MAJOR Shift in Midjourney Prompting + NVIDIA's Releases "The Matrix"

Fam,

We got some big shifts coming.

Looks like MJ V6.1 will be here this month (amongst other things)

No time for chit-chat.

Let’s jump right the hell in.

Here’s what we’re talking about today…

1. Midjourney Mastery.

2. The Major Shift in Midjourney Prompting

3. IMG Prompt vs SREF

4. Style Code Mashing

5. Beyond Midjourney: NVIDIA’s Created “The Matrix”

6. Midjourney Office Hours

Let’s Boogie.

THE MAJOR SHIFT IN MIDJOURNEY PROMPTING:

Until now…Midjourney has been primarily about text prompting…

The words you use…generate the images you want.

Don’t get me wrong…I still think text prompting is the most valuable skill you can develop as a prompter…but…

It’s becoming MORE important to use things like…SREF, IMG PROMPTS, STYLE CODES, CREF, SEED, ETC.

If you know how to leverage them…

You can become increasingly faster…and wayyyy more creative.

HERE’S AN EXAMPLE:

I wanted an image of a Boxer’s entrance to a big fight…

So I started with this text prompt…

Text Prompt Example: Everlast brand portrait photo, full body shot, Boxer facing the camera wearing a black hooded robe, walking down a stadium hallway with posse of men blurred in the background, dark vibey stadium environment with rim lighting and shadows, black trunks, high contrast and tonal balance aesthetic, 35mm film grain, leading lines, ethereal lighting --chaos 15 --ar 3:2 --style raw --stylize 200

THE PROBLEM:

Cool but…I’m asking Midjourney for a lot 

Between the boxer, the environment, the crowd behind him, the overall aesthetic.

I know what I want…but it’s not producing.

So instead of tinkering with text prompts…

Use Image Prompts + SREFs to shape the composition.

THE THOUGHT PROCESS:

Midjourney has defaults…and you need to break them.

Sometimes you need Image Prompts or SREFs to do that.

So consider utilizing an image to shape Midjourney’s imagination.

Final Prompt: https://s.mj.run/5E-c9mfNJPE Everlast brand portrait photo, full body shot, Boxer facing the camera wearing a black hooded
robe, walking down an stadium hallway with posse of men blurred int he background, dark vibey stadium Hallway with Spot
Light and shadows, black trunks, high contrast and tonal balance aesthetic, 35mm film grain, leading lines, ethereal lighting
--ar 3:1 --style raw --sw 150 --stylize 200 --chaos 15 --sref https://s.mj.run/8YkArPghNIU https://s.mj.run/5E-c9mfNJPE

COMING THIS WEEK:

This is a shift in prompting mindset…

And I’ll be doing a longer piece on it this week.

I’ll also be updating MIDJOURNEY MASTERY with videos on this image reference theory.

THE DIFFERENCES IN IMAGE PROMPTING/SREF/SEED:

Midjourney is giving you more options to utilize images as references.

USE THEM.

But most don’t know how/when to use them.

Quick Tips For Usage:

1. Image Prompts: Image Construction + Composition

2. SREFs: Image Aesthetic + Colors + Visual Signature

3. SEED: Image Construction + Subtle Aesthetics

But WTF does that mean…let’s do a quick experiment.

We’ll create a simple prompt…then we’ll use the different image reference techniques to see what they do.

Here’s the prompt:

Prompt: editorial photo of a man holding heirloom tomatoes --ar 3:2

NOW WHAT?

This image is basic…but maybe we want a little more pop.

So let’s use this image below as the reference image to influence the outputs.

REFERENCE IMAGE 👇

KNOWING WHAT TO DO:

Now you have options…but you don’t know which is better.

First, you have to decide what you want to transfer…the image construction…the image style…both…etc.

