Nick St. Pierre Profile picture
Jun 25 10 tweets 4 min read Twitter logo Read on Twitter
I've been experimenting w/ a bunch of zoom-out workflows in Midjourney.

I've found adjusting the aspect ratio with "Custom Zoom" while keeping the zoom set to 1 give you really nice control.

Here's a super simple prompt + a lil' step-by-step workflow you can use to test it out.
Base prompt to get you started. Super simple, just one word and an emoji.

Prompt: macro {🐝}

I used the bee emoji, but you can use any emoji you want. The insect emojis are great for this, but have fun with it. The eye emoji is really cool too 👁️
Upscale your favorite image & hit "Custom Zoom"

Here you can change your prompt, adjust the aspect ratio, & set the zoom level.

For your first one, keep zoom at 2 and adjust the --ar from 1:1 to something wide, like 16:9.

Then write a prompt to fill the new space with.
I prompted a "a field full of colorful flowers" to get some nice contrast against the white flower & the bee to make it pop more.

By setting a wide aspect ratio, we were able expand the sides much more than the top and bottom.

Next we'll adjust to reimagine the top/bottom
Custom zoom again, changing aspect ratio from wide back to portrait. But this time, set the zoom to 1.

Setting the zoom to 1 helps us expand the top/bottom of our frame without our subject getting super far away.

Since we're expanding the top, I prompted for a "sunny blue sky"
Here's the image. Sunny blue sky & the bee is still very clear & visible from keeping that zoom at 1.

So far we've ran 3 prompts:
1️⃣ macro 🐝 --ar 1:1
2️⃣ a field full of colorful flowers --ar 16:9 --zoom 2
3️⃣ a field full of colorful flowers, sunny blue sky --ar 4:5 --zoom 1
I did one last custom zoom out to bring it back to landscape, keeping the zoom at 1 so I didn't get too far away from my lil' bee.

The prompt I ran was "a field full of colorful flowers, sunny blue sky, butterfly --ar 3:2 --zoom 1"

Now our lil' bee has a lil' butterfly friend
You can keep going w/ this as much as you'd like.

Adjusting the --ar lets you separate the sides from the top/bottom so you can apply unique prompts to each.

Keeping zoom at 1 prevents your subject from getting super tiny while expanding.

Have fun, & follow me @nickfloats
@BitBuster01 i gotchu!
@Kieran_Does_AI_ inpainting is being worked on and might get released before v6, so in just a few weeks. it will be limited due to discord, but it will be a start. they may send you to the web when you inpaint so you can do more with it, but the ux would suck. its coming tho

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Nick St. Pierre

Nick St. Pierre Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @nickfloats

Jun 27
I asked ChatGPT "What does love look like?"

Then, I prompted Midjourney with its responses.

So let's /imagine {"What does love look like?"}

Here are 9 results.
"It's a person setting aside their own comfort to care for a sick loved one, their actions a testament to the selflessness of love."
"It's the tearful pride of a father watching his child take their first steps."
Read 11 tweets
Jun 26
Zoom-out really does give you far more control over certain details that would be otherwise difficult to fully capture in a single prompt.

Step-by-step breakdown of this image, from the first prompt to the final zoom out
Initial prompt, an extreme closeup where I can specify the facial features in much greater detail.

💬extreme closeup, kodak gold, intricate detail, face of a woman, auburn hair, blush cheeks, blue eyes, sly grin, freckles --v 5.2
First zoom out, adjusting the aspect ratio to ar 9:16 and prompting for the dress detail. I set the zoom to --zoom 2

Prompt in ALT
Read 7 tweets
Jun 25
I feel like rly good text-to-3D scene will be available 4 a mass audience much sooner than we think, & I think Luma Labs will probably get there first.

Luma was actually the AI tool that captured my interest most, more than Mj, & got me back on twitter in Feb.

My first NeRF👇
The future is generating full 3D scenes. MJ founder himself has said many times that a common misconception about Mj is that’s it’s a really fast artist, but rly “it’s more like a really slow 3D engine” & “the future isn’t generating 1 img a minute, but 60fps full volumetric 3D”
But when you think about it, Luma is gathering some of the best data in the world to train on.

They have a mobile app with many thousands of users (idk exact number) capturing the world and they’re generating an insane amount of NeRFs every day. Theyre super well positioned here
Read 5 tweets
Jun 23
Use "Custom Zoom" in Midjourney v5.2. It's super powerful.

It lets you re-prompt your zoom out, adjust the aspect ratio, and specify the exact level of zoom you want.

Step-by-step walkthrough of the workflow
First, run your prompt and upscale your favorite image.

Once you upscale, you'll see a few new options are available.

You can "zoom out 2x" "zoom out 1.5x" or "custom zoom"

We are going to be using "Custom Zoom" in this example
Now, click "Custom Zoom"

A window will pop up. Here you can "reprompt" your zoom out, set a custom aspect ratio to zoom to, & adjust the level of zoom

Delete the old prompt & write a new one. I prompted for a garden, adjusted from --ar 4:5 to --ar 3:2, and set --zoom to 1.7
Read 9 tweets
Jun 23
Last v5.2 thing for the night, but this is incredible.

/shorten command then click "show details"

It tells you which tokens are working, which are not, AND their weights.

Showing the token weight is MASSIVE. This is the most insight we've ever had on how MJ works
Look at how much of an impact your syntax has on token weight.

Same prompt, but I put "Venitian night" at front instead of end

Putting it at the front almost 10x'd the weight, plus it read it as a single token rather than splitting Venitian and Night as it did in the first 1
this is going to save so much time testing and playing with syntax by running images. Now you can test with the shorten command, and work your prompt to get the balance of tokens you need before running, saving you $$$ and fast hours
Read 5 tweets
Jun 23
Midjourney just pushed a new version (v 5.2)

Here are 8 side-by-side comparisons using the same prompts & seeds.

I also used the new "zoom out" feature for each prompt. It's honestly a total game-changer, best feature in a long time.

🧵 Prompts in ALTs. v5.1 vs v5.2
1990s street style fujifilm photo of a man, medium slate blue & electric indigo palette with subtle pops of safety orange, wearing gucci apparel on a dark Venetian night --ar 4:5 --seed 76398390

v5.1 (left)
v5.2 (right)

same prompt with custom zoom out to make 16:9

zoom out is everything I've needed. best thing so far
Read 18 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us on Twitter!

:(