LAST CHANCE + Multi-Step Midjourney Prompting.

Fam,

BIG life changes this week.

My wife gave birth to our baby daughter, Faye.

She’s cute…and has me whipped already.

That being said, I am running on fumes.

But things are happening…so here’s what we’re getting into.

1. Midjourney Mastery: Last Chance for Tier-2 Pricing.

2. Midjourney Private Rooms.

3. A Week of Sleep Deprived Prompting.

4. Multi-Step Midjourney Prompting.

5. Style Code of the Week.

6. ChatGPT 4o Release.

7. Midjourney Office Hours.

Let’s boogie.

MIDJOURNEY PRIVATE ROOMS:

Last week Midjourney released the “Private Rooms” feature.

Meaning…you can now collab with multiple people in one channel.

This is awesome for having fun with your friends.

Even better, you can create shared spaces for teams.

Here’s how to set it up:

SET UP PROCESS:

1. Open Midjourney Alpha Site.

2. Select “Rooms” on navigation bar.

3. Select “Create Private Room.”

4. Add a “Name/Description/Password.”

5. Select “Create Room.”

Now you have your own “Private Space.”

But if you want to invite people…do this…

INVITE PROCESS:

1. Open your room.

2. Copy the room URL.

3. Send URL to friends/co-workers.

4. Make sure to share the password.

This feature was missing from the discord integration.

It’s great if you’re using Midjourney in the office…shared spaces for team members, prompt collaboration, general exploration, etc.

That being said…if you’re a “Private Prompter”…

You can use it to create “Project Channels”

I loved organizing projects in “channels” on Discord…you can use Rooms in the exact same way.

You DO NOT need to invite people to the room…And everything you generate will be added to your create page.

So if you hate scrolling back through your “archive” or “create feed”…

This is a way around it.

A WEEK OF SLEEP-DEPRIVED PROMPTING:

I slept on a hospital chair for most of the week.

In my downtime…I decided to let the sleep deprivation guide my prompting.

(I don’t recommend this as a creative exercise)

But there’s some interesting concepts to play with.

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

TESTING:

Midjourney isn’t fun if you’re not pushing boundaries.

Yes, it’s nice to have structure, but…sometimes you just need to push some buttons and see what happens.

This is what I was playing with this week.

• Using SREF way more

• Text Prompting less

• Multi-step prompting (see below)

• Blending images for more unique SREFs

• Focusing on textures for more image pop

• Finding a practical use for Style Random

Let’s dive into a few of these.

MULTI-STEP PROMPTING:

Q: WTF is multi-step prompting.

A: I made it up…idk what to call it.

Everyone wants a perfect image with one single text prompt. (It’s not that easy)

If you want complex images…sometimes it’s not possible with just text…the tokens and prompt coherence are limited.

So use image references instead. (Image Prompts or SREF)…

And do this in multiple steps.

Let’s look at an example.

I wanted this man in a dragon mask to have a “petrified wood” coloration and texture.

Text Prompt: documentary photography, closeup, Chinese warrior wearing an intricate wooden dragon mask, petrified wooden textures and coloration, with traditional Chinese decoration, sharp contrasting colors, 200mm lens --ar 3:2 --stylize 450

RESULTS + NOTES:

It’s cool…but it doesn’t have petrified wooden texture/coloration.

So instead of playing with your prompt a million times…

Generate the petrified wooden texture separately…then give it to Midjourney as inspiration.

Wood Text Prompt: Chrysocolla petrified wood pattern, intricate texture --ar 3:2

RESULTS + NOTES:

Solid texture and colors…it would be hard to get this level of intricacy for both the WOOD and the MASK…that’s a lot of computation for Midjourney.

So instead of text prompting…mash them together with image prompting and/or SREF.

