Building a workflow that documents which model version, personalization profile, and parameters produced an approved asset makes it far easier to reproduce or extend that work later, even after the next version ships.

Frequently Asked Questions

Is Midjourney V8.2 free to use?

No. Midjourney has not offered a free trial since March 2023. The minimum entry point is the Basic plan at $10/month, and V8.2 is included at no extra charge across all four subscription tiers.

Do I need to specify –v 8.2 in every prompt?

No. V8.2 has been the default model since July 24, 2026, so any prompt without a version flag automatically uses it. Adding --v 8.2 explicitly is still useful for documentation and troubleshooting.

What’s the difference between V8.2 and V8.1?

V8.2 is described by Midjourney as an aesthetics-and-personalization release: bolder and more sophisticated default output, better taste-reading personalization, restored full image prompting, and new prompt tools like the Prompt Shortener and an updated /describe. V8.1 had a more restrained, V7-inspired look.

Does Midjourney V8.2 have an official API?

Midjourney does not publish an official public REST API. Automated or batch workflows generally rely on Discord automation tooling or the web app’s built-in batch features, and any third-party “Midjourney API” service is an unofficial wrapper — review its terms and Midjourney’s own usage policy before relying on it for production work.

Can I use Midjourney V8.2 images commercially?

Paid subscribers generally receive commercial usage rights to images they generate under Midjourney’s standard terms. If you need generations kept private before a launch, Pro and Mega plans add Stealth Mode, which hides your work from the public community feed — Basic and Standard do not include this option.

Why do my images look more stylized since the V8.2 update?

V8.2’s default aesthetic is intentionally bolder than V8.1’s. If you preferred the more restrained look, lower your --stylize value by roughly 50-100 points from your previous baseline.

How does native 2K resolution work in V8.2?

Unlike older versions where 2K required a separate upscale pass with some quality loss, V8.2 (like V8.1 before it) generates natively at 2K resolution, so the exported final image is closer in fidelity to the original generation.

Should I rebuild my personalization profile after the V8.2 update?

Midjourney’s own guidance suggests re-ranking fresh image pairs on the new model rather than assuming an older profile transfers perfectly, since V8.2 reads taste signals differently and has an expanded image pool to draw from during personalization.

What’s the difference between Draft Mode and Relax Mode?

They solve different problems. Draft Mode trades image fidelity for speed and lower Fast-hour cost, and is meant for rough composition testing before a final render. Relax Mode keeps full V8.2 quality but queues your generation behind other Relax-tier jobs, so it’s slower in wall-clock time but doesn’t consume your Fast GPU hour allotment at all. Relax Mode is available on Standard tier and above.

Can Midjourney V8.2 keep a character consistent across multiple images?

Yes, using the Omni-Reference parameter (--oref) with a high --oref_weight value, typically 500 or above. This is the main tool for comic panels, brand mascots, or any project that needs the same subject to appear recognizably the same across a series of separate generations.

Related Coverage

• How to Use Adobe Firefly: 12 Steps, 90 Min [2026]
• How to Use Seedream 5.0 Pro: 12 Steps, 90 Min [2026]
• How to Use the Nano Banana Pro API: 12 Steps, 90 Min [2026]
• How to Use Grok Imagine Image 2.0: 12 Steps, 70 Min [2026]
• How to Use Stable Diffusion 3.5: 11 Steps, 90 Min [2026]
• Best AI Chatbots: 8 Ranked, $200 Price Gap [2026]