Latest / Scalability School / How to Use AI to Win in Marketing

Transcript

How to Use AI to Win in Marketing · Scalability School

0:00
For the foreseeable future, in order to differentiate,
0:03
you need that sort of mushy middle human ground.
0:05
Use these tools to scrape the internet constantly
0:08
for your content and constantly be putting that into Claude,
0:11
not GVT to analyze.
0:13
And we always sort of call them PAM clusters,
0:15
Persona Angle Motivation.
0:16
Like for every persona we're generating,
0:18
we want to know which angles to utilize
0:19
and under each of those angles,
0:21
which motivations to ultimately utilize.
0:23
I think people get so bogged down with their prompts.
0:26
And I think that's the wrong mentality to have.
0:28
I'm sure your prompt is fine.
0:29
That's why when you subscribe to Mid Journey or, you know,
0:32
Beyo 3.1, I think it's like the Google Ultra subscription.
0:35
You can use your prompt unlimited amount of times.
0:38
And I promise you, you will get something that is in the ballpark
0:41
of what you're looking for.
0:43
Editors are stuck in their ways.
0:44
You know, your strategists are sort of stuck in your ways.
0:46
So we did a sort of a competition for each model.
0:51
So we would say like, take your favorite statics, animate it.
0:54
Let's present it for weeks.
0:56
We're going to judge it and whoever gets, you know,
0:57
the best output gets, you know, monetary incentive.
1:00
And then we sort of leveled that up.
1:01
Let's create a video.
1:02
And then the ending result was creating a full ad.
1:12
And now let's take a listen to the Scalability School podcast.
1:16
All right. Welcome to episode 17 of the Scalability School podcast.
1:19
Almost 20 episodes.
1:21
It's a sensitive source.
1:23
More episodes and listeners.
1:25
There it is.
1:26
That's the new tagline.
1:27
That's really good.
1:28
I actually did see that we just surpassed.
1:31
A big number of downloads.
1:32
I'm not going to share it because it's private information
1:35
only available to our to our illustrious sponsors.
1:38
But it's but it's like, oh, wow, like this is real.
1:41
And it's not inflated either.
1:42
We haven't paid for anything.
1:43
You know, some of these some of these places do that, which we're not doing.
1:47
So that's why we only have tens of listeners.
1:50
But the people who are listening are real people.
1:52
And we have we're very big on YouTube.
1:54
I don't know if you ever heard of YouTube, but we're big on there as well.
1:57
By the way, special guest today, AI expert, CEO of self-made Will Sartorius.
2:02
Will, glad to have you.
2:03
Yeah. Thanks for having me.
2:05
Stoked to us.
2:06
And then I also mean the AI captain of the Foxwell Founders membership.
2:11
So yeah, unofficial moderator of the channel.
2:15
That's right. That's right.
2:16
So yeah, well, basically we had presentations in Portland.
2:19
You weren't there for our Foxwell founders meet up, Brad.
2:22
But basically everybody made all these different tables of like,
2:25
you know, Q4, creative, you know, you get an audit and like,
2:31
Wills was about AI and there was like the majority of folks were around
2:34
Wills table the entire time looking at what he was creating.
2:36
So that's what we brought to this podcast.
2:38
Yeah. Awesome.
2:39
Yes. Doke to be on for sure.
2:41
And I feel like we should like put a put a counter up like the number of times
2:45
we say Andromeda, you know, if you're listening, you take a shot.
2:48
Hell yeah.
2:49
Yeah, I got my electrolytes right here.
2:50
So I'm ready to I'm ready to roll.
2:52
Yeah.
2:53
But like, you know, for us where this sort of entire impetus behind
2:58
getting really good at creative through AI came from was well one
3:02
at the front end we sort of perfected using social listening to collect,
3:06
you know, Facebook comments subreddit comments Twitter, etc.
3:09
Like built our own Shopify scraper.
3:11
And so like now you can plug in a brand and we can pull all this information
3:13
like corporate sonas, core motivators, angles, etc.
3:17
So that was like the front end that I developed about a year ago.
3:20
I think I talked about that in Amsterdam and now I was like,
3:22
okay, so the front end is done now.
3:24
We need to sort of figure out the back end.
3:26
Like how can we take the content that we're creating and sort of amplify it,
3:29
right and there's a lot of different ways to go about this.
3:33
There is sort of like one sort of all-star flow that I like to talk about,
3:37
but there's also a lot of little tweaks you can do which are probably best
3:42
for specific use cases whether it's you have an image and you just like
3:46
it's 10 p.m.
3:47
Your editor's offline and you just like need to resize it, but you only
3:49
have the PNG there's a solution for that.
3:52
You have statics that are maybe fatiguing and you just like want to convert
3:56
them to gifs, add a little animation, CPMs on gifs.
3:59
We always see being a lot better than statics even for the same image.
4:04
Generally in a 1 to 1 A-B test, gifs will generally always outperform statics.
4:08
So just like adding a little sort of fuel to the fire that works too.
4:11
I'm not a big clone ads guy.
4:13
I know folks like to do that and like sort of take competitors and like
4:16
insert their product like that's never something I've wanted to do, but
4:20
we can sort of talk about a flow of like finding inspiration on Pinterest
4:23
converting that to mid-journey adding your product into that image and
4:27
then that is sort of like how I like to clone ads because like at the end
4:30
of the day where AI is really helpful is again, it's a barbell.
4:34
It's like it's setting the foundation for strategists like which persona
4:37
angle motivation do you want to focus on and then it's amplifying existing
4:40
content like that messy mushy middle.
4:43
I feel like it's so innately human and like unless you want to just revert
4:46
to the mean and just like do what everyone else is doing and contribute
4:50
to you know, the sort of slot parade, you know, by all means, but like I
4:54
don't think any of us on this call or in our team are going to or aiming
4:57
to do that.
4:58
So no to that end like I talked to my team is sort of about like type
5:01
one versus type two thinking like Dan Kahneman, you know, RIP if you
5:06
haven't read thinking fast and slow you should pick it up right now.
5:09
It's the most amazing book I've ever read but like obviously type one
5:12
thinking is just like I my brain is shut down.
5:15
I'm just on autopilot type two is very much more so I'm being sort of
5:19
thoughtful in my approach and I feel like a lot of folks use AI for type
5:23
one thinking like, oh, how does it go on or fucking write this email?
5:26
Like I'm just going to do this or like, oh, you know, I need a concept
5:30
GPT help me write a concept or what have you and again, like if you want
5:33
to sort of revert to the mean and just like the average like do that, but
5:37
like if you can be thoughtful about your AI workflow and again, like
5:40
amplifying existing content and you know, sort of human messy thought
5:45
and you're going to be in a much better place.
5:48
This episode is brought to you by our friends over at Northbeam the
5:51
marketing attribution platform that a lot of us that listen to scalability
5:54
school and the host of scalability school, you know, we swear by and use
5:58
Northbeam rolled out something brand new that I think is pretty awesome.
6:00
It's real unlock I think called clicks plus deterministic views and
6:04
it's the first ever deterministic view through attribution model.
6:09
I mean, let's be honest, right?
6:10
One day click had us flying a little bit blind, you know, the drill, right?
6:14
Your TikTok or your CTV campaigns are building awareness clearly driving
6:18
site traffic and lifting sales across channels, but your attribution dashboard
6:22
still gives all the credit to Google search because it was last click.
6:25
It's like tracking who made the final shot in a basketball game, but
6:27
ignoring the five assists, right?
6:29
That got the ball there.
6:31
So clicks plus deterministic views flips that it ties actual impressions, not
6:35
guesses, but verified impressions to real conversions.
6:38
We're talking direct integrations with meta, TikTok, snap, Pinterest, and
6:42
even CTV and Northbeam uses hard identifiers like order IDs and hashed
6:47
emails to prove that someone saw the ad didn't click, but later bought.
6:52
That means you can finally measure the real impact of upper funnel
6:55
campaigns and the results are pretty nuts.
6:57
One brand saw 283% more attributed conversions on Tiktok.
7:02
Another got 175% more attributed revenue than they knew even existed.
7:05
And you know, another one drove 59% ROAS lift on Pinterest.
7:12
So if you're tired of under budgeting awareness because attribution can't
7:15
keep up, this is your chance to see the real journey and the full journey.
