Using Midjourney Video Loops & End Frames: Live Teardown

Rory accidentally finds himself on a nudist beach while Drew's making DIY sunscreen with AI. And if that wasn't crazy enough, this episode is a full live teardown of Midjourney video loops and end frame control—features built for creating cinematic AI video workflows.

Drew and Rory show how to use loops, start/end frames, and extended keyframes to build seamless sequences, plus what to avoid so you don’t burn through credits.

You’ll also learn:

✓ Keyframe Extensions – chaining multiple shots for longer, smoother videos

✓ JSON Prompting – precision timing and motion control (with live tests)

✓ Runway Act Two – motion capture updates and creative comparisons

✓ Midjourney Style Explorer & V8 Preview – what’s next for AI-driven video creation

Whether you’re a creative director, designer, marketer, or experimenting with AI video workflows, you’ll get practical prompts, iteration techniques, and creative hacks to level up your Midjourney results.

Watch now to see how these new features work, what to avoid, and how to produce cinematic AI videos faster.

---

MJ:FH Buddy (GPT)

https://chatgpt.com/g/g-68755521d2348191a5ea8f6457412d51-mj-fh-buddy

---

⏱️ Midjourney Fast Hour

00:00 – Intro & accidental nudist beach adventure

02:50 – DIY sunscreen & unexpected AI life hacks

07:00 – Midjourney video update overview (looping, 720p, start/end frames)

10:20 – Upscalers, Magnific precision, and V8 development focus

15:30 – Personalization codes & base model quality debate

17:30 – Custom GPT for Midjourney knowledge recall

21:10 – Mood boards, micro-styles, and avoiding “homogenous AI look”

24:40 – Style Explorer, aesthetic preference survey, and upcoming features

27:10 – Live first-frame/last-frame keyframe testing

38:30 – Loop functionality and extended multi-keyframe workflows

45:40 – Iterative prompting lessons and fixing motion quirks

53:30 – JSON prompting explained and social-ready video hacks

58:00 – Runway Act Two motion capture tests and impressions

01:07:30 – Sloth race cars, Trump in Lord of the Rings & other AI absurdities

01:09:40 – Key takeaways and what’s coming next

Episode Transcript

(00:00):
What's up everybody? Mid Journey Fast Hours podcast
episode 46, Rory Flynn, Drew Brucker.
We have Mid Journey video looping 720P start and end frame
all just launched within, I don't know, overnight.
Meanwhile, Mid Journey's kind ofbeen, I don't know, cracking

(00:23):
under the pressure this morning.A little bit hit or miss with
whether it works, but that's a temporary thing.
So I think we're going to talk, we're going to get into that.
We're going to get into all the new things that's coming, right.
So we still got like things likeupscale are coming with video.
We've still got some other components as they're working on
there. They've got 7 one, they've got
the 8 updates we can get into. Also some of the things that

(00:46):
we're seeing across, you know, XLinkedIn, Instagram with video,
right? Like we have runway, we have
video stuff. So I think we're going to get
into a little bit of everything today and we've got a lot to
catch up on because we've we've both been busy dads, busy dads,
busy career men. This is we'll, we'll consider

(01:08):
this our summer break. We promise Drew and I will get
back on a more regular schedule at this point.
I have been overseas for the last seemingly 2 weeks between
conferences and things like that.
So which was sick by the way, it's been, this is like the, so
the first conference I did two years ago and like the first

(01:29):
time I spoke on stage, it was anaffiliate marketing conference,
which is just like volume meta, Google creative, like it's just
a volume place. And they brought me in for
volume right for AI stuff 'causethey thought it would work.
And I was the only one talking about at that point.
And everyone's like, this is crazy this time everyone talking

(01:50):
about. AI yeah, a lot's changed.
To be to be expected, but like Iwasn't expecting like the level
of depth that some people had taken it to.
And I was like, this is sick to see like this.
Some of this stuff is really cool.
And like, you know, I always as marketers, we always bastardize
everything, but we also come up with some cool stuff.
So I was, I was amped, man. Like really some of the agents

(02:11):
that people were building like. Yeah.
It's really thoughtful, really, really useful.
Like not just building agents tobuild agents.
Like, really system, like systematic in nature.
But yeah, then I tried to go to,you know, Mallorca for three
days and just turn my phone off.I don't know.
How that? How'd that work?
That was a good idea or bad idea?
I felt great in the moment, but when I came home I was like this

(02:33):
is bad, this is awful idea. Like but dude, if you haven't
been to Mallorca I I would say put that one on the list.
That's a good one, huh? Really good, Barcelona good.
I'm excellent. Mallorca was a 30 minute flight
from Barcelona, tops. Nice.
Like get on, get off. Outside of it being like legit,

(02:56):
I do have AI do have one story for you before we get started.
This is, you know, for our for our audience here might find,
you know, Rory in an awkward situation funny.
Like we drove one day in Mallorca.
We rented a car and drove like 50 minutes to a public beach
instead of like all the the private ones.
And you know, we'd read about it.
It's like this long like 2 mile strip of beach epic, you know,

(03:17):
blue water, whatever we drive there, we get there early
because you know everything thatChatGPT told us, of course, our
Research Center. So get there early before the
crowds get there. All right, we listen ChatGPT
first one's on the beach 9:00 AM.
Also it like subtly mentioned that it could potentially be
like a nudist beach. So we're like whatever there's 2
miles strip of beach. What is the chance that we you
know, we hiked probably like a half a mile up on the beach to

