Why Try AI

Share this post

Midjourney Version 5.1: Back to Basics

www.whytryai.com

Discover more from Why Try AI

What AI can do for you. Yes, you! Hands-on, no-hype look at generative AI for enthusiasts.
Over 3,000 subscribers
Continue reading
Sign in

Midjourney Version 5.1: Back to Basics

The new "opinionated" Midjourney Version 5.1 combines Version 4's ease of use with Version 5's higher image quality. AI art beginners, rejoice!

Daniel Nest
May 4, 2023
12
Share this post

Midjourney Version 5.1: Back to Basics

www.whytryai.com
11
Share

When Midjourney Version 5 first came out in mid-March, the team was clear that it was a raw version without any “secret sauce” sprinkled in.

They called it “unopinionated.”

Well, Version 5.1. is here, and can you bet it’s got a few opinions!

Here’s the official launch summary from Midjourney CEO David Holz:

Screenshot of V5.1 announcement on Discord by David Holz

To boil all of this down to two main points, we can expect V5.1 to:

  • Return higher quality images in general

  • Make even basic prompts look pretty

Below are a few preliminary observations from my limited testing.

Why Try AI is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

📸V5 = best for photos 🎨V5.1 = best for art

Right off the bat, it’s obvious that while V5 skews photographic, V5.1 is a lot more artsy. (Just like V4 was.)

Here’s a “man riding a horse”:

Midjourney V5 results for "man riding a horse" - 4-image grid
Midjourney V5
Midjourney V5.1 results for "man riding a horse" - 4-image grid
Midjourney V5.1

All four V5 results are photos. The V5.1 grid has a more painterly feel.

But it’s more than just “photo” vs. “painting.” The other aspect is…

😐V5 = bland 👹V5.1 = dramatic

Put simply, V5 usually does the bare minimum to fulfill your requirements, while V5.1. has a flair for the dramatic.

Ask them for “modern warfare,” and you get:

Midjourney V5 results for "modern warfare" - 4-image grid
Midjourney V5
Midjourney V5.1 results for "modern warfare" - 4-image grid
Midjourney V5.1

It’s hard to argue that V5 photos don’t live up to the “modern warfare” spec. They do. But most of them feel almost like sterilized stock images.

With V5.1, stuff. Is. Happening!

You want explosions? We’ve got explosions!

Want fire? Let’s burn it all down!

Need a helicopter? Here are at least, like, three of them!

So if you’re looking to tell a story and breathe life into images, V5.1 is your friend.

🪄 Vague / abstract prompts are best with V5.1

The intrinsic drama and artistic touch of V5.1 makes it more entertaining to use with random or less specific prompts.

Back in V4 days, I used to love using song titles as prompts or experimenting with different emoji combinations.

With V5, some of that magic is gone. Emoji combo like 🌈🐷will just give you this:

Midjourney V5 results for the pig + rainbow emoji combo. 4-image grid.
“Rainbow, check. Pig, check. What more do you want?”

But do the same in V5.1 and boy, now we’re talking:

Midjourney V5.1 results for the pig + rainbow emoji combo. 4-image grid.
“This! This is what I want. Even the sheep-pig hybrid is incredible!”

This even goes for completely mundane terms.

I accidentally asked Midjourney to imagine the word “settings” instead of using the /settings command (true story), and V5 gave me this:

Midjourney V5 results for the word "settings" -  4-image grid.
Those sure are “settings” of some sort.

I then used “settings” in V5.1 and got a whole bag of fantastic stuff:

Midjourney V5.1 results for the word "settings" -  4-image grid.
Are these “settings”?! Do I care?! No!

In short, Midjourney V5.1 is simply more fun if you’re not looking for precision.

📝 Prompt accuracy: Inconclusive

I wasn’t quite sure what David Holz meant by “more accuracy to text prompts,” but I wanted to test it out. My hypothesis was that MJ V5.1 might be better at composing scenes with multiple different subjects.