7:19
So Northbeam click plus deterministic views is live now head over to
7:23
Northbeam.io forward slash demo to book a walkthrough and tell them we
7:29
have scalability scores.
7:31
So can you give like an example?
7:32
So you gave a couple examples of type one, which was like, okay, this
7:35
is the email I got, I yapped into GPT with my, my, what I think I'm
7:39
going to respond with.
7:40
Can you make me sound like less of a dick?
7:42
And then I kind of hit send.
7:43
So that's an, maybe an example of type one.
7:45
Maybe you can give an example of, of like where people are using that in,
7:48
in the creative process, but then like type two, which is maybe some of
7:51
the examples that we're going to go through, but cause I know in like
7:52
the original version of your notes here is like, that was a big
7:55
distinction for you to go through.
7:57
So maybe we spend just a couple minutes like giving some examples before
7:59
we dive into some of the workflows that you have.
8:01
Yeah, no doubt.
8:02
So I would say type one is clone this ad for me.
8:06
It's, you know, taking someone else's content that may not be applicable
8:09
to you at all and just effectively using that as your own, like maybe
8:14
creative volume is what you're after and you don't really care about
8:17
anything else.
8:18
By all means do that.
8:19
Where again, like type two sort of comes in, it's like you can outsource
8:22
a lot of thinking to AI, but I do think at least today and, and, you
8:27
know, for the first year of a future in order to differentiate, you need
8:30
that sort of mushy middle human ground, but like use apify, you know,
8:35
use Giga brain, use these tools to scrape the internet constantly for
8:39
your content and constantly be putting that into Claude, not GPT to
8:43
analyze.
8:43
And we always sort of call them Pam clusters persona angle motivation.
8:47
Like for every persona we're generating, we want to know which angles
8:50
to utilize and under each of those angles, which motivations to ultimately
8:53
utilize.
8:54
And so that framework is created for our strategist internally, right?
8:58
And so then we say, okay, strategist a here's your persona angle and
9:02
motivation run wild, right?
9:04
Like we've effectively now created a swim lane or, you know, a better
9:07
metaphor is we've put the gutters down in your bowling lane.
9:10
You know, you can't sort of shoot a gutter ball.
9:12
Now, we know that this is going to resonate with consumers, but now you
9:15
have these frameworks create a bigger swing, right?
9:19
Or, you know, conversely look at the ad account.
9:21
Maybe with this persona angle motivation combination, we haven't
9:24
done an ugly ad or maybe we haven't done like a three reasons.
9:27
Why or we haven't talked about that in a founder interview and we need
9:30
to get the founder on the horn and jam with them about it.
9:33
Right?
9:33
So it's like create these lanes beforehand using AI have the human
9:38
sort of pinpoint like where we're missing and like that hole will be
9:42
filled with AI.
9:43
Don't doubt that me.
9:43
Give me give it six months, 12 months.
9:45
Like we're going to be able to identify where the holes are and your
9:48
creative output like right now.
9:49
That's it in a neatly human thing.
9:52
And then from there, like do, you know, sort of do the work, right?
9:56
Use editors do post-production work.
9:59
And then like once you sort of have something that you're ready to go,
10:04
you know, amplify that with AI.
10:05
I will say one thing that we're trying to do and we've mandated internally
10:10
is we can't use stock anymore.
10:12
No stock photos.
10:13
You know, I got it like in this dispute with artless.io.
10:15
They were like, we're going to charge you 10k to like, you know, frickin,
10:21
you know, up your license.
10:22
And I was like, all right, sign our sucker.
10:24
Like, you know, eggs on your face, not mine.
10:27
Like I was probably going to continue paying, you know, 120 bucks a month.
10:30
But every piece of content that we're creating that would be stock is now
10:35
AI, right?
10:35
Because now we can create stock with actually the product in it and, you
10:39
know, using tools like C-Dream, you can actually make the people look a
10:42
lot less plasticy.
10:43
And now that Veo 3.1 has start and end frames, like the world is your
10:48
oyster.
10:48
You can create whatever content you want.
10:51
And so like my long-term strategy.
10:54
And I think this should be really anyone's long-term strategy is that like
10:57
you on your teams have sort of like creative visionaries, right?
11:00
Someone who's like thinking like this is a sort of vision, but this is
11:04
where we need to go.
11:05
And then we, we hired people to just where their entire job is to generate
11:08
B-roll for that brand, right?
11:11
Just using Veo 3.1, using Nano Banana, using C-Dream.
11:15
And like now our editors have this plethora of B-roll that they can
11:18
utilize that looks legit.
11:20
And you know, if you get clients saying I'm averse to using AI, like we
11:24
get this all the time and we're like, okay, cool.
11:26
Let's just do one for you and we'll prove you wrong.
11:29
Right.
11:29
And like we have clients that are doing like around five, six million
11:32
a month that were extremely averse to using AI.
11:35
We did a video for them and they're like, so where did you guys shoot
11:38
this?
11:38
And we're like, gotcha.
11:40
So it was just like one of these instances where, you know, again, I
11:44
don't want to knock on, there is a massive value add to be, you know,
11:49
going shooting content.
11:51
But like, if I just need a shot of me picking up this coffee cup, you
11:54
know, or I mean, Brad, maybe I needed you picking up that coffee cup.
11:58
It's going to be a lot easier for me to just do it myself.
12:00
Yeah.
12:01
You know, it's funny you mentioned that because literally within the
12:03
last 48, like prior to our conversation, but right before, like over the
12:07
last 48 hours, we've had somebody just like as a role, test it out, just
12:11
like building a B-roll library with, with whatever tools, all of the tools
12:15
that I'm sure we're about to talk about.
12:16
And yeah, I'm sure it'll be an iterative process to zero it out, but
12:19
especially that's why I was even more excited to talk is like, we're,
12:23
we're playing with this now a little bit more than we have been.
12:25
And I'm going to steal some of this from myself.
12:29
Yeah.
12:29
There's a lot that we can, there's, I think the frameworks already that
12:32
you've gone through will are really helpful.
12:34
I mean, I think a lot of it is so much of the centers on people talking
12:37
about Andromeda and you know, there has to be not just creative diversity,
12:40
but now it's like, you know, we had a question in the founders community
12:43
this morning of like, all right, look, how are you even reworking your
12:46
creative department to make sure, because this requires more work now,
12:50
right, that we have to come up with really net new ideas and not just
12:53
iterations necessarily upon thing.
12:56
So I think we all feel that squeeze and obviously AI can help us get there
13:00
and make that a lot easier for us.
13:02
So let's talk about, so you have a number of different workflows to go through.
13:06
Um, so static, the first one you're talking about a static to, I believe
13:11
you pronounce it Jif, like the peanut butter.
13:13
Um, I'm a gift guy.
13:14
I'm a gift.
13:15
Personally.
13:15
I don't know.
13:16
Of course I've always been, I've always been a gift guy.
13:19
I just, I just still love calling it Jeff.
13:21
So it's using mid journey.
13:23
So how does this one work in your opinion?
13:26
And by the way, if you're not listening, if you're, if you're not
13:28
watching this on YouTube, this is also, I was going to maybe even do some
13:31
screen sharing here and if you're on YouTube, if you're, if you're on
13:34
youtube.com, you'll be able to see this live, but otherwise, if not, or
13:38
you'll be able to see this recorded, if not, we'll can talk about it and talk
13:41
through, yeah, for sure.
13:42
And yeah.
13:42
Um, more than happy to sort of talk through my process as I go through.
13:46
And one thing to sort of add, you know, Brad, you mentioned that like your
13:49
billing sort of a team around this, I think that was the number one
13:52
question I got in Portland.
13:53
It was like, how are you getting your team to do this?
13:56
Cause it's a lot, it's very easy to be an agency and it'd be like,
13:59
guys use these AI tools and everyone's going to be, you know, sort of
14:03
flip you the bird, be like, why would I do that?
14:04
Like, it sounds like a huge pain, right?
14:07
Editors are stuck in their ways.
14:08
You know, your strategies are sort of stuck in your ways.
14:10
So we did a sort of like a competition for each model.
14:15
So we would say like exactly what we're sort of talking about now.
14:17
It's like, okay, we're going to start with mid journey, take your favorite
14:20
statics, animate it.
14:21
Let's present in four weeks.
