Midjourney defines aesthetic quality in AI image generation. Give the same prompt to five tools and the result most people call "the prettiest" usually comes from here.
That comes at a cost: the tool does not follow prompts literally. It works like an art director — it understands what you asked for and brings its own taste.
This guide covers the shortest route to getting started and the four parameters that actually give you control over the output.
Starting without Discord
Midjourney ran only through Discord for a long time, and that was its biggest barrier to entry. There is now a web app: you can generate straight from a browser without installing Discord or joining a server.
There is no permanent free tier. The lowest plan is 10 dollars a month for roughly 200 images. Unlimited generation exists only on Standard and above, and only in relax mode.
For a few images a month, Basic is enough. For a creator generating daily, the Standard plan's unlimited relax-mode generation is far more economical.
The four parameters that matter
What accounts for roughly eighty per cent of the output is not the length of your prompt text but using the right parameters.
Aspect ratio (--ar) sets the composition from the start; the default is square, which is wrong for most jobs. Style reference (--sref) carries the colour, texture and line character of an image you like into new generations — the most reliable way to hold a series together.
Character reference (--cref) keeps the same person across scenes; --cw sets how tightly the likeness binds. A high value also fixes the clothing, a low one only the face.
Stylise (--stylize) is the strength of the tool's aesthetic intervention. Low values stay faithful to the prompt; high values give more artistic results further from your brief.
| Parameter | What it does | Example |
|---|---|---|
| --ar | Sets the aspect ratio | --ar 16:9 |
| --sref | Copies an image's style | --sref <url> |
| --cref | Keeps a character consistent | --cref <url> --cw 60 |
| --stylize | Strength of aesthetic intervention | --stylize 150 |
| --no | Excludes an unwanted element | --no text, watermark |
Your images are public by default
On the Basic and Standard plans, everything you generate appears in Midjourney's public gallery. Most users are unaware of this.
Private generation (Stealth Mode) exists only on the Pro and Mega plans. If you work on client projects or on an unannounced product, that detail should determine your plan choice directly.
Where it is the wrong tool
Legible text inside an image: Midjourney has improved but Ideogram remains clearly ahead, particularly past three words.
Preserving a specific pose, composition or spatial geometry: not possible here at all. That work belongs to Stable Diffusion with ControlNet.
Vector output for logos and icons: Midjourney produces pixels. A logo has to be a vector to survive both signage and a favicon; Recraft is the tool for that.
Frequently asked questions
Is Midjourney free?
No. There is no permanent free tier; apart from occasional short trials, using it requires at least the 10-dollar Basic subscription. For free alternatives, Leonardo AI, Ideogram and Bing Image Creator all offer daily allowances.
Can Midjourney images be used commercially?
Images produced by paying subscribers may be used commercially under Midjourney's terms. Companies above one million dollars in annual revenue must be on the Pro or Mega plan. Copyright is separate: fully AI-generated images cannot be copyrighted in most countries.
How do I keep one style across images?
Pass the link of an image you like with the --sref parameter; new generations inherit its colour, texture and line character. If the same character recurs, use --cref.