My Complete Midjourney Review: After Three Years of Daily Use

Hands-On Verdict

Here’s the honest way to evaluate Midjourney in May 2026: stop asking whether it makes pretty demo images - it does, and it always has. The better question is whether it reliably saves you time on the work you actually ship, week in and week out.

I’ve been using Midjourney since the early beta days, back when it produced fuzzy watercolors that felt more like finger painting than professional art. Three years later, I’m still here. That tells you something. But the V8 transition - which is very much underway right now - has introduced a weird limbo period where you’re choosing between V7’s feature completeness and V8.1’s raw speed. That tension shapes everything I’m about to say.

This review was verified on May 5, 2026, roughly five days after V8.1 hit midjourney.com and Discord proper. Pricing details are a snapshot - Midjourney can change plan names, limits, and bundles without much warning, and they have done so before. If you read nothing else, here’s my rule of thumb: Midjourney is still the single best AI image generator for artistic taste and visual polish, but the gap between it and the pack has narrowed more in the last six months than in the two years before that.

The V8 Era: What Actually Changed in 2026

Let me give you the timeline because it matters for understanding where things stand right now:

• January 9, 2026: Niji 7 launched - still the strongest anime-focused model on the market.
• March 17, 2026: V8 Alpha dropped on alpha.midjourney.com. 5x faster generation, native 2K HD, text rendering that actually works when you wrap text in quotes. But no Relax mode, and premium features cost 4x more GPU time.
• April 14, 2026: V8.1 Alpha replaced V8.0. HD mode got 3x faster and 3x cheaper. Standard resolution became 50% faster and 25% cheaper. Image prompts and weights returned (they were missing in V8.0). The default aesthetic was dialed back toward V7’s familiar feel after V8.0’s look got… mixed reviews.
• April 30, 2026: V8.1 shipped to midjourney.com and Discord. Image quality improved again, particularly for SREFs and Moodboards in HD.

If that sounds like a lot of iteration in six weeks - it is. Midjourney is moving fast, and that speed has a cost: features keep getting yanked and re-added. I’ll get into that.

Here’s the important structural thing to understand: V7 is still the current default model. To use V8.1, you need to unlock a Global V7/V8 Personalization Profile - which means you have to go through the image-rating process that trains the model on your taste. It’s not optional. Without a profile, you’re locked out of the V8 series entirely.

Getting Started: Discord Is Now Optional

When I first wrote about Midjourney, signing up meant getting shoved into a Discord server with thousands of strangers, your prompts scrolling past in a chaotic river of other people’s generations. It was weirdly social and genuinely off-putting for anyone who didn’t already live in Discord.

In 2026, the web interface at midjourney.com is a fully mature product. You can sign up with a Google or Discord account, generate images through a clean visual interface, browse your entire history, organize assets, and never touch Discord at all. The Imagine bar gives you point-and-click access to parameters that you used to have to memorize as text commands. There’s a grid view for comparing batches, a lightbox for rating images, and sidebar settings that don’t block your workspace.

That said, Discord hasn’t been abandoned. The community there is still active, the daily theme channels still run, and power users who prefer chat-based workflows can keep doing their thing. But if you’re new to Midjourney in 2026, you’ll almost certainly start on the web, and that’s a good thing. The learning curve used to be two-fold - you had to learn both prompt engineering and Discord navigation. Now you only have to learn one of those things.

For V8 series models, there’s a catch: you still need to use alpha.midjourney.com (a separate portal) for some features during testing periods. It’s not confusing once you know about it, but it’s also not obvious to newcomers.

Image Quality: Where V8.1 Delivers (and Where It Doesn’t)

Let’s talk about what you’re actually here for: the images. And here’s the thing - Midjourney’s core strength hasn’t changed. It produces images with a cinematic sensibility that no other tool matches. The lighting choices feel intentional. The color palettes look like they were chosen by an art director, not a probability distribution. Skin textures, fabric draping, environmental atmosphere - V8.1 handles these with the same sophistication V7 brought, but faster.

The HD 2K mode is a genuine workflow improvement. Getting a 2048px image natively - without the two-step upscale dance you used to need - means your final candidates are usable right out of the generator. HD costs 1.33 GPU minutes compared to less than 1 minute for standard resolution. That’s cheap enough that you don’t have to hoard HD for only your best images. It’s effectively the new default if you’re on Standard or above.

Raw mode deserves special mention. By default, Midjourney applies its own aesthetic seasoning to every image - which is why most Midjourney images have that recognizable “Midjourney look.” Raw mode strips that away, giving you more direct prompt-to-output fidelity. If you’re generating product concepts, architectural renders, or source frames for video animation, Raw mode is where you should be. It’s not a minor toggle - it fundamentally changes how the model responds to your prompts.

But here’s where I have to be honest: V8.1’s image quality is not uniformly better than V7. The V8.0 alpha had a noticeably different aesthetic that many users (myself included) found weirdly flat and less charming. V8.1 dialed it back toward V7’s feel, and the April 30 quality update improved sharpness further. But if you’re especially attached to V7’s specific aesthetic signature, you might prefer staying on V7 for certain types of work. The two models have different personalities, and switching between them isn’t seamless.