14:23
We're going to judge it and whoever gets, you know, the best output
14:25
gets, you know, monetary incentive.
14:27
And then we sort of leveled that up.
14:28
Okay.
14:29
Take that gift.
14:30
Now let's put into Vayo three.
14:31
Let's create a video.
14:32
And then the ending result was creating a full ad, right?
14:35
And like, that was a great way to get buy-in from everyone into the process.
14:39
And it made it a lot more fun too.
14:40
Cause like we created teams.
14:41
Like it was just like, there was a lot of team spirit.
14:44
So it was, that was an unlock for me.
14:46
Cause I was just like, why are we not using this more?
14:48
And then someone on our team, much smarter than me suggested doing that.
14:52
So. Can I add some really, sorry, I know we're
14:55
like distracting from the workflow stuff,
14:56
which I'm very excited to get into, but do you feel like you have to like very
15:00
directly address like your perception of what you think AI is going to do to their
15:04
jobs to help get over that as well?
15:06
Because like we have, we have a bunch of designers on our team.
15:08
It's like, they see the tools and they're
15:09
like anxious about what that means for them.
15:11
And basically I tell them is like, if you
15:13
can figure out how to use this, like you're
15:15
you're replacing yourself and you continue to be here.
15:18
It's like, you're just, you're just getting better and you're, you're
15:20
leveling up and what you do it.
15:21
The same goes for me, by the way.
15:22
There's a bunch of tools that I can be able to do what I'm doing.
15:24
You know, I used to spend, we used to spend hours building ads.
15:26
Now we have tools like advantage that it takes three seconds to launch 50.
15:30
You know, so it's like, that's perpetually happenings.
15:33
Do you feel like that's the case for you?
15:34
Like, do you have to, do you find yourself explaining this to your team?
15:36
Totally.
15:37
I mean, I think I like to sort of belabor the pain points that
15:40
certain folks are feeling like, Oh, we don't have footage of this or blah.
15:45
Like, can we get the client to send us this?
15:47
It's like, well, we can, but like now you
15:50
have that sort of ability to do it yourself.
15:52
I also will say like, and this may be sort of like a hot take.
15:55
I actually think editors are going to be coming increasingly more important.
16:00
And I think sort of the top of funnel, again, like the, and when I, when I say
16:04
top of funnel, I really mean like the personas angle motivation, like
16:07
social listening discovery, like that is going to be AIified.
16:11
So like, I actually think the work up here is going to be more AI work down here.
16:16
Well, cause like, if you, I don't know if you guys have played with Sora, but
16:19
like, it's just like vibes, right?
16:22
You sort of hope you get something right.
16:23
If you try to do like an intricate prompt, Godspeed, but if you get something
16:27
that's like close and you have a good editor game over, right.
16:31
And so that's what I always sort of say to my team, like, you know,
16:34
you are sort of filling in the cracks and we are solving the pain points that,
16:38
you know, you have brought up many times in the past and everyone's creating
16:43
something, everyone's bringing something new to the table every week.
16:46
And it's just like, it's a great way to get folks like more comfortable with
16:50
these tools and then ultimately, like I said, initially, our goal is to have
16:55
creative sort of visionaries.
16:56
This is the whole ad I want to create.
16:59
I want to create a 30 second ad of this man going hunting.
17:02
And then he comes home to his family presents the deer on the table, the kids
17:06
eat, you know, whatever, and now our team is going to be way more comfortable
17:09
creating that, right?
17:11
And so like, I don't want to be confined anymore by the existing content we have
17:14
on hand.
17:15
I want our team to be able to sort of think, you know, not just outside of the
17:19
box, you know, sort of outside of any parameter that existed previously, a box.
17:25
What is bigger than a box?
17:26
I don't know.
17:26
Uh, whatever, you know, the prison universe, you get the universe out of the
17:30
prison, that's, that's, that's, that's electric.
17:32
I like that.
17:33
Yeah.
17:33
Think outside the prism, uh, you know, sort
17:36
of pink Floyd style, go to that next level.
17:39
And so we want to have a world where I had this big loft division for an ad.
17:44
Now our, our, our editors coming in and saying like, okay, cool.
17:48
Like this is fun as hell.
17:49
Like this is up to me to like, come up with this and like the, the visionaries
17:54
and divisions are not going to be too prescriptive.
17:56
So like that is where we're heading.
17:58
Candidly, you know, for, for me and a few others on my team, like this process is
18:02
quick, right?
18:02
Like I could probably do some B roll and that I feel really
18:05
comfortable with in like 20 minutes for some, someone else that's new to this.
18:09
It may take them like four hours.
18:11
Right.
18:11
So like that is like a painstaking process.
18:13
It's exhausting.
18:14
Right.
18:14
So how do you sort of overcome that hurdle?
18:17
And what we've done is like for our clients, it's like, we're not going to do
18:20
four AI videos for them a month.
18:22
We're going to do one just so you can get a little bit more comfortable.
18:25
And it's just sort of like, you know, probably about analogy, but put the
18:28
tote in the water, lowly start heating it
18:30
up so people get more and more comfortable,
18:33
you know, with getting, you know, boiled, I guess, but you know, they're going to
18:37
hop out, they're going to be fine.
18:38
Yeah.
18:39
I mean, I think it's, I think it's a, you know, a member asked me at the meetup
18:42
that, you know, look, I, we have, we have two people that are fully dedicated
18:45
to this on our creative team.
18:46
This is an agency person.
18:48
This podcast is for seven and eight figure brand owners, but I think it's,
18:52
I think it's important nonetheless, you know, how they're fully
18:55
dedicated to this.
18:55
And everyone was like, wow, we're really jealous of that.
18:57
Right.
18:57
And, and I actually think that's okay, but I actually think that's not
19:00
necessarily the greatest idea, because I think you, you want to not have this
19:04
siloed and you want to have everybody kind of talking about be integrated as
19:07
much as possible and be testing.
19:09
And I think that another piece that you didn't mention, maybe you're doing this
19:11
is that you want to incentivize people to do this, right?
19:13
You want to incentivize like, Hey, who comes up with the coolest concept?
19:15
You know, you are going to get a financial bonus aside of this.
19:18
And I think that if you have people in house or you yourself as the founder
19:21
or creating these ads, you know, giving yourself a half an hour a day
19:24
to screw around with this, right?
19:26
Like, and if you're utilizing Will's workflows, you could see all that's
19:29
happening is not necessarily dependent upon one tool alone.
19:32
It's just a number of different things where you're going back and forth between
19:34
them and refining the outputs and going from there.
19:38
And, you know, I think if you dedicate some time over a couple, you know,
19:42
couple of weeks, you're going to get pretty good at this stuff.
19:44
Even if it's 30 minutes a day.
19:46
Um, I mean, that's what I tell myself about weightlifting.
19:48
I'm like, it's, it's just 60 minutes, you know, and over time I'm
19:52
going to get big like Brad, but anyway, so I think that's, I think it's important.
19:56
I think that, uh, how do you recommend that people continue to stay on top of this
20:00
stuff?
20:00
Cause it's obviously changing super rapidly.
20:02
Like, you know, I feel like every other day there's some new thing.
20:05
How do you recommend people stay on up to date with it?
20:07
I mean, obviously your newsletter is a place that they can go to, which we can,
20:10
we can link in the show notes and stuff, but where, what are other places as well?
20:13
Yeah, obviously my Friday's newsletter shameless plug is all new models.
20:18
I always go through my workflows there.
20:20
Um, and I, you know, just sort of backtrack for one second.
20:23
How do you get people excited?
20:24
I promise you, if you give your team unlimited access to mid journey, they
20:28
will have so much fun that they're going to sort of like almost forget about their
20:30
other work.
20:31
Like we have our team going in there and constantly just like playing around with
20:34
it because it really is so much fun.
20:37
But like, obviously when you're getting much more analytical with a process,
20:40
like with nano banana, BIO 3.1, it's like very prescriptive mid journey.
20:45
Again, I can't recommend everyone subscribing their team to it, you know,
20:49
create a group email and it's probably not kosher, but like, you know, have
20:53
everyone go onto that group, me email at $60 a month and everyone can request as
20:57
much as they want.
20:57
And like, that is a great sort of starting point.
21:00
So, uh, yeah, sorry.
