BIG MIDJOURNEY UPDATES + LUMA's Huge Splash

Fam,

Well…that was a crazy week in Ai.

We seemingly had a period of silence…and then…

Everything explodes…that was this week.

Here’s what we got for today:

1. Midjourney’s New Personalization Feature

2. Midjourney Website Updates

3. Partnership Announcement

4. Luma’s Big Splash

5. Style Code Moodboards

6. Beyond Midjourney: Apple Intelligence

7. Midjourney Office Hours

Let’s Boogie.

MIDJOURENY’S NEW PERSONALIZATION FEATURE:

I was curious about how this would turn out.

After a fun runs…I’M SOLD.

Let’s take a look…

PERSONAL STYLE:

WTF is it?

This is Midjourney’s attempt to make your outputs more personal.

So instead of relying on MJ’s default aesthetic…It takes your image preferences into account and tailors it “your style.”

Here’s how to use it…

WHAT YOU NEED:

Your personalized style is based on the images you’ve rated.

If you have no idea WTF im talking about (keep reading)…

• Rate 200+ images for it to work

• Type “/info” on Discord to see your ratings #

• Or go to “Tasks” → “Rate Images” to see #

HOW TO RATE IMAGES

1. Go to the alpha site → Tasks → Rank Images

2. Rank the image pairs you are served.

3. Remember, this affects your personalized style.

So be cognizant of what you’re rating…and skip things you don’t like.

BONUS: The more you rate…the more free fast-hours you’ll get.

HOW TO USE PERSONALIZED STYLE:

1. Create a prompt

2. Add (--p) as a parameter

3. Hit submit

4. MJ will provide a custom “code”

5. Use this code for other images.

6. Adjust weight with (--s) parameter.

Example Input: freeze motion, editorial closeup of a Land Rover defender in vibrant Tokyo streets, chrome and leading lines composition, dappled sunlight --ar 4:5 --seed 856747510 --p

Example Output: freeze motion, editorial closeup of a Land Rover defender in vibrant Tokyo streets, chrome and leading lines composition, dappled sunlight --ar 4:5 --seed 856747510 --personalize lhg6k14

HOW TO MANIPULATE PERSONALIZED STYLE:

You can also control the strength of the personalized style.

**Disclaimer - This can get confusing because you use the “stylize” parameter to control the strength of the personalization influence.

• Use (--s) parameter to adjust weight

• --s 0 = OFF

• --s 100 = Default

• --s 1000 = Max

Prompt: cinematic still, Ancient tribe ritual, Mayan witch doctor wearing intricate ceremonial attire, performing an ancient ritual in the Yucatán jungle, turquoise and blacks, light rays and shadows, smoke interacting with light, --ar 4:5 --personalize lhg6k14 --stylize {0, 50, 100, 250, 400, 550, 700, 850, 1000}

QUICK THOUGHTS:

The personalization is not a static feature…it will change and adapt as you do more pair ratings. So it will be in constant motion

Meaning…rank images with intention because it will affect your personal style.

Also, your style might not be that different from the default…try rating more images and really honing in on what you like.

MIDJOURENY’S NEW WEBSITE UPDATES:

A few quick changes to some fan favorites.

Some simple…some a little confusing.

Here’s what you need to know:

WEBSITE UPDATES:

1. 'Pan' + 'Zoom' tools are now called 'Reframe'

2. 'Remix' is now found within 'Reframe' as ‘Edit Prompt’

3. 'Vary Region' is now called 'Repaint'

THINGS TO NOTE:

If you like to REMIX…remix subtle and remix strong are gone.

It’s just “edit prompt” now…and it’s not the same as it was.

REMIX + IMG PROMPTS/SREF/CREF:

If you like the add image prompts/SREFs/CREFs to your remixes…

You need to add them the same way discord.

• IMG Prompt → Place link at beginning of prompt

• SREF → Use as parameter [--sref (img link)]

• CREF → Use as parameter [--cref (img link)]

If you want to nerd out on this…

Drew Brucker and I did a LINKEDIN LIVE covering Personalized style + new website updates…here’s the LINK.

NEW PARTNERSHIP: SUPERSIDE X RORY FLYNN

I’ve been keeping this quiet for a while…but we’re making it official.

I’m partnering up with Superside.

To push the boundaries of AI x DESIGN.

Check it out HERE: https://www.superside.com/roryflynn

WHY ARE WE DOING THIS:

Last year, Phil Maggs slid into my DMs. (awesome guy btw)…We had a seemingly harmless chat that turned into a 2-hour nerd-out session.

It meshed. (almost too well)

SO WE COLLABORATED ON IDEAS ABOUT:

• Solving simple creative issues with Ai.

• Building practical workflows for creative teams

• Training 100’s of team-members to use the tools

• Creating SOPs that make operations simple + efficient.

And then...we IMPLEMENTED them.