The Gap: What V8.1 Is Missing

This is the part that frustrates me. V8.1 is the fastest model Midjourney has ever shipped, but the feature compatibility chart from their own documentation shows some glaring holes:

• No dedicated upscalers. V7 had Subtle and Creative upscalers. V8.1 has none. The “Run as HD” button works as a pseudo-upscaler, but it re-generates with minor variations rather than genuinely upscaling. That’s not the same thing.
• No Character Reference. No Omni Reference. If you need the same character across multiple images, you’re stuck on V7.
• No Multi-Prompting. The :: syntax for weighted multi-concept prompts doesn’t work in V8.1.
• No Turbo Mode. If you want the absolute fastest generation speed, you need V7.
• No dedicated Quality Parameter. V8 uses its own internal quality logic.

The official docs say that upscalers, inpainting, and outpainting upgrades are “next” on the roadmap. But as of May 2026, they’re not here yet. For professional workflows that depend on these features, you’re effectively locked to V7.

The Prompt System: Power and Personality

Midjourney’s parameter system remains one of its strongest assets. You can control aspect ratios (--ar), stylization intensity (--stylize or --s), chaos/variation (--chaos), weirdness (--weird), and experimental parameters (--exp). The --raw parameter strips the default aesthetic. The --no parameter excludes elements. The --seed parameter locks reproducibility.

But the real power move in 2026 is the style control ecosystem: Personalization Profiles, Moodboards, and Style References (SREFs). These three systems work together to give you fine-grained control over the visual identity of your outputs:

• Personalization Profiles learn your aesthetic preferences from the images you rate. Once trained, adding --p to your prompt applies your personal taste vector to the generation.
• Moodboards let you upload curated image collections that the model uses as style inspiration. Think of it as a Pinterest board that actively shapes your output.
• SREF codes are numeric style tokens that you can share, discover, and apply. The community has built massive libraries of curated SREF codes for everything from ”90s anime cel shading” to “Kodak Portra 400 film grain.” Find a code you like, slap --sref on your prompt, and you’ve inherited that style.

I use a combination of these daily. For client work, I’ll build a Moodboard from their brand imagery, pair it with my own Personalization Profile, and add a subtle SREF for finishing texture. The results are distinctive in a way that generic prompting simply cannot achieve.

The downside? Mastering this system takes time. There’s a genuine skill curve here, and the difference between “that looks cool” and “that looks like exactly what I needed” often comes down to your fluency with these parameters.

Pricing in 2026: The GPU Time Reality Check

Midjourney’s pricing structure has stayed consistent, which is both good (predictable) and bad (competitors keep getting cheaper):

Plan	Monthly	Fast GPU	Key Extras
Basic	$10	3.3 hr (200 min)	SD video only, no Relax
Standard	$30	15 hr	Unlimited Relax images
Pro	$60	30 hr	Unlimited Relax images & SD video, Stealth Mode
Mega	$120	60 hr	Same as Pro, higher limits

Annual billing knocks 20% off all tiers. Extra GPU hours cost a flat $4/hr across every plan. Companies with over $1M in annual gross revenue are required to use Pro or Mega.

Here’s my honest take: the Standard plan at $30/month is the real entry point for anyone serious. The Basic plan’s 3.3 hours goes shockingly fast - a single afternoon of experimentation can burn through half of it, especially if you’re using HD mode (1.33 min per job). With V8.1’s SD mode costing less than 1 GPU minute per job, you can stretch Standard’s 15 hours pretty far, but heavy creators will feel the ceiling.

Relax Mode - unlimited slower generation - is the saving grace of Standard and above. For batch work, moodboard exploration, or overnight rendering queues, it’s genuinely useful. Just don’t expect instant results; Relax queue times vary from a few seconds to several minutes depending on server load.

Is $30/month fair? Compared to DALL-E’s integration with a $20/month ChatGPT Plus subscription, it’s steeper. Compared to running Stable Diffusion locally with a $1,500 GPU, it’s a bargain. The value proposition depends entirely on how much you generate.

Video: Yes, Midjourney Can Animate Images Now

Midjourney added image-to-video generation: upload an image (or use one you just generated), add a text prompt describing the motion, and get a 5-second video clip. You can extend it up to four times for a maximum of 21 seconds.

This is not a Sora competitor or a Runway replacement. The video quality is solid for short loops, subtle camera movements, and atmospheric b-roll. It won’t handle complex character animation, dialogue, or multi-shot sequences. But for turning a Midjourney still into a short atmospheric clip - something you’d use as a social media teaser or a mood piece - it does the job.

Pro and Mega plans get unlimited SD video generation in Relax Mode. Basic and Standard get limited Fast video generation only (SD on Basic, SD & HD on Standard). If video is central to your workflow, you’ll want a dedicated tool like PixVerse or Runway alongside Midjourney, not instead of it.

Where Midjourney Still Falls Short

Beyond the V8.1 feature gaps I already covered, there are structural issues worth mentioning.