21:01
What was your second question?
21:02
My second question is just like talking about where else people can go to.
21:06
Yeah.
21:07
Yeah.
21:07
To date on stuff.
21:08
Yeah, for sure.
21:09
I can, like, I can share as a few Twitter handles that I follow that, you know,
21:13
sort of talk about the first thing I do when a new model is released is obviously
21:17
I go to Twitter and I say like, okay, what use cases are people talking about?
21:21
So like as Vayo 3.1 and so Vayo 3.1 dropped the day before we were in Portland
21:27
before I was supposed to do this, you know, do a demo.
21:29
I was like, shit.
21:30
Okay.
21:30
And so I went to Twitter and I sort of figured out like, what are the use cases
21:33
here?
21:33
Like if someone says they're building an
21:35
N8N flow and they're pumping out a thousand
21:38
creatives, block that person.
21:40
And it is a great tool for in-house workflows.
21:43
It is a shit tool for creative output.
21:46
I'm sorry, because like you see our process, right?
21:48
Like we have to refine the image or the video at every single stage to get
21:52
something that's actually workable.
21:54
You're adding so much risk and saying like, I'm going to use the first image in mid
21:59
journey generates and I'm going to use a first nano banana image.
22:01
I know I'm going to use the first Vayo 3.1 image and like, that's going to be your
22:04
output, right?
22:06
Well, I don't update.
22:07
I never put less than a thousand ads in when I update any campaign.
22:11
Minimum.
22:12
It's always a minimum.
22:14
Yeah.
22:15
It's, it's just the way that I do it.
22:17
One, one ASC per, uh, you know, one dollar a piece.
22:22
That's where I'm at.
22:23
Set a point, point one row as target and just let it rip.
22:26
Exactly.
22:27
So, uh, a final question that I have, Will, and a lot of this is, um, you, you
22:31
foresee that the, that the, right now you're, you're manually kind of plugging
22:35
and playing, you foresee that a lot of this is this like in what a year is all
22:39
this is going to be all in one year.
22:41
It's not going to have to do this or like, what's the timeframe on this?
22:44
Yeah, of course.
22:45
Like we're all hypothesizing here.
22:46
I would say if, if you don't learn these flows, like now you're going to be a
22:51
little left behind, I would say like get really good at these workflows.
22:54
Now in six months, maybe a year, the barrier to entry is going to be zero.
22:58
You'll be able to do effectively all of this and you know, what will be
23:02
Americans, we chat.
23:03
I don't know who's going to build it, but like that is going to sort of, you
23:06
know, be what exists and like, I would be remiss to not even talk about, you
23:10
know, the open AI browser that was just released, right?
23:13
Now you can sort of just say, buy me all these different things.
23:16
So like maybe creative will have a less of
23:18
relevance in the future in and of itself.
23:21
Right.
23:21
So like that is GTM LLM optimization, which is like a completely different animal.
23:26
But at the end of the day, like if you are generating creative right now, like
23:31
I was, I read an article the other day that like Coca-Cola paid a hundred
23:34
thousand dollars for an ad that was generated with Vayo 3.1.
23:39
So, you know, sadly we're not contracted with Coca-Cola, but what I can tell you
23:44
is that like, if you want to sign new clients and you want to really impress
23:48
big brands, do this now and you will lock them in.
23:52
If you were on a sales call, go into Sora.com, take their product, say, generate
23:56
EUGC ad with this product, show it to them on that call and they're going to sign.
24:01
Like we do this all the time.
24:02
The idea is that like you right now, the
24:05
closer, the closer move, always be closing.
24:08
One call is good stuff.
24:09
This is good stuff.
24:10
I like it.
24:10
Anyway, keep going.
24:11
Sorry.
24:11
I interrupt you.
24:12
Yeah, no, all good.
24:13
So like I will say like if you, these, these enterprise companies and these
24:19
larger companies know they need to be good at AI, right?
24:22
They know they're probably falling behind.
24:24
And so like if you lock in an annual contract or six month contract now,
24:27
when that sort of paradigm shifts and the barrier to entry becomes lower,
24:31
you're already going to be so much further ahead than ever anyone else that's just
24:35
joining when that sort of barrier to entry is lower because you're going to
24:39
know, you know, how to prescribe certain elements, even within a lower barrier
24:43
to entry, right?
24:44
And like goes back to sort of type one, type two thinking, uh, and type one,
24:48
type two sort of AI, like a lot of people are going to be generating a lot of slop.
24:52
If you know how to use these AI tools to not generate slop and actually generate
24:56
something thoughtful with actual creative prompting and strategy behind it.
25:01
Like you're going to be eons ahead.
25:03
Like when that barrier to entry, the slop gates open, you know, misinformation
25:07
is going to be rampant.
25:08
We're not going to know what's up and down, but what we will know is that like
25:12
differentiated content that doesn't fit that mold will rise to the top and
25:15
ultimately when we're, where can I get better prompting?
25:18
Yeah.
25:19
How do I get better prompting?
25:20
Again, like I, again, this is probably a hot take, but I don't think you need to
25:24
really be a good prompter.
25:25
You're just asking GPT, Hey, this is what I'm going to do.
25:28
Right.
25:29
Exactly.
25:30
And that's why like, I think people get so bogged down with their prompts and I
25:35
think that's the wrong mentality to have.
25:37
I'm sure your prompt is fine.
25:38
That's why when you subscribe to mid journey or, you know, Vayo 3.1, I think
25:42
it's like the Google ultra subscription.
25:44
You can use your prompt unlimited amount of times.
25:47
And I promise you, you will get something that is in the ballpark of what you're
25:51
looking for or conversely, you know, go to Sora and like, you know, just
25:55
vibe it, like that's an option as well.
25:57
But like, I think people just get too hung up on when it comes to creative
26:01
output, people get too hung up on prompting.
26:03
Obviously briefing is a completely different animal where prompting
26:06
actually matters a lot more, but yeah, with creative output, like don't
26:10
overthink it.
26:11
The only, the only thing that's important and I can stop my rant after this, you
26:15
know, Nana, banana, C dream mid journey text based prompts are fine.
26:19
If you're doing Vayo 3.1, Jason prompts work a lot better.
26:23
So when you're, when you're talking to GP 25, ask for a Jason output, JSON.
26:29
If you're not familiar, you probably won't recognize like this, this style.
26:32
If you're not familiar with HTML or coding, but like use that.
26:36
And then you can tweak that.
26:37
So rather than trying to engineer your prompt yourself, put your
26:41
Jason prompt into Vayo, take your output, put that back into GPT and do that
26:45
cycle we did earlier, right?
26:46
Adapt your prompt based on the output you got, right?
26:50
Don't go into the prompt yourself and like try to tweak it.
26:52
Like that will just drive you mad.
26:55
Like this system where you're just sort of like having a positive feedback loop.
26:58
I promise you, like you'll get a much better output and like you'll be, you
27:02
know, home in time to pick your kid up from daycare.
27:05
So everybody listening to this will be a year ahead because they, they listen to
27:08
this, but I'll be at least a couple of weeks ahead because I'm going to download
27:12
this and send it to my team before this goes live.
27:14
So, you know, he's going to be closing deals sitting there.
27:18
Prompt to this with UGC.
27:19
That's good stuff.
27:20
So now let's go into actually getting into some of these prompts really deeply.
27:24
If we want to, you know, if you want to learn how to do this yourself, my
27:27
recommendation for anyone is to always sort of start in mid journey.
27:31
It's the easiest platform to utilize.
27:34
So I'll just share my full screen here.
27:36
Actually probably easier.
27:36
Just show it up my window.
27:38
Okay.
27:38
So again, a journey is the safest, easiest tool to use.
27:44
You don't know anything about AI, get mid journey 60 bucks a month.
27:48
You get unlimited prompts and it does video.
27:51
It does image obviously metadata partnership with mid journey.
27:55
So I'm sure we're going to be seeing more mid journey content in there.
27:59
But again, like this is always my recommended starting point
28:02
because you can do a lot with it.
28:04
Um, and so in this sort of exercise, we can talk about, you know, sort of
28:08
taking a static ad, adding some motion to it, and you can sort of see, I was
28:11
messing around with this, you know, a little bit before our call, one quick
28:15
thing is when you start with mid journey, you're going to want to create
28:19
a new profile, um, and answer these 200 questions, you know, which do you
28:24
like more, which do you like less?
