Midjourney: My Take After Three Years in Production

In short: Midjourney turns text prompts into high-quality images and remains a reference for editorial illustration, concept art, and advertising visuals. It now runs through a web interface as well as Discord, uses tiered access plan, and only offers private mode from the Pro tier up. Paying users generally hold commercial rights, but purely AI-generated images aren't legally protectable in the US.

I started using Midjourney seriously back in 2022 on Discord, and I'm still reaching for it every single day in 2026 — for client illustrations, visual concepts, and my own articles. This piece is my field report: what I still love about Midjourney, where I pull it out first, and the cases where I switch to DALL-E or FLUX instead.

What Midjourney is, and why I still use it

Midjourney turns a text description (a prompt) into an image. Published by Midjourney Inc., an independent research lab based in San Francisco, the tool has established itself as a reference for advertising visuals, editorial illustration, and concept art. What still pulls me in after three years: the aesthetic quality and the consistency of its compositions. No other tool gives me that "agency-grade" look this fast.

Historically it was only accessible through Discord (nostalgia kept me there for a while), but the service now offers a web interface on midjourney.com — much simpler for newcomers and far more structured for my high-volume workflows. I moved over to the web entirely back in 2024.

My actual workflow

My usage comes down to a few steps I've stabilized over time. I write a prompt — say, "a serene mountain lake at sunrise, photorealistic, soft mist, golden hour, 8k". Midjourney generates four variants in a grid. I either choose to upscale one variant (high resolution) or generate variations on the same base. I tweak with parameters: ratio, style, character consistency, model version. Once I'm happy, I export a high-definition PNG.

On an average client brief, I have to iterate somewhere between five and fifteen times to land on a finished visual. It's still faster than briefing a photographer, but it's not magic.

The parameters I actually use

A few flags I know by heart: --ar 16:9 for the aspect ratio (the default is 1:1, but I almost always switch to 16:9 for the web), --v 6 or --v 7 for the model version, --style raw for less artistic interpretation and more fidelity to the prompt, --no <term> to exclude stray elements, --seed <number> for reproducibility of a given render, and --cref for character consistency across multiple images.

The engine evolves regularly and the parameters change between versions. I check the official documentation every time a major version drops — otherwise I waste time with obsolete syntax.

access plan and the public-mode trap

Midjourney runs on a monthly access plan with four tiers: Basic (a few hours of fast GPU per month), Standard (regular use), Pro (private mode + unlimited GPU in relax mode), Mega (intensive use).

The point I systematically hammer home with my clients: by default, the images you generate are publicly visible on the Midjourney feed. Private mode (Stealth) is only included starting at the Pro tier. For any confidential commercial use — a client brief, a strategic project, a competitively sensitive job — Pro is non-negotiable. I've watched marketing campaign images show up on the public feed before the official launch at a client who'd stayed on Standard.

Commercial usage rights: what I check

This is a crucial point that's often misunderstood. The current terms — which you should verify on the Midjourney Terms of Service — break down roughly as follows: paying users generally hold commercial usage rights on the images they generate. Very large companies (above a certain revenue threshold) must subscribe to the Pro plan or higher. Images generated on a open access account carry no commercial rights.

As for legal protection of the work, it's the same rule as with DALL-E: in the United States, the Copyright Office considers that a purely AI-generated image, with no significant creative human intervention, isn't protectable. For clients who want a protected work, I advise a human edit after generation.

Where I reach for Midjourney first

From my own practice, Midjourney shines for. Concept art: moodboards, visual exploration during the design phase. Editorial illustration: article visuals, social posts, YouTube thumbnails — this is my main use. Advertising storyboards: quickly presenting ideas in a client pitch. Architecture and interiors: visualizing atmospheres for real estate proposals. Fashion and e-commerce: moods, lookbooks, exploring product ranges before a photo shoot. Visual identity: exploring artistic directions ahead of a brand style guide.

Limits I still run into

A few caveats I bring up with every client. There's no official public API open to everyone: automated use through Discord/web stays manual — a blocker for my batch production pipelines. Generating legible text inside the image (typography, logos) is still imperfect — on this front, DALL-E 3 and Ideogram do better. Hands, fingers, and complex anatomy still occasionally show flaws you have to retouch. Images are public by default outside Pro. And there's no clear opt-out for artists whose style may have inspired the model (a contentious topic across the whole market, not specific to Midjourney).

Alternatives I keep on hand

Depending on the need. DALL-E 3 (OpenAI via ChatGPT Plus) for the best interpretation of complex natural-language prompts. Adobe Firefly, trained on licensed Adobe Stock — clear usage rights, indispensable for clients who want a strong legal guarantee. Stable Diffusion or FLUX (Black Forest Labs) for open source and on-premise deployment when confidentiality is non-negotiable. Ideogram is excellent at text embedded in images. Leonardo.ai for its professional interface and fine-grained controls.

My read for Trust-Vault

By my Trust Score methodology, Midjourney has a mixed profile. Reliability: visual quality recognized by a large part of the creative industry. Transparency: public access conditions, accessible documentation. Privacy: requires the Pro tier for prompt confidentiality — a major point of attention. Commercial compliance: usage rights are clarified but should be re-read depending on company size.

For a creative team in an agency or a startup, Midjourney remains a solid choice. For heavily regulated contexts (strict copyright, industrial secrecy, protected content), I'd lean toward Adobe Firefly for the contractual guarantees, or FLUX for full control over the data.

--- Sources: Midjourney Terms of Service; US Copyright Office — Policy on AI-generated works 2023; Adobe Firefly licensing documentation; Black Forest Labs FLUX release notes.

Further reading

For a complementary implementation angle, read AI Tools for SMEs: The Stack I Actually Deploy in 2026.

For a complementary implementation angle, read Prompt Engineering: The Techniques I Actually Use Daily.