Midjourney: A Real User’s Honest Overview
I’ve been using Midjourney since early 2023, and it’s become my go-to tool for generating images when I need something specific that stock photography or a quick sketch can’t provide. It’s not perfect—far from it—but it’s the most consistently useful AI image generator I’ve found for my work as a freelance writer and concept artist. Here’s what it actually does well, where it falls short, and whether it’s worth your money.
What It Does Well
Midjourney excels at producing visually striking, high-resolution images with a distinct aesthetic. The default style leans toward painterly, dramatic, and slightly surreal—think fantasy landscapes, cinematic portraits, or moody interiors. For example, I recently needed an image of a “decaying Victorian library with bioluminescent moss” for a client’s book cover. Midjourney gave me four solid options in under a minute, each with detailed textures, realistic lighting, and a cohesive color palette. The moss glowed subtly, the bookshelves receded into shadow, and the overall composition felt intentional.
Another strength is its ability to handle abstract or conceptual prompts. I’ve generated “a clock melting into a puddle of stars” and “a city made of glass and smoke”—both came out with surprising coherence. The tool understands spatial relationships better than many competitors; objects don’t often morph into each other unless the prompt explicitly calls for it.
Limitations You’ll Hit Fast
The biggest frustration is the lack of precise control. You can’t draw a rough layout and say, “put the tree here, the person there.” The prompt is text-only, so you’re at the mercy of the model’s interpretation. Want a “red car” to be a specific shade of crimson? Good luck—you’ll likely get a generic red, maybe leaning orange or maroon. You can iterate with variations, but it’s a guessing game.
Faces and hands are still unreliable. In early 2024, Midjourney improved significantly, but I still get six-fingered hands or eyes that don’t quite align. For close-up portraits, I often need to generate 10–20 images to get one with no obvious deformities. Also, the tool struggles with text—if you ask for a sign that says “OPEN,” it’ll likely produce gibberish or misspell it.
Key Workflows
The standard workflow is: join the Midjourney Discord server (or use the web app, which launched in late 2024), type /imagine followed by a prompt, and wait 30–60 seconds for four images. You can then upscale one, create variations, or use “remix” to tweak the prompt. For complex projects, I use a multi-step process:
- Initial generation: Broad prompt to get a vibe.
- Vary (subtle): Adjust the best image slightly to fix details.
- Upscale: Increase resolution to 1024x1024 or higher (paid plans allow 2x upscaling).
- Inpainting (via external tools like Photoshop): I fix hands, text, or unwanted elements manually, since Midjourney’s own inpainting is weak.
For batch work, I run multiple prompts in parallel using the “fast” mode (which costs extra on the basic plan). I also use the “style reference” feature to mimic a specific artist’s look—say, “in the style of Edward Hopper” for moody realism.
Pricing Reality
Midjourney is paid-only, starting at $10/month for the Basic Plan, which gives you 3.3 hours of GPU time per month (roughly 200–300 image generations). The Standard Plan ($30/month) gives 15 hours and unlimited “relax” mode (slower, no GPU limit). The Pro Plan ($60/month) includes 30 hours and stealth mode (your images aren’t visible in the gallery).
Here’s the catch: “fast” mode burns through time quickly. Each generation costs about 0.5–1 minute of GPU time, but upscaling and variations add up. If you’re generating 50 images a day, you’ll exhaust the Basic Plan in a week. I’m on the Standard Plan and it’s enough for my part-time use. Also, you can pause and resume billing, which helps if you only need it for a project.
Who Should Use It
Midjourney is best for:
- Concept artists and illustrators who need quick visual references for compositions, lighting, or color palettes.
- Writers and marketers looking for unique, non-stock images for blog posts, social media, or book covers (with careful editing).
- Game designers who want to generate environments or character concepts quickly.
- Hobbyists who enjoy experimenting with prompts and don’t mind the learning curve.
It’s not for:
- Photographers who need photorealistic images of real people or specific products (Midjourney’s “realism” is still uncanny valley).
- Graphic designers who need precise vector graphics, logos, or text-heavy layouts.
- Anyone on a tight budget—the free tier (Trial) is limited to 25 generations and then you need to pay.
Bottom Line
Midjourney is a powerful tool for visual ideation and striking imagery, but it’s not a magic wand. You’ll spend time iterating, editing, and accepting imperfections. If you’re okay with that trade-off, it’s worth the subscription. If you need pixel-perfect control or realistic human faces, look elsewhere or pair it with manual editing software.
