I’ve been using Midjourney since V6, and the biggest change in V8 is pretty simple: it’s faster, and it follows instructions better.
Ease of use: very low friction.
These days, you can just log in on the web and start generating. There’s no need to install models, set up a GPU, or configure a local environment.
You type what you want, wait for the results, choose an image, and keep refining from there.
Chinese prompts work too, although for more complicated requests, English still tends to be more reliable.
What the results are like:
V8 is noticeably faster than earlier versions, and it does a better job of following small details in a prompt.
For example, I used to ask for:
“A cat wearing a straw hat with a red ribbon tied around the hat.”
Older versions would sometimes put the hat next to the cat instead.
V8 is much better at understanding those relationships between objects, positions, and attributes.
The image quality is also more solid overall. For concept art, moodboards, and visual pitches, Midjourney still produces very polished results with relatively little effort.
A few issues stood out in practice:
- It’s more reliable, but a little less surprising
V8 is more accurate and more professional-looking, but some of the randomness that longtime users liked is weaker. V6 would occasionally produce something completely unexpected but strangely compelling. V8 feels more controlled.
- Old prompts don’t always carry over well
Major version updates can change how parameters, prompts, and style techniques behave. Longtime users may need to relearn parts of their workflow, which is a real cost.
- Mode support can change between versions
During testing periods for new models, some generation modes or plan-related features may be adjusted temporarily. Heavy users should also keep an eye on actual Fast GPU usage.
- Text is still a weak spot
Short English words have improved, but longer English text and Chinese typography still make mistakes. If a poster needs exact copy, you’ll usually need to fix the text afterward.
What works well
- High default visual quality. You don’t need elaborate prompts to get something that looks polished enough to use or refine.
- Low learning curve. There’s no need to understand models, nodes, samplers, or other technical concepts.
- Fast generation. Great for exploring directions quickly and producing lots of ideas.
- Mature community. There are plenty of tutorials, examples, style references, and prompt resources.
What doesn’t work as well
- Precise control is still limited. It handles overall mood very well, but details like poses, object counts, and exact positioning can still drift.
- Version upgrades can change the feel. Existing prompts and habits don’t always transfer cleanly to new models.
- Commercial copyright remains uncertain. Eligible paid users can use generated work commercially under Midjourney’s terms, but broader questions around AI training data and copyright are still disputed.
- There’s no real long-term free tier. If you want to keep using it, you’ll basically need a paid plan.
Best for
Designers and visual creators
Great for concept art, style exploration, and mood references.
Marketing, planning, and content teams
Useful for social visuals, campaign concepts, and early-stage creative drafts.
Users without a technical background
You can start by typing a prompt without learning a complex workflow.
Work where visual impact matters more than absolute precision
For example, game concepts, film storyboards, and personal art projects.
Not ideal for
Commercial work that requires precise control
Product imagery, consistent characters, or strict composition usually still need other tools in the workflow.
Heavy users on a tight budget
At $30 per month, Standard isn’t especially cheap for someone using it purely for fun.
People who want a stable workflow for years
Midjourney evolves quickly, and new versions can change both prompt behavior and generation habits.
Comments (0)