Alie Jules Profile picture
May 7, 2023 11 tweets 9 min read Read on X
Prompt study - breakdown.

Starting with the same shell of a prompt I used in the last couple of threads (see previous tweets).

Let's break down the prompt and see the results.

Thread. Image
This is the initial prompt I started with:

High fashion portrait shoot of a supermodel, in the style of curved mirrors, symmetrical lines, neoclassical style, ultra-realistic, hyper-detailed --ar 16:9

These are simply my observations. It doesn't mean this is the right way. Example of the prompt use f...
Stripping down the prompt:

I took out curved mirrors, neoclassical style.

I added "in Iceland" (random). Add anything you'd like.

New prompt in ALT. High fashion portrait shoot...ImageImageImage
Taking out another layer.

Removed ultra-realistic and hyper-detailed.

The prompt now:
High fashion portrait shoot of a supermodel in Iceland, symmetrical --ar 16:9

The girl in white looks cold. 😆 ImageImageImageImage
Next:

Removed "High" - from fashion.

The prompt now:
A fashion portrait shoot of a supermodel in Iceland, symmetrical --ar 16:9 ImageImageImageImage
Then,

Removed "Fashion" too.

Prompt now:
A portrait shoot of a supermodel in Iceland, symmetrical --ar 16:9

While I peel these layers I pay attention to what differences I see (if any).

I really like the first image. ImageImageImageImage
At this point, I wanted to see if changing the "shoot" to a "photo" would make a difference.

Note! If you do not add either - the results vary (often not photos).

Here are the results and the prompt.
a portrait photo of a supermodel in Iceland, symmetrical --ar 16:9 ImageImageImageImage
Note about "symmetry"

Symmetry seems to center the subject and, well, make the scene symmetrical.

If you remove symmetry, you get more "organic" images.

Using symmetry sometimes make a line in the middle.

Prompt:
a portrait photo of a supermodel in Iceland --ar 16:9 ImageHigh fashion portrait shoot...ImageImage
At this point, the prompt is just about as stripped as possible.

Now I can play with some other elements. a portrait shoot of a young...a portrait shoot of a count...a portrait shoot of a caref...a portrait shoot of a caref...
A few more variations.

If you take out "portrait" and insert something else - you get something completely different - more on that soon.

Also, the aspect ratio changes things. To keep things consistent, I used --ar 16:9 aspect ratio throughout.

Prompt in ALT.

#midjourneyv51 a portrait shoot of a young...a portrait shoot of a young...a portrait shoot of a young...a portrait shoot of a young...
This is an exploration.

It's not to say that this is the right way or the only way. This is simply a method I often use to help me observe what certain elements might be doing in a prompt and whether they are taken into consideration by Midjourney, and how, if at all. Image

• • •

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

Keep Current with Alie Jules

Alie Jules 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 @aliejules

May 18
Midjourney Style Reference (sref) explained

--sref codes are seeds - style reference seeds

Sref code allows you to replicate the style of that seed when you use the same --sref code with various prompts.

Why is this significant?
Using an sref makes is super easy to have a consistent style across any images.

It can be much more tedious and challenging to create this consistent style with prompting alone.

moreImage
There are 4.2+ billion srefs.

Sref seeds go from zero and up to 4294967295.

Explanation of why this top number here:


Many, many of these seeds are very similar, meaning we don't need to know/have them all.

Which is good because it would take me way, way, waaaaay, past my lifespan to go through them all. 😅
Srefs:

- Can't be combined together (possible in the future)

- Can't be used together with style you apply from an image - see screenshots

First screenshot
Add a style using the generated image (click 'style' at the bottom right).
Use the same prompt (click 'prompt' at the bottom right).
Generate the image grid.

Second screenshot
The generated image grid.

The sref 472 (in this case) as been removed and only the image style ref remains - if you look at the parameters on the right.

Not bad results, regardless!!Image
Image
Read 7 tweets
May 18
Place your subject/model in any setting - easy!

RenderNet has added a Character Lab to their suite of features making it super easy to create a character and then consistently place that character in any setting.

This is how. Image
1. Go to

2. Choose or create a character
- I chose Alice rendernet.ai/app
Image
3. Use an existing Canvas or create a new one
- I created a new one Image
Read 5 tweets
Apr 2
You can now edit your DALL·E images in ChatGPT

It's basically inpainting/generative fill.

This is how to do it 👇 Image
1. Click on your ChatGPT generated image to open the editor. Image
2. Click the icon to open the selection tool.

Once the tool is open you can adjust the size of the tool tip. Image
Read 10 tweets
Mar 17
tile parameter in Midjourney - Tutorial

Including prompts.
👇Image
When to use tile parameter?

When you want to create a seamless pattern.

Note. Using 'seamless pattern' in the prompt does not generated a 'true' seamless pattern. It may look like the pattern is seamless, but when lining up several images you'll notice that they do not create a seamless pattern.

Example:
Here I used 'seamless pattern' but not the tile parameter.

Prompt:
retro, yellow and orange, 70's patternImage
Image
How to use tile parameter

Add --tile at the end of your prompt

You can use --tile both on Discord and on Alpha Website.

Prompt:
retro abstract, circles and oval spears, lime green and aqua --tile --stylize 250Image
Image
Read 7 tweets
Oct 20, 2023
AI Image Upscaler Comparison

Paid and Free.

Let's compare.

Closeup of the original MJ image. Image
1) 2x Midjourney Upscaler Image
2) 2x with Photoshop Image
Read 13 tweets
Oct 11, 2023
Midjourney vs. Dalle-3 vs. Adobe Firefly

Same prompt for all.

10 spicy examples.

1. Image
2. Image
3. Image
Read 11 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!

:(