Let’s look at the differences…here’s the prompts:

Text Prompt: editorial photo of a man holding heirloom tomatoes --ar 3:2

Image Prompt: https://s.mj.run/1ONDJasZyqE editorial photo of a man man holding heirloom tomatoes --ar 5:6

SREF Prompt: editorial photo of a man man holding heirloom tomatoes
--ar 5:6 --sref https://s.mj.run/1ONDJasZyqE

Seed Prompt: editorial photo of a man man holding heirloom tomatoes --ar 5:6 --seed 2493580485

HERE’S THE RESULTS:

THE DIFFERENCES:

IMAGE PROMPTING: The image is very similar, but there are still carrots baked into the image…because image prompting takes elements from the reference image including construction and visual subjects.

SREF: Very good translation from image-to-image…it transfers the visual aesthetic and keeps the visual hierarchy.

SEED: Good translation with less inconsistencies, but it doesn’t really match the aesthetic perfectly.

So they all have pros/cons…but using them together is the real winner.

The main difference is you can control Image Prompts with the image weight (--iw parameter) and SREF with Style Weight (--sw parameter).

(You cannot alter the seed weight)

Combo Prompt: editorial photo of a man holding heirloom tomatoes --ar 3:2 --sref https://s.mj.run/1ONDJasZyqE --sw 50

STAY TUNED:

I’m working on a longer piece for this…and a decision matrix.

Should be coming this week.

STYLE CODE MASHING:

If you just read the last (2) sections…congrats…this one will make sense.

I thought style codes were useless…I was wrong.

I’ve been testing a bunch of these to find good building blocks.

So you can utilize SREF/SEED/IMG Prompts to build images you want/need.

Here’s (11) Unique Style Codes…and how to mash them.

If you missed it…Here’s the FULL LINKEDIN POST.

THE REAL BENEFIT:

This mashing these codes together will give you a distinct style.

Things you would not be able to text prompt for.

Process:
01. Generate Style Codes
02. Create new prompt
03. Remove style code parameter
04. Add images generated with codes as SREFS
05. Generate.

Prompt: Award winning photo of an epic ancient warrior --ar 5:6 --sref https://s.mj.run/_oQ_R2eK4Cc https://s.mj.run/DcSmi2B60QY https://s.mj.run/vty6IvFaU_U

BLENDING STYLE CODES:

Here’s the PLAIN TEXT CODES if you want to try:

--sref 302484108
--sref 3588252871
--sref 2346775807
--sref 1121645525
--sref 3474327402
--sref 2360446404
--sref 3622818906
--sref 3210011585
--sref 3496632656
--sref 1092774870
--sref 1497823114

We’ll keep pushing down this rabbit hole.

BEYOND MIDJOURNEY: NVIDIA IS CREATING “THE SIMULATION”

Alright so…I’m testing out a new section of this newsletter. It’s going “Beyond Midjourney” and into what’s happening in the Ai landscape.

(Now you know).

NVIDIA’s ACE:

Recently, NVIDIA released its “Earth 2” simulation technology to help predict climate and weather patterns.

It’s essentially a simulated world. (But they didn’t stop there)

Now they’re adding Ai into simulated characters.

For all the “Matrix” and “simulation theory” fans out there…this is how it starts.

Our simulations will eventually start to produce simulations of their own…and round-and-round we go.

You have to WATCH THIS to believe it.

Creepy…but not unpredictable.

No wonder they’re worth like a trillion dollars.

Here’s the link to the full NVIDIA KEYNOTE.

MIDJOURNEY OFFICE HOURS (06/05):

Website Updates

• Focusing on improving the website experience for new users

• They don’t want to replace Discord

• But want to provide an easier onboarding experience

• Cleaning up and combining features for a more intuitive interface

Personalization Feature

• Personalization feature is close to release

• 90-95% of testers are happy with the current version

• Aiming for a release next week

• Deciding on using personalization codes

• Personalization will be optional and turned off by default

• Users report that personalized versions feel like a significant improvement, akin to a "Midjourney 6.1" upgrade

Midjourney 6.1 Release

• Release is blocked by pixel quality improvements,

• Hands, bodies, and text enhancements are ready

• Might release a version w/o the pixel quality to avoid further delays

• Aiming to release some form of improvement this month

• The combination of personalization and 6.1 features will make Midjourney V6 feel significantly different and more serious

Have fun.

If you need help…DM me on Linkedin.

Until next time…

Rory