(03:39):
find the spot. How did we get there?
And we're like, you know, 10 people show up like all right,
cool, like beach is starting to fill up and then like I start to
see bodies. Fine.
Come to realize we picked the one little stretch of the nudist
beach. So I'm I'm walking around and

(04:00):
seeing everyone's hanging out like legit doesn't matter what
age I saw more I saw more body parts that day then, you know,
probably available on the Internet at this point.
So it was a interesting, you know, you could imagine me just.
Any. Any Roy Flynn photos surfacing
around out there? I was, I was, I was, you know, I

(04:20):
was saying to myself, I'm like Isaid to my wife, I'm like, if I,
if the sun, if the sun saw any part of my nether regions, you
know, I'm, I'm going to the hospital.
That's like third degree burn inlike 6 seconds.
That sun is like African sun. But then secondary question is,
do you put suntan lotion on those areas 'cause that kind of
sounds like it's a crime? Yeah, No, no, I don't think.

(04:43):
Sorry everyone. I'm not even a sun, I'm not even
a sun, sun lotion kind of guy. I mean, unless it's like natural
stuff. I, I've like leaned into that
just because this is going to sound kooky, but like I'm a, I'm
a big believer in like the skin is an organ so it absorbs
everything, right? So like I don't want to put any
of these chemicals on me #1 so unless like I it's, it's funny

(05:07):
man. One of the one of the cool uses
honestly that I've gotten out ofAI is like I've made my own
suntan lotion. If I made it so I've got it's
tallow. It's no well, hold on.
Wait. So I've got, I don't know.
I figured what's it's aloe. It's like, is it tojoba oil?
Is that how you say it? Tojoba oil, almond, almond oil,

(05:32):
little bit of vitamin E and carrot seed oil if I remember
correctly. Wow, OK, you get the whole
recipe down. I got and I had like I already
had all the stuff, so I just mixed it up, you know, got the
got the, the right proportions from from AI and all that stuff.
It's been great. And then I did the same thing
with like face moisturizer. So going back to the tallow
thing, I was, I learned about sort of like the tallow thing

(05:56):
being really, really good for anti aging.
And so I was like lean, I was looking into that.
I was doing research with L alarms for that.
And then I was like, OK, well, what you know, I started to see
like promotion for then the moisturizer.
I was like, this is probably something I could just do here.
I already have pallet here. And so it gave me instructions
super easy. Again, it was like 4 or five
ingredients. Everything's organic, right?

(06:18):
Made it. I've been using that as on my
face moisturizer for the last like 2-3 months now.
So this. This was The funny thing was
that like there in that sun, we're we're close to the
equator. It was about 95°.
If I, if I didn't put the heavy duty stuff on, I was going to be
in the hospital. Like no joking.

(06:38):
It was out there for if you wereout there for 10 minutes, you're
you're roasting unless you have like the perfect, you know,
European like complexion, which is yeah, like pre tanned, you
know, I I was coming in after being inside for like 3.
Months, right? You just look at the sun.
You don't go out and experience the sun.
I was, I was like, this is I could feel it.
I could feel the sizzle in aboutlike 10 minutes.

(07:01):
Sizzle. We're in a bad spot.
It's like threw some grease in apan.
You know, Oh my God. Put me in it.
Unbelievable. But all right, enough.
Sorry for the sidebar everyone. I figured you, you know, you
might want to hear what we were doing while we weren't posting.
So, you know, making suntan lotion.
I'm on a nudist beach and there you go.
So. One of the 1001 of the 1000

(07:23):
things I've done with AI that I never would have expected, but
it has insane value that no one talks about because dude, you do
one of those things one time andyou're instantly thinking, man,
I'm just going to go ahead and start my own brand.
Do everything. Grow your own food, grow your
own ingredients. You know, look.
At a business right here. Anyway, let's get into it.

(07:44):
Let's get into the stuff you guys actually want to hear
about, which is let's go into office hours first, let's cover
like new updates, what's coming,all that good stuff.
So I'm going to bring this up. Let's share the screen.
We can look at it together, we can talk through it together,
blah blah blah. Here we go all right and I got
it blown up here. So we can really see this OK

(08:08):
features, right? We just got this again, we
alluded to at the beginning, butwe just got this overnight.
This came last night. So start to end frame control.
So we didn't we didn't have end there's now looping.
You now have 720P in there and then potential turbo.
I don't know if that launched ornot.
I don't know. So we'll we'll look into that.

(08:31):
This upscaler updates ongoing. So they're training that and it
looks like this other so video model and upscaler updates.
I don't know what the video model updates are, if that's
like an existing update to what they currently have or if that
goes into like one of these other different video models
that we kind of discussed in thepast.
The high quality 1, is that whatwe're thinking maybe?