28:25
It doesn't really affect your output.
28:27
It's just, you know, trying to train its LLM of course, but this will give
28:31
you X to V seven in mid journey.
28:33
And that's the model you ultimately want to utilize.
28:35
So again, where I like to use mid journey the most, and I think the safest sort
28:39
of place to start is taking existing statics you've created and adding some
28:44
motion to them.
28:45
So where I like to do this is actually the only time I ever use chat GPT is
28:50
with, uh, prompting the sort of content we're discussing now, everything
28:54
else I use Claude for Claude is like a little bit more creative.
28:57
Uh, but chat GPT, the main reason I use chat GPT, and this is really important.
29:02
So like every prompt that has copy in it,
29:04
you need to include that copy in the prompt.
29:07
So like, or every ad rather that has copy in it, you need to use that copy
29:10
in the prompt and chat GPT does a better job reading copy on ads than Claude.
29:15
That's really the main reason.
29:17
Um, and I'm sure many folks have discovered that in the past.
29:20
So, uh, you know, here was a hollow socks ad, uh, you know, built for serious
29:25
hunters that I found, you know, in a Facebook ad library, right?
29:30
And so the first thing I want to do is I want to create a prompt.
29:34
So, you know, let's just sort of say like, I'm exhausted.
29:38
I've had a long day and I'd say, you know, uh, I want to take this static
29:43
ad and convert and add animation to it.
29:49
I will be using mid journey, give me some ideas.
29:52
So this is generally, you know, sort of brainstorming, like, um,
29:57
where, where do we begin?
29:59
Um, and so like, in this case, if it's built for serious hunters, we probably
30:03
need to like add some animation there.
30:05
And like, of course we could do something simple with, with, you know,
30:08
ruffling leaves or what have you.
30:11
Um, it doesn't really matter.
30:14
Um, so we can just sort of see if like anything here is sort of interesting at
30:17
all, uh, you know, slow motion, fog drifting, atmospheric lighting, et cetera,
30:22
into the hunt, this is, uh, I sent you guys this before, but like, you know,
30:25
maybe having some animals moving in the background.
30:28
And so like, if I had a specific idea here, like I would obviously skip this step.
30:32
Uh, but this is always a generally good place to start.
30:36
So we'll do what we did earlier.
30:38
Say, uh, actually let's animate it.
30:42
So there are a few animals in the background.
30:45
We want to create a prompt for mid journey seven, please include all of the
30:51
copy that is on the static ad in the prompt.
30:55
So now this will give me something, you know, semi usable theoretically.
31:00
Uh, and nothing like the sick of fainted GPT five.
31:03
I always miss it when I come back.
31:06
Uh, okay, cool.
31:08
So now we have our prompt, right?
31:10
So I'm just going to copy this.
31:11
I'm copy this.
31:12
I'm going to go into mid journey here.
31:13
Uh, I'm honestly going to get rid of all of this and then just go into settings
31:18
and do nine by 16, make sure our model is on version seven.
31:21
And then I'm going to, for those, for those that are listening, you're
31:24
saying getting rid of this and what did you delete there?
31:27
Sure. Sure.
31:28
So, uh, GPT five likes to add on all these sort of extra things, right?
31:32
Like aspect ratio nine by 16 version seven style, raw chaos, 10 motion,
31:37
three camera moves, slight pan, weird zero.
31:40
Like these are things that you can customize for, but ultimately it's
31:45
anything with sort of like the two dashes before them, right?
31:48
And so like, if you again, like are listening to this, what I would
31:52
recommend doing is within, um, within mid journey, there's a settings
31:57
button and you can choose all of these settings yourself within the settings
32:00
button.
32:01
So here I'm able to choose my image size.
32:03
I can just like square landscape portrait.
32:05
Uh, we'll just go nine by 16 here.
32:07
Stylization.
32:08
I just generally use the standards here standard.
32:11
The most important thing is a version seven.
32:15
Um, and always I like to do four generations.
32:18
What I have historically said is that like one out of five will be good for
32:23
gif one out of 10 videos will be good.
32:25
So, uh, you know, we'll prompt this and you can see within my prompt, it says
32:29
hollow at the top and it says built for serious hunters.
32:32
So in my prompt, I'm copying the copy over.
32:36
Um, which is, you know, definitely an essential part of, you know, and I just
32:41
realized I didn't add the image first.
32:43
So let me just do that.
32:44
So now we have, uh, the starting frame, uh, and the end frame.
32:47
So now it's just gonna, this is just going to make an assumption.
32:49
So I'm just going to cancel this one.
32:51
Cool.
32:51
Any questions so far while this generates?
32:54
No, honestly, super straight forward.
32:55
Yeah.
32:56
It's really not too complicated.
32:58
So again, like this is just a way to like, take your existing statics,
33:01
multiply them by two, five, 10.
33:04
Like we can go back to our GPT five and like, maybe there was some
33:07
other cool ideas in here, right?
33:09
Like, uh, light fog drifts, tiny particles.
33:13
Like you could do more of like a cinemagraph style, right?
33:16
Uh, so it's, again, the ability to take your existing content, amplify it,
33:21
uh, for, uh, for different audiences, if you will.
33:24
And then this one, it just put like a kind of put a deer walking behind it,
33:29
which is pretty cool.
33:30
Um, and then what the other ideas that they came up with are, it looks
33:33
like some sort of British hunter.
33:36
Yeah.
33:36
And so, yeah, this was, there's an,
33:38
there's like an alpaca idea that I came up with.
33:41
So yeah, it's kind of cool.
33:43
Totally.
33:43
And, you know, I actually did this sort of similar prompt earlier.
33:46
And so we can sort of see here, you know, the outcome of that, right?
33:50
We have these sort of four versions with, uh, you know, the buck in the background.
33:55
A few of them has have animation.
33:56
Obviously the copy was eliminated from this one.
33:58
So I would scrap that, but like something
34:00
like this, you know, I would certainly run
34:02
and you can see the copy rendering is very strong, right?
34:05
Like there's no sort of fuzziness.
34:07
Uh, everything sort of comes across super clean.
34:09
And like now I have a gift that I just created in what, uh, you know, 12 seconds.
34:16
That I can run as an ad should, uh, send it to Zach and let it rip.
34:20
See if it works.
34:21
Yeah.
34:21
So here we go.
34:21
Like here's another option, right?
34:23
Those are wild.
34:24
Yeah.
34:25
So now we have like a little bit more animation in the background and you
34:27
can see here, we lost our copy, right?
34:29
So like I said, one out of five will turn out well.
34:32
The rest will sort of be mid.
34:33
Yeah.
34:34
Honestly, that's not even that bad, right?
34:35
Like the built for serious hunters is up long enough that you might see it.
34:38
And then you get to see the deer walking more clearly behind.
34:40
I like how it's rotating.
34:42
Totally.
34:42
Yeah.
34:43
The sock is rotating.
34:44
That's kind of cool.
34:44
It's kind of interesting.
34:45
Yeah.
34:46
It's clean.
34:46
I like that.
34:47
Yeah.
34:47
I have a question about that quick before we move on though.
34:49
Do you find that you need to like, I mean, the, the, the script that you
34:53
have is not overly simplistic, right?
34:55
There's like a decent amount of detail to that, but is there, is there a balance
34:58
between like how much you should be giving it or is what you get back from
35:02
GPT, like GPT does a pretty good job at keeping it the essentials.
35:06
Yeah.
35:06
It's funny you say that I think people over engineer their prompts and they
35:11
think like this little tweak will change that this little tweak will change this.
35:14
I'm very much of the mindset that your prompt is fine.
35:17
Like what GPT spits out to you is good.
35:21
The reality is, is that, you know, sort of like human beings, AI is exceedingly
35:25
inconsistent.
35:26
So if you prompt that same thing, again, a great part of my journey, unlimited 60
35:31
bucks a month, so I could just pre-prompt this thing, you know, 20 times.
35:35
Eventually I'm going to find something I like, that's going to be a lot faster
35:38
and a lot less of a headache than me trying to, you know, prompt it.
35:41
And then I'll be like, Oh, well, this prompt worked perfectly.
35:43
But that was probably just the reality of you got a good output.