So I tested this out with “one red ball and two green cubes on a blue table”:

Midjourney V5 results for "one red ball and two green cubes on a blue table" -  4-image grid.
Midjourney V5
Midjourney V5.1 results for "one red ball and two green cubes on a blue table" -  4-image grid.
Midjourney V5.1

I can’t award any clear points to V5.1 here. If anything, Version 5 is actually better in at least rendering both the “red ball” and the “blue table” more consistently.

Let’s try something else, like “kids jumping on a trampoline made of marshmallows”:

Midjourney V5 results for "kids jumping on a trampoline made of marshmallows" -  4-image grid.
Midjourney V5
Midjourney V5.1 results for "kids jumping on a trampoline made of marshmallows" -  4-image grid.
Midjourney V5.1

No obvious winner here, either.

V5.1 renders marshmallows more accurately, but V5 respects both the “kids” and “trampoline” terms in every picture.

Maybe I’m testing for the wrong thing?

As it stands, I’d have to give a slight edge to V5 for better following directions.

Use the right version for the right purpose

One thing we can conclude quite definitively: In the absence of specific prompt descriptors, V5.1 is more artistic than raw V5.

I think it’s great to have two tools that can be used for different things.

  • Midjourney V5 is perfect for photo prompts, creating realistic stock images, or situations where you need a high degree of control over the end result.

  • Midjourney V5.1 is great for art, fantasy concepts, storytelling, or just having some good old fashioned fun with less predictable but awesome-looking results.

To drive the above point home, I’ve added V5.1 to my V1-V5 comparison for the terms “forest hut,” “hamster photo,” and “hoverboard”:

Midjourney V5 results for "forest hut," "hamster photo," and "hoverboard"Midjourney V5 results for "forest hut," "hamster photo," and "hoverboard"Midjourney V5 results for "forest hut," "hamster photo," and "hoverboard"
Midjourney V5
Midjourney V5.1 results for "forest hut," "hamster photo," and "hoverboard"Midjourney V5.1 results for "forest hut," "hamster photo," and "hoverboard"Midjourney V5.1 results for "forest hut," "hamster photo," and "hoverboard"
Midjourney V 5.1

There’s just more life in the V5.1 images, isn’t there?

How to enable Version 5.1

You’ve already done it!

Yup, if you’re reading this after May 4, 2023, Version 5.1 is now the default mode:

David Holz announcement about V5.1 becoming default

If you’ve previously set up another version via /settings, you can switch to V5.1 using the same process:

Settings menu in Midjourney Discord with V5.1 enabled

Have fun and enjoy those epic scenes!

Why Try AI is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Over to you…

Did you do more to test the “text accuracy” claim? Have you found version 5.1 to be better at interpreting input than V5? I’d love to see some examples!

Feel free to share any other general observations about Midjourney V 5.1. You can leave a comment on the site or shoot me an email (reply to this one).

12
Share this post

Midjourney Version 5.1: Back to Basics

www.whytryai.com
11
Share
Previous
Next
11 Comments
Share this discussion

Midjourney Version 5.1: Back to Basics

www.whytryai.com
Charlie Guo
Writes Artificial Ignorance
May 5Liked by Daniel Nest

Wow, that rainbow pig example is wild. Somehow it never occurred to me to use emojis in my prompts.

Expand full comment
Reply
Share
1 reply by Daniel Nest
Gordon Mickel
Writes Byte-sized Brainwaves
May 5Liked by Daniel Nest

Great writeup!

I've been having success using Midjourney Version 5.1 with the new "Raw" setting when I'm going for photo-realism. My initial - very unscientific - tests suggest that the images seem to be slightly better than with 5.0 while not being overly stylised.

Expand full comment
Reply
Share
1 reply by Daniel Nest
9 more comments...
Top
New
Community

No posts

Ready for more?

© 2023 Daniel Gniazdo
Privacy ∙ Terms ∙ Collection notice
Start WritingGet the app
Substack is the home for great writing