Multi-Step Prompt: https://s.mj.run/BJkiX8ttgpE https://s.mj.run/SZdk-VnY83I documentary photography,, Chinese warrior wearing in intricate wooden dragon mask, wood texture, with traditional Chinese decoration, sharp contrasting colors, 200mm lens --chaos 15 --ar 3:2 --sref https://s.mj.run/SZdk-VnY83I --sw 30 --stylize 450 --iw 0.7

RESULTS + NOTES:

In this image…I used the “wood” as an image prompt and an SREF to really lock in the aesthetic.

Ultimately, you have understand the tool’s functions to manipulate the images. Things like Image Prompting, Image Weight, SREF, Style Weight, etc.

And you have to test different parameter variables to get your aesthetic.

(I’ll do a longer piece on this…stay tuned)

So think about it like parts of a whole…instead of one prompt.

STYLE CODE OF THE WEEK:

I’m still trying find a practical use for “style random” (it’s not going well)

It’s more fun than functional.

“Random” is the opposite of “control”..so it’s a slot machine.

If you have no idea what i’m talking about…

Add “--sref random” to the end of your prompt and it will generate a random aesthetic.

QUICK TIPS FOR STYLE RANDOM:

1. Use shorter prompts so you can see the “style”

2. Longer intricate prompts = less style identification

3. If you like a style…use it as a SREF on other prompts

4. Or use the “style code” to consistent gen in that style.

Weekly Winner: --sref 3690843001

I tested this one on a few car models…and it produces with minimal effort.

Prompt Structure: [car model] --sref 3690843001 --s 250 --c 5

Best thing about the codes…if you find one you like…

You can transfer it to any image.

Prompt: Cinematic Samurai --sref 3690843001 --s 250 --c 5

I’ll start dropping some here weekly.

Stay Tuned.

CHATGPT 4o RELEASE:

Unless you live under a rock…you know ChatGPT released some big updates.

Here’s the standard talking points:

• OpenAI released GPT-4.0 (GPT-4 Omni),

• Multimodal model that can handle audio, video, images, and text.

• GPT-4.0 is faster, more capable, and available for free ChatGPT users with certain limitations compared to paid users.

• GPT-4.0 API is 2x faster, 50% cheaper, and has a 5x higher rate limit for developers.

• New ChatGPT desktop app allows screen sharing and voice interaction, with emotive and expressive responses.

• GPT-4.0's image generation capabilities were not featured in the keynote but are impressive, including image editing, 3D object synthesis, and font generation.

Here’s some of the more creative things people are doing:

• Creating consistent character images across different scenes

• Generating commemorative coins and animating them in 3D

• Transferring logos and designs onto objects in images

• Producing entire font sets based on text descriptions

• Creating realistic 3D object renderings from multiple angles and animating them

• Placing brand logos realistically within generated images

Matt Wolfe does a great job of breaking it all down.

MIDJOURNEY OFFICE HOURS (5/15):

Midjourney Version 6.1 Release

• Dramatic improvement in image quality, texture, coherence, and text rendering

• Default aesthetics will remain similar to version 6.0

• Significant update, but not a complete overhaul like version 7

• Debating whether to include personalization features in 6.1 or release separately

Personalization Feature

• Optional mode that personalizes model biases based on user preferences

• Learns user biases over time to generate more tailored images

• Initial release will be simple, with one personalization style per user that can be shared

• Conducting surveys and data analysis post-launch to understand connections between aesthetics, demographics, and user preferences

Midjourney Version 7 Development

• Increased knowledge compared to version 6, with more data

• Improved prompting capabilities, handling multiple objects without entanglement

The Future of Creative Mediums

• Midjourney is not just replicating existing art forms like photography or filmmaking

• The goal is to create a new medium that culminates in simulated worlds, or "Sims"

• Sims represent the next step in creative mediums, beyond passive movies and interactive video games

• In Sims, there is no hard distinction between creator and player, allowing for a more immersive and collaborative experience

• Focusing on developing this final creative medium rather than iterating on intermediate steps

Have fun.

If you need help…DM me on Linkedin.

Until next time…

RF