What it actually is
Midjourney is an image generator with a strong and deliberate point of view. Where most tools try to render whatever you describe as literally as possible, Midjourney interprets — and its interpretation is consistently more attractive than the literal thing would have been. It also now runs as a proper web application rather than living inside Discord, which removed the single biggest complaint long-time users had.
What it’s genuinely good at
- Aesthetics, plainly. Lighting, composition and colour come out looking considered rather than generated. This remains its defining advantage.
- Stylistic consistency. Once you find a look you like, you can hold onto it across a whole set of images — essential for anything that has to sit together.
- Cinematic and painterly ranges. Concept art, cover art, mood boards, and atmospheric illustration are where it is comfortably ahead of everything else.
- Style references. Feeding it an image to steer from is well implemented and is the fastest way to get a coherent series.
Where it falls short
Text inside images is bad. Not slightly imperfect — genuinely unreliable. Words come out misspelled or as plausible-looking gibberish. If your image needs a slogan, a label or a title, do not fight this; use Ideogram, which exists precisely to solve it.
It interprets rather than obeys. Ask for a specific product on a specific background and you will get something in the spirit of your request rather than the thing itself. That is a feature for art and a serious problem for product work.
No free tier. You cannot properly evaluate it without paying. Everything worth judging about it is behind the subscription.
Parameter overhead. Getting the most out of it means learning its flags and settings. Casual users get decent results; consistent results take study.
Who should use it
Designers, concept artists, illustrators, and anyone who needs images that look beautiful rather than images that match a specification. If the brief is “make this look striking”, Midjourney is the answer.
Choose something else if: your image must contain readable words (Ideogram), you need literal accuracy to a detailed brief (DALL·E 3), or the output will be reviewed by a legal team (Adobe Firefly, which is trained on licensed material).
Pricing
Subscription only, with no meaningful free tier. Entry plans start around $10/month for a limited amount of generation time, rising through tiers that add faster rendering and stealth mode for private work. For occasional users this is poor value; for anyone generating images weekly it is easy to justify.
Alternatives worth considering
- DALL·E 3 — follows instructions far more literally, and handles simple text.
- Ideogram — the clear winner for typography, posters and logos.
- Adobe Firefly — commercially safer training data and deep Photoshop integration.
- Leonardo.ai — better suited to game assets and consistent character work.
Our verdict: still the quality king for anything artistic. Midjourney is not a precision instrument and was never meant to be — use it when the goal is a beautiful image, and reach for a different tool the moment the goal becomes a specific one.
Frequently asked questions
Do I still need Discord to use Midjourney?
No. It runs as a standalone web app now, which is a substantial improvement for anyone who found the old workflow awkward.
Is there a free version?
Not in any meaningful sense. There is no free tier you can properly test image quality with — you have to subscribe.
Can I use Midjourney images commercially?
Yes on a paid plan, subject to their terms. On any free or trial access the rules differ, so check the current licence before using images commercially.
Why does Midjourney spell words wrong?
It renders text as visual pattern rather than as characters. Models trained specifically for typography — Ideogram above all — handle this correctly.