Midjourney

What it actually is

Midjourney turns a text prompt, and optionally a reference image, into generated artwork — that’s the entire premise, and unlike most of the tools on this site, it doesn’t try to be anything else. There’s no chat mode, no document analysis, no general-purpose assistant sitting alongside it. What’s grown around that single premise is genuinely deep: Omni Reference for holding a character or style consistent across many images, Draft Mode for cheap fast iteration, Personalization that learns your aesthetic from images you rate, and — since 2025 — a video model that can animate a still image into a short clip.

Four pricing tiers sit on top of that: Basic ($10/month), Standard ($30/month), Pro ($60/month), and Mega ($120/month), each buying a block of fast GPU time rather than a fixed image count. Standard and above include unlimited Relax-mode generation, which just runs on whatever GPU capacity is available rather than the fast, prioritized queue — slower, but effectively uncapped. Companies earning more than $1 million in annual revenue are required to be on Pro or Mega regardless of how much they actually generate, which is worth knowing if you’re evaluating this for a larger organization.

Where it’s genuinely good

The aesthetic quality is the whole reason people choose Midjourney over technically comparable alternatives, and V8.2 has made that gap harder to close rather than easier. Lighting, texture, and composition come out looking considered by default in a way that used to require careful prompting to coax out of earlier versions. Omni Reference is the feature that’s made the biggest practical difference for working creatives this year — being able to lock a character or product across eight or ten different scenes, reliably, is the difference between a series of nice individual images and an actual usable campaign.

Draft Mode is an underrated addition for exactly this kind of professional workflow: exploring five different creative directions used to mean burning real fast-GPU time on each one, and now it doesn’t. And the fact that the company has stayed self-funded, with a founder who’s been public about not building for investor exits, shows up in the product itself — feature development has been steady rather than lurching toward whatever a funding round demanded.

Where it falls short

The business model around evaluation is the most immediate friction point. There’s genuinely no way to try Midjourney properly before paying — no free tier, and a refund policy narrow enough that it doesn’t function as a real trial either. Combine that with GPU-hour billing instead of a simple image quota, and a first-time user can burn through Basic’s allowance faster than expected without ever quite understanding why.

The pace of model releases is the other real cost, even though it’s also evidence the product is improving quickly. Four default-model changes in roughly a year means the specific parameters and prompting habits that produced great results six months ago may not transfer cleanly to the current version, which is a real relearning cost for anyone using this professionally and consistently. And more broadly, the question of how large AI image models were trained — and what that means for anyone using their outputs commercially — remains a live, unresolved debate across the entire category. It’s not a Midjourney-specific problem, but it’s a legitimate thing to factor in before betting a commercial workflow entirely on outputs from any model trained this way.

Which plan to actually get

Basic, at $10/month, is really only enough for occasional, light use — 3.3 hours of fast GPU time runs out faster than the price suggests once you’re iterating on real work. Standard, at $30/month, is the correct default for most people: unlimited Relax mode means you’re never fully blocked, you’re just choosing between fast and free-but-slower. Move to Pro, at $60/month, specifically if you need Stealth Mode for confidential client work, or if 30 fast hours a month genuinely isn’t enough — not just because it sounds like the “professional” tier. Mega, at $120/month, is built for studios doing serious volume, particularly video work, which consumes GPU time far faster than stills.

Ease of use

The interface itself, especially since the move to a full web app alongside the original Discord bot, is straightforward: type a prompt, generate, refine. What takes real time to learn is everything underneath that simplicity — how GPU-hour billing actually translates into a per-image cost, how the parameters and reference systems interact, and how to write a prompt that gets you a specific result rather than a pleasant surprise. Midjourney rewards deliberate practice more than most AI tools; the gap between a first-week user’s output and a six-month user’s output is genuinely large.

Who it’s actually for

Creative professionals, art directors, and marketing teams who need images with real aesthetic judgment baked in, and who are producing enough volume that a subscription pays for itself quickly. It’s also the strongest pick for anyone specifically doing character- or brand-consistent work across a series, given how far Omni Reference has come. It’s a weaker starting point for someone who just wants to experiment casually before spending money, or a team that needs to share a small number of seats across many occasional users — the per-person subscription model and lack of a real trial both work against that use case specifically.