The short answer (July 2026): Midjourney is a browser-based AI image generator at midjourney.com that requires no software installation. Plans run from $10/month (Basic) to $120/month (Mega), and there is no free trial. The current default model is V8.1, which became the standard on June 11, 2026, and produces standard-definition images in about 4 seconds. You sign in with a Discord or Google account, type a description at midjourney.com/imagine, and receive four image variations to upscale or edit.
Last verified: July 25, 2026, against official docs and pricing pages.
What Each Plan Costs
Midjourney sells GPU compute time, not image slots. Your monthly fast GPU hours determine how many images you can generate at full speed. On Standard, Pro, and Mega plans, once those hours run out you shift into Relax mode, a slower shared queue that costs nothing extra from your allotment. Basic has no Relax fallback.
| Plan | Monthly price | Annual price (billed upfront) | Fast GPU hours | Relax mode (images) | Stealth mode |
|---|---|---|---|---|---|
| Basic | $10 | $96/yr ($8/mo) | 3.3 hrs | No | No |
| Standard | $30 | $288/yr ($24/mo) | 15 hrs | Yes | No |
| Pro | $60 | $576/yr ($48/mo) | 30 hrs | Yes (includes SD video) | Yes |
| Mega | $120 | $1,152/yr ($96/mo) | 60 hrs | Yes (includes SD video) | Yes |
A standard V8.1 image costs 0.8 GPU minutes. That puts Basic's 3.3 hours (198 minutes) at roughly 247 standard images per month. HD images (2048px, about 12 seconds each) cost 1.3 GPU minutes, so the same Basic budget yields roughly 150 HD images. Fast hours reset each billing cycle and do not roll over. Extra fast GPU time costs $4/hour on any plan.
Standard works out to $2/GPU hour versus Basic's $3.03/GPU hour, making it the better value if you generate regularly.
Getting Access
There is no free trial as of July 2026. Midjourney ended its free trial in 2023. You can browse the public community gallery at midjourney.com without an account, but image generation requires a paid subscription.
To subscribe: go to midjourney.com, click Subscribe, and sign in with either a Discord or Google account. There is no email-and-password option. After choosing a plan, billing goes through Stripe and access is immediate.
The Midjourney Discord server is still accessible for free. You can watch community generations there and ask questions before committing, but you cannot generate images in Discord without a paid plan.
Your First Generation
Navigate to midjourney.com/imagine. The prompt bar at the top reads "What will you imagine?"
Type a description and press Enter. V8.1 returns a 2x2 grid of four variations in about 4 seconds at standard definition. From there:
- Click any image in the grid to expand it
- Choose Upscale to get a single downloadable high-resolution image
- Choose Vary (Subtle) or Vary (Strong) to regenerate close or distant variations of that specific image
- Choose Edit to open the Canvas editor for regional editing
The web interface also has sliders for aspect ratio, stylize, and chaos in the settings panel. You can adjust these visually instead of typing parameters. Either method works.
To change the model version for a single generation, add --v 7 or --v 8.1 at the end of your prompt. To set a persistent default, open the settings panel. V8.1 is the default. Use --v 7 if you need Omni Reference (any image as a visual guide, a V7 feature not yet fully available in V8.1) or --niji 7 for anime-style output.
Writing Prompts That Get Results
V8.1 understands natural sentences. Keyword-spam prompts ("cinematic 8k HDR dramatic moody lighting") still work but produce less precise output than clear descriptions. Write what you actually want to see.
A structure that works consistently: subject + visual details + environment + style reference + parameters.
Example: A woman in her 50s reading a newspaper at a rain-soaked Paris cafe, soft window light from the left, 1970s documentary photography, Kodachrome grain --ar 4:5 --s 200
A few specifics worth knowing:
Put the most important visual element first. Midjourney weights the beginning of the prompt more heavily than the end.
To include readable text inside an image, put that text in double quotes within your prompt: A storefront sign that reads "Grand Opening". V8.1 handles text rendering substantially better than previous versions.
For style transfer, append --sref [image URL] to your prompt. Control how strongly the style applies with --sw between 0 (light influence) and 1000 (heavy). This works with any publicly accessible image URL.
To exclude elements, add --no [element] at the end: --no shadows, --no text, --no people.
