Midjourney - Professional AI Art and Image Generation

Midjourney is the model artists reach for when a picture has to look composed, with strong light, texture and style range across any aspect ratio you pick.

PromptDescribe what you want Midjourney to create

Translate your prompt to English for better results.

0/5000

Aspect RatioOfficial

Native

Square

Portrait

Landscape

Cinematic

Native, portrait, landscape, and cinematic aspect ratios render directly with Midjourney without reframe delay or extra credits.Official ratios: 1:1 • 9:16 • 3:4 • 2:3 • 4:5 • 16:9 • 4:3 • 3:2 • 5:4 • 21:9

Midjourney Controls

Adjust Midjourney speed, version, stylization, and variety to match your creative direction.

Generation Speed

Model Version

StylizationStylization: 100

Stylization 0-1000 • Higher values produce more artistic interpretations.

VarietyVariety: 0

Variety 0-100 • Increase in steps of five to explore diverse grids.

Output Format

This Midjourney generation will cost 60 Credits

Available Credits: 0

Type a prompt to generate

Output Gallery

Webhook-completed Midjourney results backed up to R2 appear here instantly

About Midjourney

Midjourney is the model to pick when the image has to look designed rather than photographed. This workspace exposes the controls that change its output: version, speed, stylization and variety, plus up to five reference uploads for image-to-image work. Jobs run asynchronously and land in the gallery when the webhook reports back.

The version dropdown is the first decision. Version 7 is selected by default, and Version 6, Version 6.1 and Niji 6 are there for work that has to match an existing set or lean illustration. Speed sits alongside it: Relaxed is the default, with Fast and Turbo for a quicker turnaround. These are not cosmetic settings. The version and speed are sent with the job, so a look you settle on can be reproduced on the next request by keeping the same combination.

Stylization is the dial most people miss. It runs from 0 to 1000 and starts at 100, which is a middle setting: low values follow the prompt closely, high values let the model interpret the wording more freely. Variety runs from 0 to 100 in steps of five, and raising it asks for a visibly different take rather than a cleaner version of the same idea. Change one of the two at a time; adjusting both makes it hard to tell which setting produced the result you liked.

Everything runs as a queued job. The workspace shows an estimate while it waits, and the render appears in the output gallery when Midjourney calls back. The finished file is copied to our storage, so the gallery keeps working after Midjourney's own link expires. If the job fails, the credits return without a support ticket. Other pages here specialise: Flux Kontext for surgical edits to one photo, Nano Banana for cheap iteration on a reference set. This page is for the frame that has to carry an art direction.

Stylization is quantized to steps of 50, so the values you can land on are 0, 50, 100 and so on; nudging it by a few points does nothing. That makes it a coarse dial for the overall treatment, while the prompt carries the specifics. Variety is the other half: it opens at 0, meaning a first submission is not asking for a new composition at all. If you want two genuinely different arrangements, raise variety rather than stylization, because the latter changes the treatment, not the layout. Settle the layout first, then the treatment; doing it the other way round means redoing both.

The prompt is translated before it is sent, and that switch is on by default. A short brief written in Chinese is converted to English by the translation step, and if the translation fails the original wording is used rather than blocking the run. Turn the switch off when the exact phrasing is part of the brief — a brand name, a quoted line, a deliberate misspelling — because a translation can smooth those away. The version, speed, stylization and variety settings are sent as separate fields, so switching translation off does not affect how the look is configured.

How Midjourney works

1
Set version and speed
Version 7 by default. Drop to 6 or 6.1 for consistency with older work, or Niji 6 for illustration. Leave speed on Relaxed unless you are waiting on a deadline.
2
Write the prompt and add references
For text to image, describe subject, light, lens and mood, and let the aspect ratio settle the composition. For image to image, upload up to five references and describe the transformation rather than the source.
3
Tune one dial, then submit
Pick a ratio from the official list, adjust stylization or variety one at a time, and send the job. The credit cost is shown on the button before the job leaves the page.

What Midjourney is best for

Key art with a fixed direction
Lock a stylization value and a version, then keep producing frames that belong to the same campaign instead of drifting with every new prompt.
Illustration and character work
Niji 6 handles the illustration-leaning pieces while Version 7 covers the photographic ones, from the same workspace and the same reference uploads.
Image-to-image restyling
Upload a layout or a rough and describe the finished treatment. References are capped at five, which keeps a set from pulling in five unrelated directions.
Comparisons you can defend
Raise variety to see a genuinely different interpretation of one prompt, then choose with evidence instead of another round of guessing.