35:47
You know, luckily with that prompt.
35:49
So I would say your prompt is fine.
35:52
Unless you're like, obviously, you know, there's something really explicit in here
35:55
that doesn't make any sense.
35:58
Um, like in this case, if I wanted the copy to sort of stay on longer, what I
36:02
would do is I would go back to my GPT, I would download this as a GIF.
36:08
And then I would say, I mean, we can do it.
36:10
Um, I would download this as a GIF.
36:12
I would go back to my, uh, my GPT chat.
36:16
And I would say this was the output.
36:19
The copy disappears.
36:23
Um, can you correct the prompts to have the copy remain?
36:29
And so like, that's how I would prompt engineer, right?
36:31
I would take your output.
36:32
I would put it back into GPT and say like, this is what I got.
36:36
This is what I want to fix.
36:37
And like, this will make more sense when we sort of like do product placement.
36:40
Because like you sometimes you just like really need to isolate one component.
36:45
Um, and so in this case, like what I did for those listening is like, I, you
36:50
know, downloaded the GIF, I put in the GPT.
36:52
I said, this is the output.
36:53
The copy disappears.
36:54
Can you correct the prompt to have the copy remain on screen for
36:56
the entirety of the animation?
36:57
So now I'm going to go back.
36:59
I'm going to do this sort of the same process.
37:01
I'm going to delete our little double dashes that have AR 16, uh, you know,
37:06
the aspect ratio, I'm just going to make sure everything looks good here.
37:08
I'm going to add our starting frame with the built for series
37:11
hunters, and then I'm just going to re prompt it.
37:14
Um, so yeah, that's sort of the gift flow.
37:18
And like, there's so much you can do here, right?
37:20
Like this is your playground.
37:22
Uh, you know, if you were to sort of scroll down in here, I would say like my
37:25
team like uses, like this was, uh, an example of an ad we created in and a
37:30
banana, and then we wanted to add some motion to it, right?
37:32
Like made the fire or made the guy move around like, Oh, you know, cracking a can,
37:37
you know, what have you.
37:38
Like our team is in mid journey all the time.
37:40
It's just the easiest.
37:41
It's the lower low spirit entry.
37:44
And when I will get here in a second, but also if you need to replace a product,
37:47
it's a great way to create a good product backdrop.
37:50
So like, if I will say mid journey sucks at product rendering, like that's why,
37:55
you know, we don't use mid journey for anything product specific.
37:58
It's really for vibes or for creating gifts,
38:01
but mid journey is best as sort of step one.
38:03
So let's sort of, you know, move into sort of workflow too.
38:07
So next, you know, next sort of step in
38:09
the process, let's talk about sort of product
38:11
insertion.
38:12
So, uh, I was doing this before.
38:14
So again, like what I like to do is go to
38:17
Pinterest and it seems like, you know, for
38:19
Zach and his brand, like hunters are very
38:21
much a big part of, uh, you know, who he's
38:23
targeting.
38:24
So if I want to, you know, I did this before, but like Hunter, you know, wearing
38:28
socks, right?
38:30
And this was my sort of reference image.
38:32
Uh, maybe there's something a little bit better.
38:34
Uh, let's just say maybe we want something like a little bit more vintage.
38:41
Uh, I would say Zach's probably looking for something more like Merica.
38:45
Got it.
38:45
Okay.
38:46
You don't know what I'm saying?
38:47
Like, uh, like a bearded beefy man, you know, somebody like, somebody like Brad,
38:54
but with full facial hair that is like, that loves, loves guns and loves America.
39:02
Got it.
39:02
Loves his wife.
39:04
Maybe a little more of a gut drives a truck.
39:07
Loves it.
39:07
Loves a fresh, a fresh beer ski.
39:10
Okay.
39:11
Loves a burger.
39:12
Do you need more?
39:12
Definitely right there.
39:14
That's a back field.
39:15
This one left one more.
39:17
That was actually not both.
39:18
Both of those are bribed out.
39:19
This is, this could be our guy.
39:21
That seems like that's my guy.
39:23
Honestly.
39:24
All right.
39:24
So, so for those of you that are
39:26
listening, it's a, it's an older gentleman that, uh,
39:28
is wearing socks.
39:29
He's got jeans on and he's not smiling.
39:32
Um, at all.
39:34
And this is actually just so we don't get sued.
39:37
You know, this is actually not the image we're going to use.
39:39
So what we're going to actually do is we're going to, you know, we're going to
39:42
save this image into the void.
39:46
Uh, and then we're going to go into this little nifty tool called free
39:50
image, uh, to prompt generator.
39:52
Love this guy.
39:53
Uh, and now I'm going to upload that image that we just found.
39:57
Oops, not that one.
39:58
Here we go.
39:58
Here's our friend and I'm going to generate a mid journey prompt for him.
40:02
Right.
40:02
So obviously you can do this in GPT as well, but I find that like this tool,
40:07
you get five free a day.
40:09
It's just like, it gets it really, really close.
40:11
So I'm going to take this mid journey prompt that we just generated with
40:14
image prompt.org to go back to our friends and mid journey.
40:17
We can also look at these.
40:17
Oh, look, our copy is now saying the entire time.
40:20
Like, great.
40:21
We just tweaked that little thing.
40:23
And now we have an ad, maybe that one's a little wonky, but
40:25
you sort of get the idea, right?
40:26
Like we, all we did was we took our prompt, we put the output in and
40:30
we sort of cleaned it up a bit.
40:32
Okay.
40:32
So now I'm just going to take that prompt that we just
40:35
generated an image, prompt.org.
40:38
I'm going to do the same thing.
40:38
I'm going to do nine by 16, you know, version seven, whatever.
40:43
And then I'm going to generate that.
40:45
So this image that we generate is going to be our baseline image that we utilize.
40:51
Um, and what we're going to do is we're going to take our image and we'll
40:57
probably have to re prompt this because we need to be able to see a socks
41:00
more a little bit, um, but we're going to be taking this image and
41:05
this is going to be our baseline.
41:06
And another thing that my journey doesn't do so well is it like gets people.
41:11
People look too like Plato eat like they're too plastic.
41:13
It's like very clearly AI.
41:15
And so we're also going to be using, um, you know, seed dream to clean it up.
41:20
So you can sort of see this based on that image, this is what we got.
41:23
So I'm going to actually take one of these images.
41:27
I'm going to do exactly what I did before.
41:30
I'm going to go into our GPT chat.
41:33
I'm going to paste that in here.
41:34
I'm going to paste our initial prompt in here.
41:37
And this is an image of an older gentleman sitting in front of a bunch of guns
41:40
with then standing next to somebody with a flannel and he's downloading the image.
41:46
And now you're taking it into a different place into chat GPT.
41:52
For sure.
41:52
You went, you went Pinterest to the free, the image to prompt generator.
41:56
And then you went image, your prompt generator into mid journey mid journey.
42:00
You kind of miss the socks.
42:02
It didn't generate exactly what you wanted with the socks showing obviously.
42:05
And that's kind of what the product is.
42:06
So you're taking that back to GPT and you're asking it to show the socks better.
42:10
Right.
42:11
Bingo.
42:11
Exactly.
42:12
Oh yeah.
42:12
Nailed it.
42:14
Okay.
42:16
So let's redo this.
42:19
Let's see what we get.
42:20
No socks are the main event here.
42:21
Okay.
42:22
Nice.
42:22
You can tell we're getting some, we're getting some movement.
42:25
It's a real guy's guy.
42:26
This guy loves to crush Budweiser after work.
42:29
Yeah.
42:29
Yes.
42:30
I like what I'm saying.
42:32
He would, he, between the choice of buying a new truck or a older F one 50,
42:36
he would buy the older one.
42:38
Cause that's when they, that's when they, they made them better.
42:41
The 1995 whatever version.
42:44
Andrew, that might be you in like 50 years.
42:47
Yeah.
42:47
It looks pretty, looks, I mean, what we're
42:49
looking at here for those of you that are
42:51
listening is a photo of an incredibly handsome man.
42:54
He could be a model.
42:55
Actually.
42:55
Like a handsome older man.
42:58
And, uh, he, we're going to put some socks on him.
43:00
Uh, some, some hollow socks, of course, cause those are the best.
43:03
And because this model slash handsome man loves hunting and he likes to have the
43:07
ability to have his feet be dry.
