Midjourney
The image card. Midjourney still makes the most beautiful pictures in the deck, and in 2026 it finally got a real editor and a video mode. It also remains the only major AI creative tool with no free entry point, prices its plans in GPU hours, and publishes everything you make unless you pay extra. Here is what each plan buys and where it loses.

- Rating7.7/10
- Read time8 min
- Data tables2
- StatusChecked September 2026
Midjourney
The best-looking image generator you can subscribe to, sold by the GPU hour, public by default, and with no way to try it before you pay.
- Best for
- Concept art, illustration, editorial and campaign imagery, anyone who cares about aesthetics more than text or product accuracy
- Price
- Basic $10/mo; Standard $30/mo; Pro $60/mo; Mega $120/mo; 20% off billed annually
- Free tier
- None. The trial ended in March 2023 and has not returned
- Made by
- Midjourney, Inc., South San Francisco
- Platforms
- Web app (midjourney.com and the alpha site), Discord; no public API, and the terms ban automated access
- Trains on your data?
- The privacy policy lists data collected through model training, the terms grant Midjourney a perpetual license to your prompts and images, and rating images earns you free GPU time
Rating reflects Midjourney's documented plans and capabilities as of September 2026 plus our own sessions. Some vendor links on this site are affiliate links; they never affect a rating. See our disclosure and method.
What it does
Midjourney turns a text prompt into a grid of four images, and its whole reputation rests on how those four images look. Rivals compete on instruction following or text rendering; Midjourney competes on taste. An August 2026 comparison at Gizmo Times put it plainly: Midjourney V8.2 is the pick for stylized artwork, concept art and aesthetic exploration, while ChatGPT Images 2.0 wins conversational editing and text-heavy graphics, and Google's Nano Banana family wins for people already inside Gemini.
The model moved fast this year. V8.1 became the default on June 11, 2026, replacing V7 with better prompt adherence, better text and an HD mode that renders at twice the size and four times the resolution of V7, with SD images in about four seconds. V8.2 took over on July 24 with a focus on aesthetics, fewer random low-quality results and a much better personalization pass. Then on August 27 Midjourney opened testing of its first V8.2 edit model: edit an image with plain instructions, generate from up to four reference images at once (this replaces omni-reference), inpaint a region, or expand the canvas. It runs on the main site, on the alpha site, and in Discord with a parameter.
Personalization is the feature that separates a casual user from a good one. You rate pairs of images, Midjourney builds a profile of your taste, and every prompt leans toward it. V8.2 draws its rating pool from a larger, cleaner set of images, and a profile code made in V8.2 still works in V8.1 and V7. Rating images also earns free Fast GPU time on every plan, which tells you how valuable your taste data is to the company.
Video is an add-on to images, not a separate product. You pick a still you made or uploaded, choose low or high motion, optionally add a text prompt, and get four 5-second clips. Each clip can be extended four times in 4-second steps to a maximum of 21 seconds, with loop and end-frame options. Output is 480p on all plans; Standard, Pro and Mega can switch to 720p in Fast Mode. There is no text-to-video.
Where it wins
- Aesthetics. Nothing in the hearts suit produces a more striking first draft. Lighting, composition and texture arrive looking finished, and the stylize and raw controls dial the house look up or down.
- Personalization. Once you have a few hundred ratings under a profile, the model reliably makes your kind of image. A paid reviewer at Geeky Curiosity found the V8.2 rating pool a clear step up from V7's, with fewer three-armed distractions and faster payoff.
- The edit model. Instruction editing, multi-image references, inpainting and outpainting were the features Midjourney lacked against ChatGPT and Nano Banana. As of late August 2026 it has them, and they work inside the same personalization and style-reference system.
- Relax Mode economics. On Standard and above, image generation in Relax Mode is unlimited. If you can wait a few minutes per grid, $30 a month buys as many images as you care to make.
- Commercial terms that are easy to read. Subscribe at any tier and you can use your images and videos commercially. The single carve-out, a Pro or Mega plan for companies over $1 million in annual revenue, is stated on the plans page.
Where it loses
- No free tier. ChatGPT, Gemini, Canva and every other creative card in this deck let you try before you pay. Midjourney ended its trial in March 2023, briefly reopened it for about 25 images in August 2024, and has been paid-only since. You cannot test it on your own brief before spending $10.
- Public by default. Everything you make is viewable and remixable by the community unless you have Stealth Mode, which costs $60 a month on Pro. Client work on Basic or Standard is visible to anyone who looks.
- Your content, licensed forever. The terms grant Midjourney a perpetual, irrevocable, sublicensable license to your prompts and outputs, and the privacy policy lists data collected through model training. There is no toggle to opt out. Cursor, Descript and Notion in this deck offer a switch; Midjourney does not.
- Video is a finishing step. Image-to-video only, 480p unless you are on Standard or higher and in Fast Mode, and a four-clip SD batch costs 8 minutes of Fast GPU time (26 minutes in HD). A guide at AI Video Sensei that ran eleven passes on one still got four clean clips, four usable ones and three with visible warping. It animates an image; it does not make a film.