(08:53):
Maybe. Man, I want to hear more about
that. To me, that's now whoa.
Random drop for them to just like kind of go silent on it for
a few weeks, but that was I'm here's what they.
Do that. They do that sometimes.
Yeah. Is an upscaler going to be like
useful? I don't know, like I find I
don't. Know yeah, I don't like 1 I
don't know man I because you could take it into cap cut right

(09:16):
and then it's like and you can also change the frames per
second it's not so choppy right it's like.
Yeah, you can use toe pads for that stuff.
I love, I love Astra for what itis.
I still think like there's no real magnific for video yet,
which is like the, you know, let's add this, let's add the
skin detail, let's add the little micro sort of textures,
which is what we need. But we'll see.

(09:37):
You think they're working on that?
It's A to me, it's like the biggest market.
Opportunity. I feel like that would be
something that they would be at least thinking about by now, if
not already in development. Just kind of like hush hush
maybe. Yeah, 'cause we've said this
before, like, you know, give it to me in in 480.
Like I don't really care. I don't know if, but if you have

(09:59):
something that can do it, like let's let's see it.
Magnific did drop that update with the precision upscaling.
It's good. I got a gnat right here doing
this live. There we go.
Yeah. Magnificent.
Yeah, magnificent. Like I played around with that a
little bit. The precision thing, which I
think is cool. I don't like, I don't know, to

(10:23):
me, I don't know if I've explored the depths of it
enough, but I feel like if I didn't like want the creative
part, right, like a lot of what I've done or I think other
people are probably doing too, is just slide the creativity
meter all the way down, resemblance up a little bit and
the image doesn't change, right?But there is some, you know,

(10:43):
refined controls on precision where you can add in a little
bit of grain, you can control the sharpness.
I guess that's maybe the big thing.
I played with both. But yeah, interesting update.
The only problem with with I find with Magnific is like if
you have text or graphic or something like on a T-shirt or

(11:05):
like logo or whatever, like if you use any sort of level of
creativity like what 01 even like it more, it changes it.
So it's like, you know, it's theonly pain in the ass for me is
like that kind of stuff. Like it changes some micro
details. Yeah, we shouldn't have to guess
so much, right? I just want to like, highlight a
part, either like mask it and belike, don't upscale this or

(11:26):
like, don't change this. You know what I mean?
Yeah. Other tools, but yeah, like
sensitive around certain part like especially with things like
maybe around, I don't know, smaller facial regions that you
just don't want it to more for hallucinate on Yeah, but.
Yeah, we'll we'll see what this one holds.
I. Think we'll see what?
I trust them now. Yeah, let's they've done.

(11:47):
Yeah, let's see what happens. I mean, like the the video, I'm,
I'm still really high on what wehave, you know, with mid
journey, again, it's it's very boiled down, almost like MVP
version, but I kind of like thatin some ways.
But I am curious about that, that more development package
that we were talking about a couple weeks ago.
They're talking a lot more aboutV8 recently.

(12:11):
V8 described as ridiculously good model.
I don't know why I just read that in like Will Ferrell's
voice. I don't know why it just like
popped into my head. I mean ridiculously good.
I don't know. Did they ever talk about V7 as
ridiculously good? I'm not so sure.
So like, I was thinking early days.
Did they late? They didn't at all.

(12:31):
They weren't really like it didn't seem like they were
excited beyond. Well, I elated.
Elated is the word. Yeah, this that sounds
encouraging. Some V8 features may backport to
create V-71 main team focus currently on V8 development.

(12:53):
What do you think about that prioritization with V8 knowing
what we know, I guess versus better things like improvements
around user experience and UI, kind of like just getting to
things faster without having to kind of know all the insurance

(13:16):
and out. Because I still feel like mid
journey is unless you know it, it's still probably one of the
harder ones to figure out in terms of like just image Gen.
creation. I'd say like like, you know, and
then feature, like I know we gotalready have a lot of feature,
but I'm just trying to think like, what about like efficiency
measurements, right. So like again, like I'll bring

(13:39):
up this idea that I think I brought up a few times a long
time ago. But just like, what if you're
trying to create a story or a film or something with
connecting pieces? And there's an easier way right
to based on a prompt to give youlike multiple scenes that could
go within a storyboard really quickly?

(13:59):
Yeah. You know, like stuff like that,
like I just, I, I feel like, I wonder if that's going to be a
priority. Just like what is VA?
Is it just around quality and coherence like the same things
we keep hearing because. Yeah, I think.
I think it's already good, I just don't like my
personalization code, you know what I mean?
Here's a weird thing. I, I for a training I was doing,

(14:20):
I went back because I had to take my personalization code off
because like I can't, like, I feel like it's ingenuine to show
what the images look like without with my personalization
code because that's not what they really are on the base
model, right? So like going to the base model
and like trying to run stuff andget it to where I want it.
Like I wasn't I wasn't getting any level that I was like happy

(14:40):
with not like V6 where if I before personalization.
Well, it's got no direction. It's got no direction.
Now, so I was like, I was like, huh, is this model really just
based on personalization? Because I'm just curious of the
base model for version 8, right?Like if that's going to be just
overall better because like I look at, you know, I look at
some of the other. I still don't, I mean like like

(15:02):
quality wise, like I still thinkflux and ChatGPT and everything
else like sort of looks, still looks AI still think I can get
the most realistic stuff out of mid journey with the most
detail. But like I'm not I wasn't.
I just got going back and working up the base model.
I was like, this isn't that giant of a leap from V6 point 1.