The pure diffusion architecture is becoming a liability. Competitors like Nano Banana Pro and GPT Image 1.5 use hybrid autoregressive-diffusion approaches that dramatically improve complex prompt adherence. The Decoder tested V8 against these models with a deliberately tricky multi-subject prompt - “a horse riding an astronaut” - and V8 consistently failed where the AR-blended models succeeded. For simple aesthetic prompts, Midjourney wins. For complex compositional instructions, it’s falling behind.

No official API. Midjourney still doesn’t offer a developer API. Every “Midjourney API” you see advertised is an unofficial third-party wrapper that risks account bans. If you need programmatic image generation at scale, you’re looking at alternatives.

The Personalization gate for V8 is annoying. I understand why Midjourney wants users to build taste profiles - it dramatically improves output quality. But forcing every V8-curious user through a rating exercise before they can even try the model is user-hostile. There should be a “skip for now” option.

Content moderation remains opaque. Prompts get blocked with no clear explanation of which word triggered the filter. Similar prompts can pass or fail unpredictably. For professional users with tight deadlines, this uncertainty is maddening.

Commercial Use and Copyright: Still Murky

Midjourney’s terms say that paying subscribers get “General Commercial Terms” - meaning you can use generated images in commercial work. But Midjourney explicitly states they cannot offer copyright guidance, because the legal status of AI-generated images varies by jurisdiction.

The US Copyright Office has maintained that purely AI-generated images without sufficient human authorship aren’t copyrightable. This means an unmodified Midjourney output likely has no copyright protection - anyone can use it. You can use it commercially per Midjourney’s license, but you can’t stop others from doing the same.

For most creative professionals, this isn’t a dealbreaker - the images are usually part of a larger copyrighted work or get modified enough during post-production to meet the authorship threshold. But if your business model depends on owning exclusive rights to AI-generated assets, you need a lawyer, not a Midjourney subscription.

Comparing to the Competition in 2026

The landscape has shifted significantly. Here’s where things stand:

DALL-E / ChatGPT Image Generation: OpenAI’s image models now use AR-blended architectures that handle complex prompts more reliably than Midjourney’s pure diffusion approach. If your priority is prompt adherence - getting exactly what you described - DALL-E is often better. But Midjourney’s images still look more polished and aesthetically intentional. It’s a taste-vs-precision tradeoff.

Nano Banana Pro (Google): Currently the best photorealism model on the market. Skin, fabric, lighting - it edges out Midjourney for hyperrealistic human subjects. But Midjourney’s artistic range is broader. Nano Banana is a scalpel; Midjourney is a paintbrush.

Ideogram: The undisputed king of text-in-image rendering. If your images need embedded typography, logos, or labels, Ideogram is better. For pure visual artistry, Midjourney leads.

Leonardo.Ai: Has narrowed the gap significantly in 2026 - its photorealism now rivals Midjourney’s, and its editing/refinement tools are more practical for production work. If you need workflow tools more than peak aesthetic quality, Leonardo may be a better fit.

Stable Diffusion / Flux 2: Open-weight models give you unlimited customization, no content filters, and one-time hardware costs. Flux 2 in particular produces excellent images. But the setup complexity and hardware requirements are substantial. Midjourney’s managed experience still produces better results with less friction for most users.

Adobe Firefly: The safest choice for enterprise commercial work - trained on licensed content, integrated with Creative Cloud. But its artistic output is noticeably less distinctive than Midjourney’s. It’s a production tool, not a creative exploration tool.

Who Should Use Midjourney - and Who Shouldn’t

Use Midjourney if: you care about visual taste more than pixel-perfect accuracy, you’re building mood boards or concept art, you create marketing visuals where aesthetic polish matters, you enjoy the craft of prompt engineering, or you work in an artistic field where “the look” is the deliverable.

Skip Midjourney if: you need free access, you need pixel-accurate text rendering, you require programmatic/API access, you want a beginner-friendly template-based tool, or your workflows depend on features currently missing from V8.1 (character consistency, dedicated upscaling).

My Recommendation

Midjourney in May 2026 is the best artistic image generator available - but it’s a tool in transition. V8.1’s speed improvements are genuine and transformative for iteration-heavy workflows. HD 2K output eliminates the upscale tax. The web interface makes the whole experience accessible to people who would’ve bounced off the Discord requirement a year ago.

But the V8 series is incomplete. Missing features that professionals depend on aren’t bug fixes - they’re structural gaps. The pure diffusion architecture, while still producing the most beautiful images in the industry, is showing its age against hybrid approaches for complex compositional tasks.

If you’re already in the Midjourney ecosystem, upgrade to Standard, build your Personalization Profile, and use V8.1 for most work while keeping V7 handy for character consistency and dedicated upscaling. If you’re evaluating AI image generators for the first time, Midjourney should be your first stop - just go in knowing that you’re buying into an opinionated, artist-first tool that expects you to learn its language, not the other way around.

At $30/month for Standard, Midjourney remains my recommendation for serious creative work. But the recommendation comes with an asterisk in 2026 that it didn’t carry a year ago: the competition is catching up faster than Midjourney is pulling away.