Midjourney is the AI image generator that defined the category’s aesthetic. It launched in 2022 as a Discord bot, hit cultural saturation in 2023, ditched Discord for a proper web app in 2024, and shipped v8 with native video in March 2026. After four years of generating over 50,000 images for client work and personal projects, here’s the unfiltered review of where Midjourney stands today.
Key takeaways
- Still the aesthetic king — for “make this beautiful,” nothing else is consistently as good.
- FLUX and Imagen 4 closed the gap on photoreal; Ideogram 3 still wins typography.
- v8 added video (5s clips) — fun, not yet a Runway/Sora competitor.
- $10/mo Basic is enough for casual; serious users want Standard ($30) for unlimited slow generations.
- Buy if aesthetics matter most; consider alternatives for typography, photoreal portraits, or commercial precision.
What Midjourney is in 2026
A subscription-only AI image generator with a polished web app (midjourney.com), a Discord bot (legacy but still maintained), an iOS/Android app, and v8 features that include:
- Image generation with v6.1, v7, and v8 models selectable
- Video generation (5-second clips, image-to-video)
- Personalization — train Midjourney on your aesthetic preferences
- Style references (
--sref) — match a reference image’s style - Character references (
--cref) — consistent characters across generations - Editor — inpainting, outpainting, upscaling, vary
- Web app — gallery, organizational tools, rooms (collaborative spaces)
Pricing
| Plan | Price (monthly) | Fast hours | Relax (unlimited) | Stealth mode |
|---|---|---|---|---|
| Basic | $10 | ~3.3 hours (~200 images) | No | No |
| Standard | $30 | 15 hours | Unlimited (slow) | No |
| Pro | $60 | 30 hours | Unlimited (slow) | Yes (private generations) |
| Mega | $120 | 60 hours | Unlimited (slow) | Yes |
Yearly plans get a 20% discount. The pricing logic: Midjourney charges by GPU time. Fast = priority queue. Relax = wait 5-30 min per image. Stealth = your generations aren’t public on midjourney.com (which the others are by default).
Most people land on Standard ($30/month) because the unlimited Relax mode means you can iterate freely without watching a meter. Basic at $10 is fine for casual; Pro/Mega only justify themselves if you need privacy (Stealth) or hundreds of Fast generations weekly.
What Midjourney still does best
1. The aesthetic
This is the one thing that hasn’t changed: Midjourney’s default look is more visually pleasing than any competitor’s. Skin tones, lighting, composition, painterly quality — it leans toward editorial/cinematic in a way that’s instantly recognizable. For “make me a beautiful image of X,” it’s the safest bet.
FLUX and Imagen 4 have caught up on technical quality. They have NOT caught up on this feel. Midjourney’s training on aesthetic preference (literally — users vote on which generations look better, and the model learns) is a moat.
2. Style reference (--sref)
Drop in a URL or seed, and Midjourney matches the style across new prompts. This is the single most-used feature in professional workflows. Want every image in a brand’s visual identity? --sref + a brand image, then generate freely. The consistency is genuinely good.
Random style codes (--sref random) let you discover unique aesthetics you couldn’t describe in words. This is one of the most fun features I’ve used in any AI tool.
3. Character reference (--cref)
Want the same character across multiple images? --cref works well enough for serial illustration, comic-style work, and ongoing campaigns. Not flawless — small facial details drift — but better than competitors at maintaining a recognizable identity.
4. The editor
The web app’s editor lets you inpaint (replace regions), outpaint (extend canvas), upscale to high resolution, vary specific areas, and remix prompts. The UX is polished and the results are good. The “remix” feature alone is responsible for a lot of my best images.
5. Speed at scale
On Fast mode, generations complete in 30-60 seconds. On Relax (Standard+ plans), queue waits are 5-15 min but unlimited. For batch work, the unlimited Relax queue beats per-image pricing models like DALL-E.
Where Midjourney has lost ground
1. Photorealism — FLUX has caught up
FLUX Pro 1.1 (Black Forest Labs) hit photorealism parity in late 2024 and arguably surpassed Midjourney on technical fidelity by 2026. For “realistic photograph of X person doing Y,” FLUX often produces more believable output, especially for skin, hands, and product photography. Imagen 4 (Google) is similarly strong.
Midjourney still has the aesthetic edge for “stylized realism” but pure photoreal is no longer its category.
2. Typography is still bad — Ideogram 3 wins
Midjourney v7 and v8 improved text rendering but it’s still hit-or-miss. “A poster that says HELLO in serif” works ~70% of the time; complex typography fails more often than it works. Ideogram 3 generates legible, varied typography on first try, every time. For text-heavy work, Ideogram is the right tool.
3. No API
There is no official Midjourney API. Third-party wrappers exist (via Discord automation) but they’re unstable and against the ToS. If you need programmatic image generation in a product, FLUX (via Replicate or Together), DALL-E, Imagen, or open-source Stable Diffusion are the options. This is a deliberate Midjourney choice — they want humans, not pipelines.
4. Public generations by default
On Basic and Standard, your generations are publicly visible on midjourney.com. For personal work this is fine; for client work or sensitive material you need Pro+ for Stealth Mode. This is a friction Discord users were used to but new web app users sometimes miss.
5. Video is a fun toy, not a competitor
Midjourney v8’s video feature generates 5-second clips from a still image. The motion is okay; it’s good for “animate this one frame I love.” But Runway Gen-4 and Sora 2 are categorically better for any serious video work. Midjourney video is best treated as a “fun extra” feature, not a Runway replacement.
