Last month I was sitting in my study, staring at a Discord window with a cup of coffee that had already gone cold.
A client had just emailed me. They wanted visual concept images — "Eastern aesthetics, futuristic vibe." Tight budget, but had to look premium immediately.
I opened Midjourney, typed the prompt, uploaded three Song Dynasty landscape paintings as references.
By the fifth image, I already knew most of the batch would work.
The lighting transitions were smooth. Textures were rich. Cyberpunk elements blended with classical mountains and rivers in a way that felt almost inevitable — like the two styles had always belonged together.
When I reached the twentieth image, I stopped scrolling.
Not because the images weren't good enough.
They were too good.
Good enough that I sat back, looked at the screen, and felt something I didn't expect: unease.
Am I still creating something — or just selecting ready-made assets in bulk?
Midjourney has been with me since v4. It saved me countless outsourcing dollars and helped someone who literally can't draw deliver professional visual concepts to clients.
But v7 was the first time I asked myself a genuinely different question: is this tool helping me — or slowly replacing me?
What actually changed
I spent about half a month with v7. I wasn't testing parameters like --stylize or --chaos. I usually don't bother with that.
I focused on one practical question: is it still worth the subscription?
The biggest improvement is style consistency.
One of the most frustrating things about previous versions was what I started calling "style drift." You generate ten images hoping for a coherent tone. The first looks like cold cyberpunk. The second suddenly feels like steampunk. To get a usable set, you might generate a hundred images and cherry-pick maybe eight.
With v7 and Omni Reference, that problem is dramatically reduced.
I uploaded those same three Song Dynasty paintings with a prompt like "cyberpunk reconstruction, neon mist over misty mountains." Out of 50 generations, about 48 looked like they belonged to the same visual universe. Colors stayed consistent. Brush textures matched. Lighting logic held across the whole set.
My first reaction was pure excitement. For solo creators or small teams, this is enormous. What used to take days of manual tweaking now produces a full proposal set in under an hour.
The creative depth also improved. I ran a deliberately cinematic prompt:
"A lone astronaut on Mars discovers an old CRT television half-buried in red sand. The screen flickers with childhood memories — fading from cold blue to warm yellow, evoking nostalgia and isolation."
In v6, something like this would need 30 to 40 attempts for a single good result. v7 gave me six images. Five of them nailed it — the astronaut's posture felt lonely, the CRT reflection worked, the warm color gradient through Martian dust looked genuinely emotional.
For a moment I felt almost smug. The tool was making me look like a designer with real artistic vision.
Then the feeling shifted.
When efficiency becomes the problem
When the output quality gets this high, something strange happens to your role.
Before, using Midjourney still required real judgment. This image has better lighting. That one has awkward composition. The act of choosing was a form of creative participation — it was where your taste actually lived.
Now v7 regularly gives you a pile of images that are all usable. Out of fifty, picking ten becomes almost mechanical. This one is brighter. This one is cooler in tone. The rest essentially decides itself.
I called a friend who still works in branding and vented about this. He was quiet for a moment, then said: "The tool solves how to draw. But it can't solve why draw it this way. If you let it decide the 'why' too, you're not creating anymore — you're operating a machine."
That landed harder than I expected. Because I realized I hadn't been asking "why" as often lately.
There's a second issue that bothers me more, honestly.
v7 locks visual style extremely well — but it does so by learning from reference images. Which leads to a simple question: if I can replicate this aesthetic with three reference files, so can everyone else.
Your "unique visual style" stops being unique the moment someone runs similar references through the same model. I kept thinking about something Naval Ravikant said — that real wealth is something that generates returns while you sleep. If you have to constantly operate something, it's a job, not an asset.
Generating images with v7 still feels like a job. Building a repeatable visual system might eventually become an asset — but v7 hasn't fully solved that. It locks style well, but it doesn't prevent everyone else from locking the same style.
Which leaves a question I'm not sure how to answer: am I building creative assets, or helping train a tool that eventually makes my aesthetic easier for anyone to copy?
So is it still worth it
For illustrators, concept artists, and visual storytellers, v7 is still genuinely excellent. The lighting, textures, and artistic depth are among the best in AI image generation. The community alone is worth staying for — just browsing other people's work teaches you new ideas every day.
But for brand designers, indie creators, or the "visual person" on a small team, the story gets complicated. v7 is almost too efficient, and that efficiency raises uncomfortable questions about creative identity.
I've noticed some teams quietly moving toward local Flux 2 Pro models or hybrid workflows using GPT Image for editing. Those tools feel stronger for precise control and building long-term assets. Midjourney v7, by contrast, feels like driving a very fast car where the steering is just slightly less responsive than you'd want.
Where I actually landed
I'm still on the fence.
Midjourney has been part of my workflow since v4. A few hundred dollars a year saved me thousands in outsourcing and changed what I could offer clients. That history is real and I don't dismiss it.
But v7 was the first upgrade that made me think seriously about creator identity — not just workflow efficiency.
There's a line I keep coming back to that Luo Yonghao said years ago: tools should liberate you, not define you. v7 absolutely liberated me from how to draw. The part I'm still working out is whether it's quietly starting to answer what to draw and why — and whether that's a problem or just the next stage of how creative work evolves.
I don't have a clean answer. I'm not rushing to renew for another year, but I'm not canceling either.
Maybe v8 will settle it. Or maybe the question will just get harder.