WOW RUNWAY GEN-3 + Industry Changing FIGMA Reveal

Fam,

The game has been changed.

Because RUNWAY GEN-3 is starting to roll out.

And it will change what you think about Ai.

Here’s what we got for today:

1. RUNWAY GEN-3 Tests + Thoughts

2. FIGMA’s New Ai Play

3. MINDBLOWING Ai-Enhanced Creative Systems

4. MIDJOURNEY Campaign Development

5. BEYOND MIDJOURNEY: OPENAI’s Curious Hire

6. MIDJOURNEY OFFICE HOURS

Let’s Boogie.

RUNWAY GEN-3 TESTS + THOUGHTS

Runway GEN-3 is wild. (It’s creativity on steroids)

I was lucky enough to test GEN-3 as part of their Creative Partners Program.

Typically, they give us access 24-48 hours before they release.

(So it’s coming)

Here was the first test: FULL LINKEDIN POST

RUNWAY GEN-3:

This was Runway’s answer to SORA → Mission accomplished.

It’s still developing…but…The product is really impressive.

INITIAL THOUGHTS:

• HUGE improvement from GEN-2

• Natural motion improvement is crazy

• Only text prompting (at the moment)

• Prompt coherence is strong

• Small prompt details = big difference

• Options for (5)-(10) second video output

• Advanced camera motion/creative control

• Gen time is surprisingly fast

• Output resolution 720p (1280x768)

TESTING PROCESS:

• Push tools with complex ideas

• See where the rough edges are

• Run prompt multiple times 

• Test same/different seeds

TEST PROMPT: low-fi handheld camera footage of a man transitioning into a black werewolf, set in the forest of the Pacific Northwest

OUTPUT THOUGHTS:

• Really surprised at the output

• Handled a complex visual well

• (Can’t be much training data on this)

• These were 4/10 generations 

• Using the same seed

So this wasn’t a 50-generation test (It handled it well…right away)

PUSHING IT FURTHER:

The prompt coherence is also impressive.

Considering you can only TEXT PROMPT at the moment.

(No image reference yet)

I’ve tested some more seemingly complex concepts from a language perspective and it responds.

Complex Prompt: Starting from a ground-level view of a desert road leading towards a tunnel inside a mountain, the camera tracks smoothly along the road into a short dark tunnel. As it emerges on the other side, the camera rapidly ascends, revealing the road continuing to a beach dirt road emerald colored water on the left and palm trees on the right

ADDITIONAL THOUGHTS:

The human face is complex and we notice everything that’s wrong.

It impressively handles human motion like “a smile” or things like skin tone, pores, cracks, etc.

It’s super-defined and well done.

Prompt: selfie closeup shot of a woman's smiling face, she is spinning happily, 360 degree orbiting view, snow shimmers surrounding her.

MULTIPLE OBJECTS:

This was (again) eye-opening. 

AI has typically had issues with crowded images.

It’s TOO MUCH stimulus for it to focus on.

(This applies to most video + image generation tools)

Below it handled → Multiple cars, shifting lanes, complex environment.

Prompt: the camera follows black sports car as it races down a busy city street, the car weaves through traffic, 35mm cinematic shot

QUICK TIPS:

• Focus on the camera movements

• Think start point + end point for the shot

• Text prompting vid generators is harder

• Because of 3-Dimensional space

• Take that into consideration and describe the motion

• Be specific with details → Very similar to Midjourney

• If you’re having trouble…check the RUNWAY DISCORD

Be on the lookout - I would expect GEN-3 to be released THIS WEEK.

And when it does…say goodbye to your day.

More on this to come.

FIGMA’s NEW AI PLAY:

Guessing there are more than few FIGMA users on this NL.

I’m relatively new to the platform (16 months), but I realize I was far behind.

They just unveiled some updates at CONFIG that are actually really useful.

(Looking at you FIGMA presentations)

Here’s the QUICK RECAP:

FIGMA + AI:

• If you’re NOT a designer → These updates may not seem big.

• If you ARE a designer → You know how helpful this stuff is.

• If you manage a creative team → Think about how much this will help

AI UPDATES:

• Visual Search: Upload an image and find old designs instantly.

• Asset Search: AI-powered search finds the most relevant components within your boards.

• Content Generation: Generate realistic copy and images (the copy function allows you to adjust tone → Super helpful)

• Background Removal: Remove backgrounds without switching tools.

• Make Prototype: Turn static mockups into interactive prototypes

• Rename Layers: Rename layers with a single click using AI-generated contextual titles. (A personal favorite)

• Make Designs: Generate UI layouts and components using text prompts, creating the first draft of a design for manual editing.

The MAKE DESIGN feature is really wild…it generates an entire layout and gives you starting point (at a minimum) and it’s fully editable.