- Legal weather. Disney and Universal sued Midjourney for copyright infringement in 2025 and Warner Bros. followed. In July 2026 the fight was over discovery, with Midjourney arguing that training on copyrighted characters is fair use. Nothing has been decided and nothing stops you using the tool, but a cautious legal team will ask.
- Interface churn. The web app is far better than Discord ever was, but the alpha site was rebuilt in August 2026 and the changelog for the next two weeks is a long list of fixes: settings that reset, Relax flipping back to Fast, Upscale buttons that vanished.
Plans and pricing, September 2026
| Plan | Price | What you get | Who it is for |
|---|---|---|---|
| Basic | $10/mo, $96/yr ($8/mo) | 3.3 hours of Fast GPU time (200 minutes), no Relax Mode, 3 concurrent image jobs, 1 concurrent video job, SD video only, extra GPU time at $4/hr | Trying it, occasional use |
| Standard | $30/mo, $288/yr ($24/mo) | 15 hours of Fast GPU time, unlimited images in Relax Mode, HD video in Fast Mode, 3 concurrent jobs | Most working creators |
| Pro | $60/mo, $576/yr ($48/mo) | 30 hours of Fast GPU time, unlimited images and SD video in Relax Mode, Stealth Mode, 12 concurrent Fast image jobs, 6 Fast video jobs | Client work that must stay private, companies over $1M revenue |
| Mega | $120/mo, $1,152/yr ($96/mo) | 60 hours of Fast GPU time, everything in Pro, 12 concurrent Fast video jobs | Studios and volume producers |
GPU time is the currency. Fast Mode spends your monthly hours and returns a grid in seconds; Relax Mode queues your job at lower priority and, on Standard and above, costs nothing for images. Pro and Mega extend Relax to SD video. Every plan can buy more Fast hours at $4 an hour; blocked jobs never cost GPU time. Annual billing is a 20% discount, paid up front.
Video cost, checked September 2026: a default batch of four SD clips costs 8 minutes of Fast GPU time, and four HD clips cost 26 minutes. Drop the batch size to one with --bs 1 and it is 2 minutes for SD or 7 for HD. Extending a clip costs the same as generating one. On Basic's 200 minutes, that is 25 four-clip SD batches a month if you do nothing else.
Score breakdown
| Criterion | Score | Why |
|---|---|---|
| Capability | 9.0 | Best aesthetics in the suit, a real edit model since August; video is 480p and image-to-video only |
| Ease of use | 8.0 | The web app is simple; personalization is homework and the alpha redesign is still settling |
| Value | 7.5 | Unlimited Relax images at $30 is generous; no trial, and privacy costs $60 |
| Trust | 6.0 | Public by default, perpetual license on your content, no training opt-out, active studio lawsuits |
| Ecosystem | 6.5 | Web and Discord only; no API, automated access banned, no integrations with design tools |
Verdict: the artist's card, held with both eyes open.
If the look of the image is the point, subscribe to Standard, do the personalization ratings in your first week, and live in Relax Mode. Skip Basic unless you will make fewer than a couple of hundred images a month. Go to Pro only for Stealth Mode or because your company's revenue requires it. And if you need text in the image, product accuracy, a free tier or a promise not to train on your work, read the Canva and ChatGPT cards first.
Alternatives
- Canva if you need images inside a design workflow, with a free tier and clearer data controls.
- ChatGPT if you want conversational editing, legible text in images and a free way to start.
- Gemini if you already pay Google and want its Nano Banana models bundled with storage and a chatbot.
Quick answers
Is Midjourney free?
No. Midjourney ended its free trial in March 2023 and, as of September 2026, the cheapest way in is the Basic plan at $10 a month or $96 a year. You can browse the public gallery without an account, but you cannot generate anything without paying.
Can I use Midjourney images commercially?
Yes, on any paid plan, under what Midjourney calls General Commercial Terms. The exception is companies with more than $1 million a year in gross revenue, which must be on Pro or Mega to own their images. Note that everything you make is public by default unless you pay for Stealth Mode on Pro or Mega.
Does Midjourney make video?
Only from a still image. Midjourney animates an image you generated or uploaded into a 5-second clip that you can extend four times to 21 seconds. Output is 480p on every plan and 720p on Standard and above in Fast Mode. There is no text-to-video, so it is a finishing step for images rather than a rival to a full video model.
Sources
- Midjourney Docs — Comparing Midjourney plans, checked September 2026
- Midjourney Docs — Video: batch size, GPU cost, extend, resolution, checked September 2026
- Midjourney Updates — Version 8.2 announcement, July 2026
- Midjourney Updates — Edit model for V8, August 2026
- Midjourney Docs — Terms of Service, effective May 27, 2026
- TechCrunch — Midjourney's discovery fight with Disney, Universal and Warner Bros., July 2026