Midjourney v8 vs the alternatives
| Capability | Midjourney v8 | FLUX Pro 1.1 | Imagen 4 | Ideogram 3 |
|---|---|---|---|---|
| Aesthetic / “beautiful” | Best | Very good | Very good | Good |
| Photorealism | Excellent | Best | Best | Good |
| Typography | OK | Good | Very good | Best |
| Style consistency | –sref is best | Limited | Limited | Limited |
| Character consistency | –cref works | OK | OK | OK |
| API access | None | Yes (Replicate, BFL) | Yes (Google AI) | Yes |
| Video gen | 5s (basic) | None | Veo (separate) | None |
| Pricing entry | $10/mo | $0.04/image (API) | $0.04/image | $8/mo |
Pros and cons
Midjourney pros
- Best default aesthetic of any image generator
- Style reference (`–sref`) is industry-leading for brand consistency
- Character reference (`–cref`) for serial illustration
- Polished web app with great editor (inpaint, outpaint, upscale)
- Unlimited Relax mode on Standard+ plans
- Commercial use included from $10/mo
Midjourney cons
- No official API — can’t integrate into products
- FLUX and Imagen 4 surpass on pure photorealism
- Typography still weaker than Ideogram
- Public generations by default (need Pro for Stealth)
- v8 video is basic vs Runway/Sora
- Subscription-only — no pay-per-image option
Who Midjourney is for
Buy Midjourney if:
– Aesthetic / “make this beautiful” is your priority
– You work in marketing, design, content where look matters more than technical realism
– You want consistent brand visuals via --sref
– You want unlimited iteration via Relax mode
Don’t buy Midjourney if:
– You need an API to integrate into a product — use FLUX
– Your primary need is realistic product photography — FLUX or Imagen
– You generate posters / typography-heavy designs — Ideogram
– You want serious AI video — Runway or Sora
Four years and 50,000 generations later
What’s changed in how I use it:
– 2022-2023: Daily, for everything. Wow factor.
– 2024: Daily, with --sref becoming the workflow center.
– 2025: 3-4× per week. FLUX took over for photoreal product work.
– 2026: Still my default for editorial/aesthetic work. Ideogram for posters, FLUX for product photography, Sora for video.
The honest summary: Midjourney went from “the only good option” to “the best option for a specific kind of work.” That kind of work — aesthetic, stylized, brand-coherent — happens to be a big slice of professional creative output, which is why it still matters.
FAQ
Is Midjourney still the best AI image generator?
For aesthetic / stylized work, yes. For photorealism, FLUX and Imagen 4 tie or beat it. For typography, Ideogram beats it. There’s no single “best” anymore — pick by use case.
Do I still need Discord to use Midjourney?
No. The web app (midjourney.com) has been the recommended UX since 2024 and now has all features. Discord still works but is no longer the primary interface.
Can I use Midjourney images commercially?
Yes — all paid plans (Basic and above) include commercial use rights. Free tier images cannot be used commercially.
Is Midjourney v8 worth upgrading from v7?
The aesthetic step is modest; the big additions are video (5s clips) and improved prompt adherence. If you’re already on v7 and don’t need video, v8 isn’t a must-upgrade — both ship in the same subscription.
Does Midjourney have an API?
No official API. Some third-party services wrap the Discord bot but they’re against the ToS and unstable. For programmatic use, FLUX (via Replicate / Together AI / Black Forest Labs API) is the production-grade alternative.
What’s the difference between Fast and Relax mode?
Fast uses GPU time from your monthly quota (~3.3 hours on Basic = ~200 images). Relax is unlimited but queues your generations behind paying-Fast users — 5-30 min wait per image. Standard and above include unlimited Relax.
How does Midjourney compare to DALL-E 4?
DALL-E 4 (OpenAI, late 2025) has stronger prompt adherence and is excellent at text-in-image. Midjourney has stronger aesthetic defaults and --sref style control. They’re not direct competitors — different strengths. Heavy users often subscribe to both.
Will my generations be private?
On Basic and Standard, no — they’re visible in the public gallery on midjourney.com. Pro and Mega include Stealth Mode for private generations. If privacy matters from day one, budget for Pro at $60/mo.
Is Midjourney worth the subscription?
If aesthetics matter to your work — marketing, design, content, branding — yes; the $30/month Standard plan with unlimited Relax generations pays for itself quickly versus stock imagery or a designer’s time. If you only need the occasional image, or you need an API to integrate generation into a product, a pay-per-image option like FLUX is better value.
Is there a free version of Midjourney?
Not reliably. Midjourney has run limited free trials in the past but suspends them when capacity is tight, and free-trial images can’t be used commercially. In practice, plan on the $10/month Basic plan as the entry point — and note that Basic makes your generations public, so budget for Pro ($60) if you need privacy.
Bottom line
Midjourney v8 in 2026 is still the best AI image generator if you value aesthetic above everything else. It’s no longer the only good option — FLUX has eaten its photorealism lunch, Ideogram its typography lunch, Runway and Sora its video lunch — but the core aesthetic that made it culturally dominant is still uniquely good.
For most professional creative work, the Standard plan at $30/month is the right entry point — unlimited Relax mode means you can iterate freely, and the editor + --sref + --cref features give you real control over output.
If you only subscribe to one image generator, Midjourney is still the safest bet. If you do this for a living, you probably want Midjourney + FLUX (via Replicate) + Ideogram, totaling ~$50/month, to cover the full range of what AI imaging can do in 2026.
Midjourney didn’t win every battle. But four years in, it’s still the tool whose output you’d hang on a wall.