(15:22):
So I think they probably didn't.They say they for V7 they they
use the V6 data to train V7. I don't.
I don't get out. I don't remember.
I don't remember. But that was in like the waiting
days when we were like, when thehell is this thing coming out?
This is a good plug though. This is a good plug for this.

(15:45):
So what we did was we took all the transcripts from all
previous 45 episodes. Mid journey, fast hours.
There we go. And we also imported some just
like basic understanding of mid journey, like codify black and
white things right around parameters and this and that,
you know, features, controls, blah blah, blah.

(16:06):
And now you can go in here and you can ask any question you
want. I mean ask, ask something, give
me something. What does the quality parameter
do? All right, so let's run this.
Searching its knowledge. All right.

(16:35):
Would you look at that? Look at this.
Just a perfect little mid journey fast hours buddy.
So hey, hey, if you're looking for something we may have said
in the past, I think that's where I really like.
I want to make sure that it it does pull those things as as as
well as it can. I did this in the the best

(16:59):
possible, most most accurate waythat I can for this like initial
version. Obviously, at the same time, I
didn't spend hours on this, but I would love, you know, if you
guys have any feedback, we'll we'll drop this in the episode
description so you can use this if you ever need it again.
You can ask anything mid journeyrelated.
You can ask anything for recall based on past transcripts of
this podcast. I I like that it said what they

(17:22):
said on the podcast. Drew called one of those sneaky
mfers because people forget about it.
That's that's good. That's good reference knowledge.
Look. Even tested the F1 car shots,
remember that. Yes, that's weird.
That is. Weird.
Let's go. Man AI how?
About that anyway, we'll throw that in so you guys can

(17:44):
reference. I think that's like super
valuable. Anything you want to know about
Midge, you don't have to ask us.You don't have to go back and
try to find it, OK? Like it's there for you.
Throw that in your sidebar. Free to use and if bran over
you. Guys, if you guys missed Faye,
she's back and she's back with avengeance to very chatty.
Right now, I mean, we're she's overdue.
She didn't get it. She didn't get her say for the

(18:05):
last few and I think the last maybe few that we recorded she
wasn't around for. Yeah, now she's back and she's
she's one given 100% so and she's she.
Can say the word fish now. She can and if you're back and
if you're here for the first time, my daughter makes regular
appearances on the show. She's 14 months.
She has a lot to say, so just. OK.

(18:27):
I want to know when you're goingto stop doing the months?
Is it after 18? I figured 1 1/2.
Well, 1 1/2 feels like. You're not going to say, yeah,
like once you get to 18 and it'slike 1 1/2 or 18, but then then
you're not going to say you're not going to say 19 months.
No 1 1/2. Once it hits 1 1/2, I've
realized that's my. That's my yeah, yeah.
So. That'll be cool.
Yeah, we. Don't have to say months no

(18:48):
more. Kind of like, yeah, it's like
the same thing with the, it's like the same thing with the
dogs, right? You got to eat up to months and
then it's years and then that's it.
Yep, I'm with you. I'm with you, man.
OK, well, I mean, let's let's goback to let's go back to notes
here. We were talking about V8.

(19:10):
I asked you that question and I I think we kind of went off
track, but I am still curious about what you think here, you
know? We.
Expect them with V8. You know, like if we were to
think about what would be different.
All we've heard for these last models up to this point is
quality coherence. Still need to work on Omni
reference like it just needs to be better.

(19:31):
I like it, I think it's good, I just don't think it's as
consistent as it can be with theoptions that.
Are available there. Now, and regardless like it, it
gets me where I need to go, but I still have to like play with
it. Remember, and I don't, it's
probably worth revisiting too, but remember when we, when they
did launch that we had the episode, we'll we'll link that
too. But where we actually played

(19:52):
with the Omni reference, wait, but we also played with Exp,
Remember we were actually getting a little bit better
results with adding in just a little bit of Exp on there too.
Yeah, which is interesting. It's just like now.
Why do you have to do that? Yeah.
Which had GBT in context of runway references where it's
just like put this person in thescene and it makes it that.
Easy. I don't even think we need to
wait right? Like we don't even need Omni

(20:14):
reference waiting. I don't know it's it.
It should be so good that it just you don't even need to
wait. You don't even need to add Exp
Boom boom. Because that's the, that's like
the next step in getting towardswhere they want to go, right?
It's like they want to build real time or they want to build
like immersive worlds. Like then you need consistent
characters, objects, locations. And I'm not is that you can do

(20:36):
it. It's just not as straightforward
as it is in some other places. But I think like I'm looking for
I'm looking for the same thing that they always have done with
VA, which is the mid journey factor.
It's like where it's just, it's just more insane, more
versatile, more like widespread than any other tool out there.

(21:00):
So I don't even know like I haven't thought about when that
would launch, what I would think, like how I would react to
it. I'm just like so in it right
now. I think it's coming this year.
I mean, it's coming this year. Yeah, mood boards have been, I
think they improved mood boards again.
Like I still think mood boards is like the best thing that.
No one's no one's even talking about mood boards anymore.

(21:20):
Which is. Crazy because I use the micro
mood boards now like a lot like when I'm doing things like I've
been creating them as tokens more so than as styles where
it's like all right, I want likeoverhead view.
Like I use that as a mood board.I want like a harsh flash type
style. I want like this, that is so
it's more so like piercing together.
That with micro. Yeah, like fisheye lens like.