Midjourney at a glance

Model: Midjourney V7 by default; V6, V6.1 and Niji 6 selectable
Modes: Text to image; image to image with up to 5 references
Controls: Speed Relaxed/Fast/Turbo, stylization 0-1000, variety 0-100
Aspect ratios: Auto, 1:1, 3:4, 9:16, 4:3, 16:9
Credit cost: Shown on the generate button; charged once per submitted job
Output: Delivered by webhook and backed up to our storage

Midjourney FAQ

Which Midjourney version should I choose?: Version 7 is the default and the best starting point. Use 6 or 6.1 when a new image has to sit next to older Midjourney work, and Niji 6 for illustration. The version travels with the job, so keeping the same selection keeps the look stable.
What do stylization and variety actually change?: Stylization sets how much freedom the model has to interpret your prompt, from literal at 0 to strongly artistic at 1000, with 100 as the default. Variety asks for a different composition of the same idea in steps of five. Move one at a time so you can tell what caused the change.
Can I work from a reference image?: Yes. Image to image accepts up to five uploads, 10 MB each in PNG, JPG, JPEG or WEBP. Describe what should change rather than what is already visible, because the reference carries the subject and the prompt carries the transformation.
How do I know when a render is ready?: The job is queued at Midjourney and the workspace shows an estimate while it runs. The result appears in the output gallery when the webhook fires, and the file is copied to our storage so the link does not expire.
What happens to my credits if a job fails?: They are returned automatically when Midjourney reports a failure. You are charged for a submitted job that completes, not for one that fails, so a broken queue does not turn into a support ticket.
How long does a Midjourney job take?: The page starts a 45-second estimate when you submit, and that figure is a starting point rather than a promise: the job queues at Midjourney and the wait depends on load. Because the render runs on Midjourney's side, closing the tab does not cancel it; you can leave and check the gallery later. If you resubmit, keep the same version and speed: both travel with the job, so the look stays reproducible.

About Midjourney

How Midjourney works

Set version and speed

Version 7 by default. Drop to 6 or 6.1 for consistency with older work, or Niji 6 for illustration. Leave speed on Relaxed unless you are waiting on a deadline.

Write the prompt and add references

For text to image, describe subject, light, lens and mood, and let the aspect ratio settle the composition. For image to image, upload up to five references and describe the transformation rather than the source.

Tune one dial, then submit

Pick a ratio from the official list, adjust stylization or variety one at a time, and send the job. The credit cost is shown on the button before the job leaves the page.

What Midjourney is best for

Key art with a fixed direction

Lock a stylization value and a version, then keep producing frames that belong to the same campaign instead of drifting with every new prompt.

Illustration and character work

Niji 6 handles the illustration-leaning pieces while Version 7 covers the photographic ones, from the same workspace and the same reference uploads.

Image-to-image restyling

Upload a layout or a rough and describe the finished treatment. References are capped at five, which keeps a set from pulling in five unrelated directions.

Comparisons you can defend

Raise variety to see a genuinely different interpretation of one prompt, then choose with evidence instead of another round of guessing.

Midjourney at a glance

Model

Midjourney V7 by default; V6, V6.1 and Niji 6 selectable

Modes

Text to image; image to image with up to 5 references

Controls

Speed Relaxed/Fast/Turbo, stylization 0-1000, variety 0-100

Aspect ratios

Auto, 1:1, 3:4, 9:16, 4:3, 16:9

Credit cost

Shown on the generate button; charged once per submitted job

Output

Delivered by webhook and backed up to our storage

Midjourney FAQ

Which Midjourney version should I choose?

Version 7 is the default and the best starting point. Use 6 or 6.1 when a new image has to sit next to older Midjourney work, and Niji 6 for illustration. The version travels with the job, so keeping the same selection keeps the look stable.

What do stylization and variety actually change?

Stylization sets how much freedom the model has to interpret your prompt, from literal at 0 to strongly artistic at 1000, with 100 as the default. Variety asks for a different composition of the same idea in steps of five. Move one at a time so you can tell what caused the change.

Can I work from a reference image?

Yes. Image to image accepts up to five uploads, 10 MB each in PNG, JPG, JPEG or WEBP. Describe what should change rather than what is already visible, because the reference carries the subject and the prompt carries the transformation.

How do I know when a render is ready?

The job is queued at Midjourney and the workspace shows an estimate while it runs. The result appears in the output gallery when the webhook fires, and the file is copied to our storage so the link does not expire.

What happens to my credits if a job fails?

They are returned automatically when Midjourney reports a failure. You are charged for a submitted job that completes, not for one that fails, so a broken queue does not turn into a support ticket.

How long does a Midjourney job take?

The page starts a 45-second estimate when you submit, and that figure is a starting point rather than a promise: the job queues at Midjourney and the wait depends on load. Because the render runs on Midjourney's side, closing the tab does not cancel it; you can leave and check the gallery later. If you resubmit, keep the same version and speed: both travel with the job, so the look stays reproducible.