The most intriguing part…FIGMA gives you the option to “opt-in” to design training.

»So YOU control your data/IP«

And i’m a fan of them being transparent about this.

Here’s the link to the FULL KEYNOTE

MINDBLOWING AI-ENHANCED CREATIVE SYSTEMS:

This week Phillip Maggs from SUPERSIDE took the stage at CONFIG.

(And he did not disappoint)

If you’re interested in the FUTURE OF AI-CREATIVE SYSTEMS…

WATCH THIS VIDEO.

PARTNERSHIP WITH SUPERSIDE:

I recently PARTNERED WITH SUPERSIDE to work on exactly this.

The Problem:

• How can we scale an entire creative team with AI?

• Making processes and systems more efficient

• Take all aspects of our client’s “brand + data” into consideration

• While providing a service-focused approach

• (And not downsizing the team)

The Result:

• Across 500+ client projects…

• We saw 200% faster completion from brief-to-delivery

• 40-85% reduction in hours spent on projects

• SAVING clients over $1.4 Million.

The VIDEO above explains the breakdown of HOW we’re doing it.

It’s not just prompting Midjourney and generating an image.

It’s a fully scaled (ai-enhanced) system that accounts for the hiccups in the design process.

I’ll be breaking down one of the processes on Linkedin this week.

(Stay Tuned)

If you’re looking for a design team → WE’RE HERE.

MIDJOURNEY: CAMPAIGN DEVELOPMENT

Maybe you’re not in a position to implement a full ai system.

There’s still things you can do to utilize Ai in your workflows.

Check this mock campaign: FULL LINKEDIN POST.

THOUGHTS:

• IMO, Midjourney is utilized best in small format/short shelf life channels.

• Examples: Email Marketing + Social Media + Paid Media.

• It works best when you need asset volume to continue production.

• Think shots on goal vs. lasting message (like print/billboard)

• If you’re going to do this…you need to operationalize MJ a bit.

CONSISTENCY IS THE GOLD STANDARD:

Utilize prompt structures or formulas to get consistent output.

Like this one below:

Base Prompt Structure: Photo Type: [shot type] Subject Focus: [subject + action] Tone: [tonal balance] Colors: [color schemes] Texture: [texture description] Mood: [emotional tone] Environment: [setting details] Lighting: [lighting details] --ar 2:1 --personalize [your code] --stylize 200

The entire campaign can be built off (1) prompt structure + SREF.

Quick Thoughts:

• (--p) parameter helps with consistency

• Personalize + SREF is a powerful combo

• Use Img prompts for keeping image structure

• Use SREF for transferring aesthetics

• Text prompting is still important

• Structuring your prompts = fast iteration

If you want to dive deeper into Storyboarding…check out this post

BEYOND MIDJOURNEY: OPEN AI ADDS EX-NSA TO BOARD

I hype up AI a lot…but…I also want to provide some alternative perspectives.

I’ve been vocal about this OpenAi + Apple integration. (I don’t like it)

Now…OPENAI has added a former Director of the NSA to their board.

Consider this the TIN FOIL HAT section of the NL.

Here’s the breakdown:

TIN FOIL THOUGHTS:

It could be nothing…it could be something.

I’m just not comfortable with OpenAI having access to my emails, texts, browsing history, photo library, health data, geo-location, apple-pay, etc.

I’m sure the former Director of the NSA has zero use for this type of data or knowledge.

Sorry for the “conspiracy mindset”…We’ve just seen Big-Tech overreach time and time again…it’s concerning.

I love Ai for small/medium business and how it can exponentially level the playing field, but this is the side I don’t like.

More info here → AP Article.

MIDJOURNEY OFFICE HOURS (6/19):

Model Updates

• Midjourney 6.5: In stage 2 of 3-4,

• Midjourney 7: Delayed for new dataset requirements.

  Goals: better knowledge, improved prompts, faster performance.

• 3D and Video Development

  • Progress: 3D advancing faster, focusing on quality.

  • Recent Progress: Promising developments in 3D.

Demographic Surveys and Aesthetics

• Data Analysis: From user surveys.

• Future Features: Demo aesthetics for gender, age, and countries.

• Observations:

  • Current aesthetic is 75% male-dominated.

  • Male vs. female aesthetics differ greatly.

  • LGBTQ+ aesthetics are unique.

  • American and European aesthetics are similar.

  • Asian vs. Western aesthetics are distinct.

• Future Plans: Balance default aesthetics, encourage personalization.

Personalization

• Future Developments: Next-gen tests, possible reset option, and personalization profiles.

Style Exploration

• Random Style Explorer: Under consideration.

• User-Driven Pages: Potential "hot" and "top" style pages.

Have fun.

If you need help…DM me on Linkedin.

Until next time…

Rory