(21:43):
Not even. Not necessarily just thinking
about color. Yeah, more so thinking about
like the shot type or the camera, like if it's a super
wide angle lens, like with the fish eye, you know, or like soft
edges or things like that. Right.
Where, you know, you see all these people on Instagram that
are that are Hawking a lot of their products with like basic

(22:04):
like photography terminology. And then, you know, they're
$1800 for a list of camera angles and.
So this is not new. This is not new.
No, it's not new, but it's also,you know, you can just apply
that to a that becomes the style, like the fisheye lens
becomes the style. And then, yeah.
Prompted, I think we were when we were first talking about mood
boards, 'cause we did a few episodes on those.

(22:26):
Yeah, we were talking about thatidea of of breaking these down
like you certainly had like the color style sort of component
thinking about it a little bit more in terms of like style
reference mood boards. But then you we were even
talking about that. Yeah.
Like the the micro elements, youcould focus on specific

(22:46):
elements, right. So that could be to your point,
camera angle, that could be, youknow, the the type of lens, you
know why you know, we talking about close up, you know, you
could go all the way down the list of things that you would
typically structure with your prompt.
Yeah, and it's because. You can just keep keep trimming
the text prompting down to just like the meat and it doesn't

(23:07):
have to be all the additional pieces to get that creative
direction. It can be like, here's my
vision, like attach mood board and then X person doing this,
whatever. So I think it's, you know, it's
a different way to think about it instead of like always having
to write the text prompts. But I think produce the best,
you know, like I still think if you have a really good text
prompt, it listens. It's just getting there, you

(23:29):
know, and then like having the little additional pieces to make
it sort of vibrant and pop because look, here's a here's
one of the things I presented onwhen I was on stage.
I was just saying, like, if everyone's using ChatGPT and
everyone's using flux and everyone's using all these
commercial models that are like just readily available on every
tool, everything's going to start to look the same.
So you're going to have like this homogeneous sort of look

(23:51):
and feel of everything. We all know that ChatGPT has a
look, that yellow green look. The very specific like I can
pick out a ChatGPT. And so fast, so fast.
Out of a lineup very easily. Not saying it's bad, it's just
very easy if you don't like you know, that's that's the thing
with a hyper coherent model, right?
Like if you, I ran these tests where I was showing like one

(24:14):
word promise for each sort of image generator and like ChatGPT
is by far and away the worst. It looks like nothing, but it's
because it needs a shit ton of detail.
And if you don't give it to it, it just stays bland.
But if you give it to it, then it gets there.
It's just like you have to give it all the detail.
But that's why, you know, I think we have to, we still have
mid journey is going to have a place because things are going

(24:35):
to start to look the same if everyone's using those models.
Yeah, so far Mid Journey has done a really good job
navigating the the the space that they're in without BC
money. They're like the only ones, you
know, that that really aren't heavily backed that I know of.
I don't know about like other ones I don't like, I don't know
much about Higgs field in terms of like behind the scenes

(24:57):
business wise or or even RE for that matter.
But you know, Mid Journey, you got to admit, I mean, they've
been around for a long time and they've adapted really, really
well considering that such a bigconstraint.
Let's let's get into this like next piece of this.
So there's style explorer stuff,style explorer features coming

(25:18):
after the video updates. There's a survey active for
style, age, gender, country, aesthetic analysis.
Really. Let's go take a look at that.
Let's see what that is. I don't see that.
You completed surveys, rate ideas, demographic.
Well, there's your demographic, Sir.

(25:38):
Oh, you completed those? Yeah, but none, none of those
are recent. Those are all those.
Yeah. So, OK, you've got that.
This feature will explore how demographics affect aesthetic
preferences. OK, interesting API stuff.
I don't know if we want to go down that route or not.

(25:58):
They're talking about partnership philosophies.
OK. And then secret projects.
So still don't have any details on those secret projects yet.
Timeline decision. Pending announcements in one
month or a few months. Leaning toward faster
announcement schedule. OK.
That sounds like that sounds like when you get a snow in the
forecast where it could be 1 inch or 30 inches, Yeah.

(26:22):
That's like a thought in somebody's head.
I don't know what I should do with that info but.
Yeah, we should, We should. Should we test the?
Should we test the the 1st frame, last frame?
See how that looks live? Yeah, let's go for it.
You, you got it pulled up. Pull it, pull it up on pull it
up on yours for the only becauseI have to top up again.

(26:42):
Guys. I've I am so I am so in the mid
journey hole with the with my credits this month.
I have been this is my fifth topup.
So I'm I'm on relax mode and it's it's just bad.
If you're if you're listening tothis, go into here and go into
rate ideas and let's find we need to find give hours away.