CREATING AI-ENHANCED WORKFLOWS FOR:

• Static Ad Creative + Motion Graphics

• Custom Model Training + Brand Asset Expansion

• Rapid Branding + Concept Development 

• To name a few…

This was one of them:

MOST IMPORTANTLY…

These ideas needed to be backed by DATA. (Not hopes and dreams)

Soooo for 12+ months we tested systems, processes, and tools to find the sweet spots. And make these workflows consumer-ready.

For clients like Amazon, Reddit...and Salesforce.

THE RESULT?:

• Across 500+ client projects…

• 200% faster completion from brief-to-delivery

• 40-85% reduction in hours spent.

• Saving clients over $1.4 Million.

This is what happens when you have a team of world-class traditional designers and you put this type of power in their hands.

So I'll be working alongside the team at Superside to

• ANALYZE current marketing and design problems 

• R&D Ai solutions and workflows

• And…SOLVE PROBLEMS.

With a team of 700+.

If you have creative needs or questions...SHOOT ME AN EMAIL »[email protected]«

Or peep the landing page HERE: https://www.superside.com/roryflynn

No question is out of bounds…no project is too big or small.

LUMA’S BIG SPLASH

Did you even “social media” this week…If you didn’t try Luma?

Luma’s new image-to-video is cool.

But, as usual…There’s hype-value attached.

Check out FULL LINKEDIN POST HERE

DECENT & EXCITING, BUT…

We got so juiced by Sora’s possibilities that anything remotely close will satisfy.

I tested a bunch of images And it’s good…but is that because it’s new/different?

I guess we’ll see…here’s how to use it…

PROCESS:

1. Go to LUMALABS.AI

2. Sign up with your email.

3. Add and image reference and/or text prompt

4. Hit generate.

Pretty easy…let’s be honest about it though…and look at some PROs and CONs.

LUMA PROS:

• Unique output

• Creative camera motion/stability

• (Somewhat) realistic object motion

• Stays true to image input

• Prompt coherence is decent/good

• Complex motion much better than others

• 10 Free generations/day

LUMA CONS:

• You lack control over output

• Slot machine generations

• Prompt heavily influences output (also a pro)

• Object permanence is inconsistent

• Long queue times (really long)

• Watermarked downloads

PROMPTING TIPS:

• Be detailed

• Describe the shot and motion

• Ex (orbing shot, panning left, etc)

• Describe characters actions/motion

• Ex (helicopter flying toward camera, etc)

LUMA PROMPT EXAMPLES FROM VIDEO:

SHOT 1: cinematic orbiting shot of a helicopter flying towards and exploding volcano, smooth camera motion

SHOT 2: cinematic panning shot, volcanologist with the reflection of a helicopter in his visor, exploding volcano in the background, dramatic scene

SHOT 3: cinematic fast zoom, volcanologists working, massive exploding volcano in the background, chaos, dramatic scene

SHOT 4: cinematic shot, helicopter flying towards a volcanologist waiting to exit the situation, ethereal lighting, dramatic moment

THOUGHTS:

This has the most complex motion from a tool we’ve seen and it’s powerful…but overall it lacks control.

As usual…this tool is new…and needs time to develop…we’ll see where it goes.

MIDJOURNEY STYE CODE MOOD BOARDS:

I know Midjourney style codes feels like 6 months ago

But they’re still useful and kinda awesome.

So here’s a quick tip:

Tip: Use mood board prompts to understand your style codes look & feel.

MOOD BOARDS + STYLE CODES:

Instead of running a ton of gens…and wasting countless fast hours.

Use mood board prompts to generate or test your style code.

PROMT STRUCTURE: multi-panel mood board, split into multiple images, including lifestyle images and color swatches --ar 5:6 --sref [random or your code].

BEYOND MIDJOURNEY: APPLE INTELLIGENCE

Well we got the big introduction…and it’s well…(i’ll let you be the judge).

IDK how I feel about a company having that much access to my data.

It goes beyond texts/emails…they have access to health data/credit cards/location/etc…it sounds cool in theory…but i’ve watched too many science fiction movies.

Matt Wolfe does an awesome job of breaking it down…

Again…I’ll let you be the judge.

Even thinking about icloud, connected devices, photos, documents…this is the first time i’m uneasy.

If you want to watch APPLE’s FULL KEYNOTE here’s the link:

MIDJOURNEY OFFICE HOURS (5/29):

Recent Releases

• Personalization feature - Discussed above!

Upcoming Releases

• Combining style codes and personalization codes for easier blending

• Hopefully releasing next week

• Updating personalization more rapidly with user actions

6.1 or 6.5 version

• Final training run in-progess (will take at least 2-4 weeks)

• Testing a lot of V7 features in V6

• Potential upscaler update

V7 version

• Not in final training yet

• Will take at least a month once started

Website Changes

• Many small changes being made, some good and some bad

• Focused on simplifying for new users who find Discord confusing

• Slow rollout with small cohorts to track metrics vs Discord

• Moving towards MVP for open public access between V6 and V7

Have fun.

If you need help…DM me on Linkedin.

Until next time…

Rory