Key Parameters
| Parameter | What it does | Range | Default |
|---|---|---|---|
| --ar | Aspect ratio (width:height) | Any ratio | 1:1 |
| --s / --stylize | How much Midjourney's aesthetic overrides your prompt | 0 to 1000 | 100 |
| --c / --chaos | Variation between the four grid images | 0 to 100 | 0 |
| --no | Exclude listed elements from the output | Text list | None |
| --sref [URL] | Apply the visual style from a reference image | Any public image URL | None |
| --sw | How strongly the style reference influences output | 0 to 1000 | 100 |
| --draft | Activates Draft Mode (V8.1 web only): 24 images at lower GPU cost | Flag | Off |
| --v | Model version selector | 7, 8.1, niji 7 | 8.1 |
Common aspect ratios in practice: --ar 1:1 for square social posts, --ar 16:9 for landscape and video thumbnails, --ar 4:5 for Instagram portraits, --ar 3:2 for DSLR-style photos.
Stylize at 0 to 100 keeps output close to your description. Stylize at 400 and above lets Midjourney's aesthetic sense take over, which often produces more striking images at the cost of prompt accuracy. When you need precise adherence, lower it.
Chaos at 0 (the default) keeps all four grid images visually related. Raising it to 50 to 75 during early ideation gives you genuinely different directions to choose from rather than four variations of the same composition.
After Generation: Editing, HD, and Draft Mode
Vary Region (inpainting): After upscaling any image, the Canvas editor lets you draw a freehand or rectangular selection over any part of the image, type a description for that region, and regenerate only that area while preserving the rest. Midjourney recommends selecting a substantial portion of the image. Selections that are too small tend to be ignored.
Rerun as HD: Any V8.1 image can be re-rendered at 2048px (4x the pixel area of standard output) directly from the editor. HD images cost 1.3 GPU minutes and take about 12 seconds. You do not need to regenerate from scratch; click "Rerun as HD" on any upscaled image.
Draft Mode: Added in June 2026 and available on the web only (not Discord), Draft Mode generates 24 images at 512px per prompt for 0.4 GPU minutes. That is six times the images at half the GPU cost of a standard four-image prompt. Use it at the start of a project to explore creative directions cheaply, then drop the --draft flag once you have a direction and generate at full quality.
Adding --sref random inside a draft prompt generates 24 images each with a different randomized style reference, useful for visual brainstorming across style directions at once.
Commercial Use and Privacy
Every paid Midjourney plan includes commercial use rights. One threshold applies: if you work for a company that earned more than $1 million USD in gross revenue in the prior calendar year, Midjourney requires a Pro ($60/mo) or Mega ($120/mo) plan. Individuals and companies below that threshold can use Basic or Standard for commercial work, including client projects, print sales, and merchandise.
Stealth Mode, exclusive to Pro and Mega, makes your generated images private rather than visible in the public Midjourney gallery. On Basic and Standard, every image you generate is publicly visible in the community gallery by default.
If privacy matters to your workflow, Stealth Mode is the primary functional reason to upgrade from Standard to Pro.
FAQ
Is there any way to try Midjourney before paying?
No. The free trial ended in 2023. You can browse the public gallery and watch community generations in the Midjourney Discord server without an account, but generating images requires at minimum the $10/month Basic plan.
What is the practical difference between Fast and Relax mode?
Fast mode uses your monthly GPU hour allotment and processes immediately. Relax mode puts your job in a slower shared queue at no additional GPU hour cost. It is slower but effectively unlimited for images. Relax is available on Standard, Pro, and Mega; Basic has no Relax fallback.
Can I use Midjourney on a phone?
Yes, through any mobile browser. There is no native iOS or Android app. The web app at midjourney.com/imagine works on mobile, though the Canvas editor for regional editing is more practical on a desktop screen.
Which plan makes sense for a freelancer doing occasional client work?
Standard at $30/month covers most independent professionals: 15 fast GPU hours, unlimited Relax mode images, and commercial rights for businesses under $1 million in revenue. If your images need to stay private, that is the functional reason to move to Pro.
What does V8.1 still lack compared to V7?
Omni Reference, a V7 feature that lets you use any image (a character, an object, not just a style source) as a visual guide for generation, has not been fully ported to V8.1 as of July 2026. Midjourney has indicated an improved version is in development. If Omni Reference is central to your workflow, add --v 7 to those specific prompts.
Found this useful? Get AI Weekly free, three issues a week, read by 40,000+ practitioners.