(27:05):
Yeah, because I have something Ican give.
I'm not running out. I did get close last month.
I was down to like 19. That's impressive.
So I take it down to just like like baby for like a normal
person like 4. Months, generations.
OK, well, all right. So, so where we want to start,
man? We want to start with here.
What's just, I mean, you know, so if everyone, if you've never

(27:26):
played with keyframes before, just as a general sort of
consensus here or general sort of understanding 1st frame, last
frame means you have your first frame generated, which will be
able to start generating from that image and then your last
frame is where it'll end. So it's going to generate
everything in the middle. Now that brings up a whole
different set of issues. If you're trying to put two

(27:49):
different things in there, sometimes it'll combine that.
Like if you try to put like an image of like a basketball
player and then like an illustration of something else,
like it will somewhat seamlesslytransition that sometimes,
sometimes it'll just cut. A lot of times they have to be
related like the two images. So it has to be something that

(28:09):
is like, you know, I start here,car driving on a road and then
car driving around like a turn, right?
Like, think about it like that. If you get like something like
this. But Mid Journey's a little bit
more creative, so it could handle, you know.
That's the standard. Process with keyframes to the
right which says Andy. This is where I was having

(28:30):
issues with this. Oh.
My God. OK, where?
Where did my first one go here? I don't know if these are the
best 2, but at least they have the same style attached.
Geez, there you go. OK, that was harder than it
should be. Correct.

(28:50):
Then you have to add a prompt. Yeah, that's the only.
Yeah. And these were yeah.
And these were things I was justprompting for, for style codes.
So it's not like an accurate prompt.
So let's just put young. Is when they're doing.

(29:11):
So let's sync up something somewhat creative on the fly.
Maybe take out young because they've been super vigilant on
that's a yeah 'cause they, I've been getting flagged for
everything, even if it's nothinglike even, you know, it's not
like my personalization code where I'm asking for whatever we
were doing two weeks ago with the the, the horror, the horror

(29:36):
images and the masks and stuff like that.
But see suspiciously down the street.
Yeah, right. Just run that.
See how it see how it works. Anything else I need to think
about here? No.
OK. All right, so you have the
seven. 20 or 480P option there. Oh, good question.

(30:02):
No, I think it maybe. I wonder if that's the default
now. Let's see.
OK, no. Run it.
See what this is. Let's see what we got because
this is like I said, on the key frame side while we're waiting

(30:24):
for anyone who's not used to working with them, it's a lot of
times if especially if you're trying to do things with same
character or like the same objects or what's what not if
they're not exactly the same onefor one, you're going to have
like this weird morphing transition because it just it's
that's the way it is. So.
You know. It might be more fun for more
creative uses, but like that? Like that'll work like that's.

(30:50):
Well, I'm going to run it. I'm going to run it without.
Yeah. And then oh fuck, it didn't take
it. OK, cool.
Why didn't it take the last one?Yeah, yeah, yeah.
I look at this. It's get out of here.

(31:13):
Get out of here. I just had a white screen last
night for six hours. I couldn't even.
Get on. Well, at least you can kind of
do that cinema. Next shot of a detective from
behind she walks is eerie.

(31:49):
EE RY or EE RIE? I think it's EE.
RIE is our SAT word of the day. Let's see what we got on this.
It says cinematics walks carefully down they it'd be cool
if they added like an auto prompt.
I mean oh they have the auto prompt feature for the videos

(32:11):
but it's. Oh my God, can't do that either.
Wow. OK, All right, so let's do.
Let's see what we got here. That works.
Hey. That worked really well.
Like that second one in from theright the.
Middle on the right side, yeah. Just.
And this one, if she didn't walk, walk in the door backwards

(32:34):
and then we got motion high, motion high, kind of a little
bit funky. But again, like with this is a
very simple prompt. Like we get I, I don't think on
some of these you would expect the one shot, but we could
obviously get a little bit more.I got a perfect example for that
one. Go back though, so like.
Yeah. For specific details on video

(32:56):
right, for anyone using video like not the the one above it,
not that not that the one where the girl was like walking the
wrong direction. So see how the girl is facing
right And we have the prompt says Asian girl is walking
suspiciously down a city street.Often times this this will
happen all the time with the cardriving with people walking like
even if they're facing One Direction, if you if you just

(33:18):
type in walking right, often times that like corrects the
whole thing. So it's like these little micro
things with video that can screwit up.
So like that's these are like the things and you see it and
you're like, why is she walking the wrong way?
It's like, well, it has to be. It's like so literal when it
comes to video. If you want certain control, if

(33:39):
you don't want control, just want to let it ride, let it
ride. Like this is just if you're
trying to get something specific.
Yeah, that's a good, that's a good call.
Yeah, especially. So no prompt here.
This was just seeing what it would do.
OK. And then this one we do have a
prompt for. That one's pretty good.

(33:59):
Not bad. That.
One on the right, some people over here doing something.
It's like a bunch of people justlooking.
But yeah, so again, like, but you know what?
The other part of this right andI this is pretty obvious, but
you would want to look and see what things are wrong to, you

(34:20):
know, like you were just talkingabout to then go back and
iterate on the prompt with that's super obvious, but just
in case we need to call that outbecause you do that a couple
times, you should get what you want.
And I think that's the expectation we should have
versus like getting this right out-of-the-box, right?
It's all about an iterative game.
Same with like Chachi BT you know, there are certain things

(34:40):
where you would expect to get itwhat you want, but it's all
about the context. And sometimes with this
visually, you need to see what it's going to give you before
you can make that change. So.
Especially if you're doing something more complex, like
just knowing that you're going to have to go and iterate and
actually have a detailed eye forthe prompt to go find the
problem area and then go readjust it.
So like, did it. What is the girl?