43:11
That's where I've given him hollow socks.
43:12
Totally.
43:12
And it looks like we have a little, uh, some horse saddles in the background.
43:15
So maybe we also maybe want to prompt this one more time to add some, you
43:20
know, rifles or, you know, not a big gun guy, but maybe that's rifles.
43:25
I mean, what do people shoot with?
43:27
Those guns.
43:28
Yeah, I'm not sure.
43:29
Okay.
43:29
I'm not sure either.
43:30
I'm not, I'm not, I'm the opposite person to ask about hunting.
43:33
Okay.
43:33
Okay.
43:34
Um, yeah, me too.
43:35
I live in New York city.
43:38
Okay.
43:39
So let's redo this one more time and say, uh, and again, every time what I'm doing
43:43
when I'm refining my prompt, I'm just taking the image that we just generated
43:46
mid journey, putting it back into GPT five and saying, uh, in this case, I'm
43:51
gonna say in the background, I would like more, you know, hunting paraphernalia.
43:57
Nice.
43:58
Um, so I'm just going to redo this, you know, one final time and hopefully,
44:03
I mean, what's interesting about it to me is like, it's, it goes back and forth
44:06
between all these different tools and you're utilizing them to just refine
44:10
one another, um, which like you said, it's obviously going to be in one place,
44:14
but I think what's, what's cool about it is it actually is putting together
44:18
really good outputs that, you know, so if somebody, so let's say that I run
44:23
one of these with this old guy, well, just to be clear, you have to, you have
44:27
to say, this is created with AI in your ads.
44:30
If it's utilizing some like person, you have to disclose that
44:34
from an FTC standpoint, right?
44:35
Yeah, I believe so.
44:36
I think that is, uh, you know, that you have to say, this is AI generated.
44:40
Same thing you'd have to do.
44:42
Like if you use an UGC actor, right?
44:44
Like you'd have to say paid actor sort of same idea, right?
44:47
You have to just, just FYI to everybody out there that you do have to do that.
44:51
Yeah.
44:52
Yeah.
44:53
Everybody totally does that too, especially with UGC creators.
44:56
Oh, totally.
44:57
Everyone does this.
44:58
Of course.
44:58
And it's a hundred percent compliance across the board, but I do know that the FTC,
45:02
you know, we all follow that, uh, that lawyer guy on, on, on that, uh, or X,
45:07
uh, I mean, it's case after case of like, Hey, they, you know, they
45:11
utilize this decision to disclose this.
45:13
Like this stuff's coming out more and more.
45:15
And if you're a bigger company, that's scaling, you just really got to
45:17
see why to make sure you're okay.
45:20
For sure.
45:21
Okay.
45:21
So I actually feel great about this image.
45:23
I feel like there is, you know, we're sort of hitting the nail on the head.
45:25
We can see the socks really well.
45:27
So being able to swap the product out will be great.
45:29
He's got some camo gear, you know, he's got some buck horns.
45:33
Uh, he's got, you know, some rifles.
45:36
Uh, so yeah, I feel pretty good about this.
45:38
And again, you can sort of see like, if we zoom in on, on our frontier, like it
45:42
still feels like a little plasticy.
45:44
So that's another thing we're going to have to adjust for.
45:46
Like, especially these ones, these are much more offensive.
45:49
Like you can very clearly tell these are AI.
45:51
This guy's actually not horrible.
45:54
Um, but ultimately, you know, we will want to sort of clean that up.
45:56
So, okay.
45:57
Next step in the process, I'm going to download this image and here, let
46:00
me actually share my full screen here.
46:02
Um, so you can see me do this.
46:05
Okay.
46:06
So I've now downloaded this image.
46:07
What I'm going to do now is I'm going to, uh, is this our friend?
46:12
I'm going to open this in preview.
46:14
And then I'm going to go to tools annotate, uh, rectangle, and I'm going to
46:20
highlight justice socks, I'm going to do it for both socks.
46:24
Okay.
46:24
This is a mastery share right here.
46:26
I've done a lot of what you're talking about, thanks to you teaching me,
46:29
but I haven't done this, which is literally just creating a box around the
46:34
socks to say, Hey, this is what I'm looking for.
46:37
Exactly.
46:38
And the, the reason we're doing this is like, we actually won't use the box
46:42
when we're generating the image prompt, but we will use the boxes for GPT to
46:48
write a succinct prompt to ensure that nothing else changes, but the socks.
46:53
And so we'll sort of see what that looks like.
46:55
So we're going to go back to our GPT that we've been working with.
46:58
We're going to grab our little friend, the Hunter, big manly friend, sorry.
47:02
Um, we're going to put that in here.
47:04
And then what I'm also going to do is grab a photo of hollow socks.
47:09
So I'm just going to, you know, grab, uh, something without copy on it.
47:13
I guess like, is this, this is probably, this is the one that I'm testing for
47:17
hunters, so maybe I'll just save this one.
47:18
Mm-hmm.
47:19
It's a nice one.
47:20
Yep.
47:20
Orange and black, high top crew.
47:23
Yep.
47:23
Perfect.
47:24
So what I'm going to do now is we have our Hunter friend.
47:27
I'm going to take this, uh, hollow socks image and say, okay, next step.
47:35
Uh, please write a nano banana prompt to, uh, put the socks on, uh, hollow socks
47:48
in the prompt, please include all copy that appears on the sock.
47:53
Please note, we will not be uploading the red boxes.
47:58
So our prompt needs to be succinct enough to change the size of the
48:02
element enough to change just the element and nothing else.
48:07
Um, and then, uh, I think that's, I think I'm not missing anything there.
48:12
So let's do that.
48:14
Um, Oh, I will be uploading both the image of the elderly
48:20
gentlemen and the hollow socks.
48:22
Uh, okay, cool.
48:24
So now what I'm going to do is go back to our image.
48:28
Maybe I'll save this one as, or I'll just, you know, now we have this in GBT.
48:31
Delete this.
48:32
So I'm going to save this, save this.
48:35
Okay.
48:36
Cool.
48:37
This is for nano banana, not mid journey seven.
48:40
Sometimes they will think you're doing one thing, then another again.
48:43
We need to pinpoint and change just the socks.
48:47
Okay.
48:47
This is a little short.
48:49
I'm not super bullish on this, but we'll see what happens.
48:52
Um, okay.
48:53
So now I'm in a tool called fell about AI.
48:56
If you've ever spoken to me about AI creative, I've definitely
49:00
mentioned this to you, it's a Higgs field competitor.
49:04
Uh, effectively what it is is just like, it combines all of the AI models into one
49:09
place.
49:09
So it's just like very easy to, you know, use anything.
49:14
The only tool I don't utilize in, uh, in foul, actually to mid journey is because
49:22
obviously mid journey doesn't have an API.
49:23
So, uh, foul couldn't use it and Vayo three, Vayo 3.1, I guess I use directly
49:30
in Google deep mind, uh, and, and what they call flux because there you're able
49:35
to, if you pay like 120 bucks a month, you get unlimited Vayo three fast.
49:40
And in this case, it's sort of pay as you go.
49:42
Okay.
49:42
So now we're in fell.ai.
49:44
And nano banana, I'm going to put our prompt in here.
49:48
I'm going to add our images.
49:49
Oh, I don't know why this is showing up.
49:53
I'm going to generate four images because
49:55
again, I would say one out of five is okay.
49:58
Aspect ratio nine by 16 and I'm going to run it.
50:01
And so I guess again, for those of you listening, I did not include the little
50:05
red boxes around the guy's feet.
50:08
Uh, you know, the only thing, uh, the red boxes are used is to make sure your
50:13
GPT prompt is proper, right?
50:16
Uh, so.
50:18
Oh yeah.
50:19
Looks like he's wearing some hollow socks there.
50:21
Yeah.
50:21
But again, like this, like not a great first output, right?
50:25
Cause it just says hollow.
50:26
But as you can see, sort of nothing else changed.
50:28
So now I would go back to our GPT copy this image and paste it in here and say,
50:35
uh, we need to include more detail in our prompt.
50:39
Uh, as the output did not match the sock exactly, please analyze the skew image
50:50
of the sock and include that detail in the nano.
50:55
Banana.
50:56
You know, it, what's interesting about this is that it's utilizing all these
51:00
tools in combination with one another.