(35:02):
Like that one worked better to me.
Let's see. Walking to the right, ripping
back. Here now still got to still got
to like still all of these have a little bit something wrong
with them going in the wrong side of the door where the
handles at right? That's why keyframes is hard.

(35:23):
I think everyone wants keyframes, but a lot of times
it's like things have to be so consistent.
I mean this this was just a re roll, so this was without the
additional prompt. This one works pretty good.
You know, walks right out of thescene.
That's cool. What?
What else did you re roll on here?

(35:44):
Just the other. This was the non prompt.
This was the one sneaking into the door.
Let's see do I have this one on low too?
I'd like to get this one on low so let's go prompt.
Should I lose my air loading image?

(36:07):
Yeah, I don't know. We're having some 7 some issues
here mid journey. But it's been been glitchy for
the last day so. Let's see what we get out of
this. So this was just rerunning with
motion high. This won't be a motion low.
Let's run this one again too. And we didn't do this one on

(36:30):
low, so prompt. So yeah, it's not taking in that
first image. So you're going to drag this
back in. Let's go low.
I've had issues with the 1st frame like adding things to the
1st frame. Seems last frame is fine but
whenever I try to drag somethinginto first frame it's just like
I don't know this is weird. It might just be glitchy because
it's it's one day old but. Yeah, yeah, This one, she's

(36:54):
going in the wrong side of the door where the handle's at.
So I like, you know, if like if we were to iterate on that, I
would probably iterate on that, walks into the left side of the
door, grabs a handle just to give it some additional
direction. Let's see on this one, this is
low motion. That one was was this one worked
pretty well. Yeah, that one worked pretty

(37:14):
well. This one right here.
Bang, there it is. And that looks like a 2D flat
animated style. Like that's perfect.
Even the shadow on the door. Yeah, notice, notice that.
Yeah. Oh yeah.
There you go. There's there's little things
that it does that are really impressive.

(37:35):
I I actually did AI did a VO3 like video on Trent like to
trying to do Transformers and then I ran the same prompt in
mid journey. It actually was pretty good
wasn't as realistic, but like itlistened.
It was a really long, like really long prompt.
And I was like, cool, it's like it, yeah.

(37:56):
Wasn't expecting that, but. Yeah, watch this 4th frame here.
Watch this right here. Watch this guy.
Let me get out of the frame. I can't be in the last frame
bro. It's like the CEO at the
Coldplay concert. Like oops.
Look at him. Look, I got to get out.
I got to get out of here. I didn't know it was on camera.
Oh, it's amazing. OK, and then you then you had

(38:17):
this this loop functionality too, right?
So now you can go into do, do dodo I mean, this is not going to
be a good well, that's not a good example.
I just ran these right before. I didn't even really I have a
chance here but to look at these.

(38:46):
Sidebar of mid journey definitely has some of the if
not the top one or two most likemicro micro detail movements.
Like there's little things that happen within this that are.
The car. The car messes it up, damn it.

(39:09):
It was a good, good shot though.Yeah, interesting, interesting
things happening. But that's now here, right?
So you can go into any image yougot and now they've replaced.
So this was used to be the manual, so it was auto and
manual here. Now they've got loop here.
Now you got to click into animate manually to do and it

(39:33):
and it takes you up here. So you have that.
But like again, we could do thisauto or this is where you could
really dial it in, right? So again, right, Like we could
come up with a creative idea here.

(39:53):
Click on the where creator actions it says more options.
Click on that Honestly if they. Have this is like pan and remix
and all that. All right, they didn't add the.
They didn't add like use as first frame, use as second
frame. No, no, that would.
That's a good call though. Yeah, that should be here.
You shouldn't have to drag it. Right, yeah, 'cause the drag is

(40:16):
annoying and then like it doesn't, it never like seems to
work well. I have that issue with describe
to just like never does what I want it to do.
Describe a. Little Infinity symbol.
Wow. Very, very UXUI efficient there.
Yeah, Infinity loop. Infinity loop, baby.
Dude, super cool. Like I I still just love it.

(40:37):
You can generate in any aspect ratio.
To me that's just like one of the best things because there's
nothing worse than having to go into a tool and it's like, oh,
it just doesn't fit. You have to go like reformat it
and then. It's pretty good without any
direction right there, right out-of-the-box.
See, look at the micro movements.

(40:57):
Like it doesn't like like just move to the side.
Like you get blinking, you get like, right?
Like that one? Damn, that one's a pretty good
one. I like that it's a good like.
Might be a good cover for one ofyour your Twitter threads.
Yeah, there you go. There it is.

(41:20):
This is with motion high. Too much going on here probably.
You know, this is like one of those things too, where it's
like, OK, think about the image you have here.
This is kind of a simple scene. It's one guy standing there and
you're not prompting it. So probably not a great fit for
Motion High, but we'll just see what we come up with here.
I mean it looks natural. Still kind of cool.

(41:42):
Yeah, like even just the lookingleft and right, that's like
looks pretty. I couldn't believe that.
Dope. Yeah, it's cool.
I'll show it. I have like 3 examples that I
ran this morning and that's it. I'll show just 'cause they're
they're more like disparate, like let's do the most random
things together. Agree.