51:02
And it's also getting some really good outputs, um, very quickly.
51:06
Like we've only been doing this for essentially like 30 minutes and we have
51:10
you know, a number of things that we can use.
51:12
Okay.
51:12
So again, like not perfect.
51:14
I would probably want to prompt this a few more times to get the socks to be
51:17
exactly right, but let's just say we were happy with this so we can sort of switch
51:21
topics, but most important thing, not the most important thing, but
51:24
an important thing, this is just like a little bit of a hack is again, making
51:29
him look a little less plasticy.
51:31
Um, and that's what I use the C dream model for.
51:35
Um, so, uh, and we can include, well, by the way, for, for podcast listeners,
51:41
we can include, um, notes from you to make sure that people can like walk
51:45
through this a little bit themselves or the things that you've covered in
51:48
a couple of other ideas for sure.
51:50
Just so they can guide through them, guide through themselves.
51:53
Definitely.
51:54
And like, again, this is one of these things where you just want to do it
51:58
yourself, like that's the best way to learn, uh, you know, watching me do this
52:03
is probably helpful, uh, but it's only starting point.
52:06
So there's one very simple prompt to make people look a little less plasticy.
52:10
I put it in our notes over here, but it's just this.
52:13
So, uh, it's image underscore three nine eight four.cr two.
52:18
The, as I understand it, the C dream model was trained on Canon raw files.
52:22
So when you include the dot CR two within the prompt, it makes it HD.
52:27
And then I say, please upscale, uh, you know, this person showcase pores,
52:32
et cetera, and make them feel less plastic.
52:34
And so I'm going to, again, let's do four images, uh, and we'll, we'll
52:41
do portrait nine by 16, uh, see dream takes a second.
52:45
So I can sort of pause there, you know, show you guys the output.
52:49
It'll probably take a minute or two and, you know, we can sort of switch topics.
52:52
Yeah.
52:52
I'd be curious, like before we switch to
52:54
the operational stuff and come back to that,
52:56
like, why don't we just everything to be helpful to list out some of the workflows
52:59
that you have with the specific tools.
53:02
So like you started with the static to Jeff Jeff peanut butter,
53:06
static to get with mid journey.
53:08
The next thing you did was you, um, I mean, you kind of combined some of those
53:11
things, but then you also use nano banana for product insertion, fixing the
53:15
plastic nature of them is using C dream.
53:19
Is there anything else with those three tools that are worthwhile to chat?
53:22
And, or do you want to just like quickly
53:23
mentioned the other couple that you've got?
53:25
Yeah.
53:26
So C dream, uh, again, I would say it has two really good use cases.
53:29
One is like making people look like people and two it's image resizing.
53:33
If you, and, uh, it's a very simple prompt.
53:36
If you have like a nine by 16 image, upload it to GPT, they, you know, please
53:41
read all of the components of my nine 16 image, take that prompt, put it into C
53:46
dream and then change the aspect ratio directly in C dream.
53:49
And that way you can like, very, I think I did this actually ahead of
53:51
our call for, um, and I can show you this quickly for these guys, right?
53:55
So like, this was a nine by 16 image.
53:57
Uh, I said, please resize this exact at one by one square format while preserving
54:02
every visual element composition, text hierarchy, keeping the two black and
54:06
orange, hollow, you know, one labeled over the calf, et cetera.
54:09
Like what I did is I uploaded my nine by 16 image into GPT.
54:14
I had GPT read every piece of copy on it and then include every piece of
54:18
copy on the prompt as well as the different components.
54:20
So like, of course some of these, like, they're going to change a little bit,
54:24
but like, there we go.
54:25
Right.
54:25
That's a perfect, you know, one by one image of that exact same thing.
54:29
So like, if you just like don't have the time, energy, or effort to just
54:34
like quickly resize something, sometimes we'll just do this and it works very well.
54:39
Uh, the one sort of caveat is make sure you're using HD, the HD feature
54:43
directly in C dream.
54:45
Um, cool.
54:46
So now we sort of go back to our guy here.
54:48
Again, he lays looks a lot better, right?
54:50
Like he actually feels, you know, a bit more human.
54:53
Uh, like I think this is probably the best.
54:56
It's just like a great way to get like more
54:57
consistent coloring, um, across the board.
55:00
And so like, if I were to go, you know, one step further, my sort of final
55:04
step in this process would be taking this image, going to VO 3.1 and using
55:10
this image as a start frame.
55:12
And then I would, I would take this image as well, upload it, you know, back
55:16
into nano banana and have this man maybe, you know, hunting and then using
55:20
those as a start and end frame.
55:21
And now I have a great piece of B roll of the man sort of sitting in his shop.
55:25
Uh, you know, just maybe he's like moving a little bit and then it transitions to
55:29
him, like walking around with the socks in the woods, uh, actually like hunting
55:33
dough or whatever it is.
55:34
Yeah.
55:35
Or he's like pulling up the sock, putting his boot on exactly.
55:37
I don't know like that.
55:38
Yeah.
55:38
Yeah, exactly.
55:39
So yeah.
55:40
So like the, it's all you would really do is take this image that we just
55:44
generated in C dream, put it back into nano banana and say, using this man,
55:48
create XYZ scene, and now you have an end frame so you can create like a whole
55:53
piece of B roll with the person wearing, you know, the proper socks.
55:57
Yeah.
56:04
I mean, I think, I think there's a ton of ton to say here.
56:06
We'll appreciate your time.
56:08
Appreciate you being here.
56:09
Uh, we'll put a share your document to anybody in the show notes on
56:12
scalability school.com.
56:14
Um, it'll also be in the show notes on youtube.com forward slash scalability
56:19
school, or you can find us, um, or you can email brad at sexy at brad.net.
56:24
No, I'm just kidding.
56:25
That's not as email.
56:27
Can you email me at Andrew at foxwell, digital.com.
56:31
And I'm happy to connect to you and the resources that you need.
56:34
Will bang your episode.
56:36
Yeah.
56:36
Thanks for having me.
56:37
Really appreciate it.
56:37
This episode is brought to you by Homestead's email and SMS service.
56:47
Honestly, we've, we've done an episode with, with Jacob from Homestead.
56:51
He's the man and he's, he's got to be probably one of the smartest
56:54
email and retention operators, like on the agency side, hands down.
56:57
I look forward to seeing the tweets that he puts out basically every single
57:01
month with like a recap of the designs and as amazing as the designs themselves
57:06
are, it's the actual like tactical behind the scenes work where he just
57:08
like knows what he's doing and the team at home said like, really know what
57:12
they're doing on email and SMS probably, if not the number one, very, very
57:17
close to competing for number one retention agency in the DC space.
57:20
So just killing it.
57:21
Yeah, absolutely agree.
57:22
You know, if you are looking at your emails and you're like, these are terrible
57:26
and you know, you can optimize the flows and you know, you can do SMS better.
57:30
You got to look at the Homestead team.
57:32
So check them out, reach out and they can definitely help.
57:35
Make your entire email and retention and SMS program so much better.
57:40
And I'm not just saying that we've all seen the work.
57:43
I actually had a member who sat down with, with Jacob at the founder meetup
57:47
that we had, and he came over to me and said, I just talked to him for 10 minutes.
57:52
And literally he, he's revolutionized my entire email department.
57:57
So agencies are modeling themselves off of what Jacob is giving them for advice.
58:02
So anyway, onwards to hopefully working with them.
58:10
The only way that we grow this podcast is by you sharing it with your friends.
58:15
Honestly, like reviews kind of don't really mean anything too much anymore.
58:18
They're really meaningful, but they don't do a lot for the growth of the podcast.
58:21
Um, and so sharing YouTube links, sharing Spotify links, sharing Apple,
58:26
whatever we call it under the podcast app now, anything you can share the
58:31
better we're going to be guys, anything else you want to say on this?
58:35
Yeah.
58:35
Please go check us out on YouTube, rack up those views for us.
58:38
We'd love to see it.
58:39
And then subscribe.
58:40
Make sure to subscribe on YouTube as well.
58:41
And I relentlessly refresh the YouTube comments because it dictates
58:44
my mental health for the day.
58:45
So please say something nice about all of us.
58:48
Thank you everyone.
58:49
Thanks for listening.
58:50
Honestly.