(42:03):
Agree. Yeah.
So we did a flat, we did a flat.We did a sort of like more of a
more of a cinematic. Yeah, OK, let me let me pull
this thing up. I'll just, I'll just show you
what I already ran so I don't just continue to generate in
relax mode so everyone has to sit here for 15 minutes.
So like this one right? The start start frame is this

(42:28):
and the end frame is this right like?
So you're asking a lot here. Exactly.
So it's sort of just like transitions into the other
scene, but like. This is one of bad, right?
Like it's, I mean those are two totally different.
Yeah. So I think this is what you can
expect for this kind of stuff ifyou're doing it right.

(42:48):
Like I like that it just went into a tunnel and then it ended
up in. The I do too.
I think that actually was now give us a sense of like how many
times like this was something that you didn't manually prompt,
right? You just autoed it and then how
many re rolls did you have to doto get something like this?
I just ran it one time. I wrote.
Dream skating. That's pretty, pretty dope.

(43:11):
This one takes the guy's head and puts him on a boat.
This little head rides through over into here, but you know, I
was like, let's just test it out, see how it works.
And with two totally random things like, that's not bad.
Hell yeah. That would be on most other
tools that have keyframes, just a jump cut transition.

(43:31):
They'd just be like, here's the image and then we're over here
and that's all you get. So that would be this one more
like this one is more, but you get to see what I'm talking
about here with the car driving backwards because.
I didn't specify it. Right.
So started with like the close up of the F1 racer and then I

(43:52):
wanted to sort of like pull out.I didn't say anything.
I just wanted to see how it would work right.
So you can see like the IT starts forward and then we end
up going backward because cars and people if you don't define
direction, it just ends up doingthis a lot, which is weird, but
like you, you get some initial start, which is cool.
This one I thought performed better which was like starting

(44:16):
and close up on the skier and then going downhill.
Just typed in downhill skiing action shot.
Yeah. This one, maybe it was this one
that works. This one see, she went backwards
right? And then she gets into like her
little thing. This one might have worked
better. There you go.
It's a quick, quick transition. But if you have, you know, 2

(44:39):
images like this is what I mean,like these aren't so close that
it's going to create like a weird transition.
Like this obviously is a close up.
And then this is a little bit further out.
So you can sort of see like the natural transition from her face
mask down to her full body. But you know, sometimes this is
really hard for anything 2 totally random images and say

(45:02):
just make something, right? So that's and then if anyone's
looking to create like longer scenes, the little hack here, I
know this probably sounds it's like stupid to people who've
been listening to this podcast, but then you just flip and put
this as the first, you put this final image as the first frame
and then keep building from there.
So you could build like a sort of like an infinite pipeline of

(45:24):
stuff. So that's that's just a little,
you know, again, for maybe we have some new listeners are used
to hearing our our voices aren'tused to hearing face scream in
the background. That's, you know, some of the
stuff with video. But you know, I like these
things. I always like having if we could
do. I don't know if can you do just
last frame and not have a first.Frame I was wondering that I was

(45:46):
looking at that it looked like you could so.
I've always found that to be more useful to me personally
than just having two key frames.If you can do last frame and end
it somewhere. So.

(46:06):
Starting frame. What if OK, what if I do this?
And just run. That will it work?
Let's Nope. Nope.
All right. Damn.
Because that was a. What if I click?
What if I click? No.

(46:30):
Yeah. Damn.
OK. Hold on, wait.
Go to one of these real quick. Go to click into this, just like
one of these that you just generated this, this, or
actually with the 2 with the twokeyframes.
Sorry, click into one of these and then hit.
Does it have an extend option? It does.
So click click hot like one of these.
I just want to see what happens when you when you hit this if it

(46:51):
if it gives you an option to addanother frame.
Like with you talking about extend manually and see.
Because we're going from image 1to image 2, all right?
So it gives you the starting frame, which is where it ended.
And now you can add another one,all right?
So this is something here. So you can now add a different
ending frame and now it's going to extend like keyframe to
keyframe to keyframe to keyframe.

(47:11):
Does that make sense? Yes.
Yes. So let's go back into here.
Let me see if I can find something that looks something
close. I don't know.
I like this image. Let me just take this out.

(47:34):
Let's run. Still pulling that last.
Why is it still pulling that onein?
Weird. It's weird.
So that's like another glitch currently.
Let's try this with high, OK? So now you should have 10 second
video, basically three key frames.
So if you can build. There we go.

(47:56):
Right. Like we had our first image,
then our second image, and then at the end of the second image,
it's like, let's add a third image.
So you can keep going this way, which can be super useful if you
know what you're doing for longer shots.
That's good to know. I didn't even.
Think about, yeah, yeah, that's a good call because you're
seeing it right here. This was the end of that frame,
right? It's not pulling back this one.

(48:17):
This is the end of that frame, so.
Now we'll give you the full 10 seconds.
Here is my is my question. Is it just going to give you the
next one like in one shot? Yeah. 9 seconds there it is
that's good so this is good to know I now now I have something

(48:39):
to play around with now this canbe now this is like where the
the mid journey like creativity factor comes in.
You know string 5 random. String them together.
Yeah, Key. Frames together and see what it
can do like that should be there's a whole like little
format there. Well, you guys might be
something on Twitter later there.
We go, yeah, there it is, there it is.

(49:01):
All right, so we got to walk into the car, back Outback and.
Back. Oh, there we.
Go, yeah, hold on. Let's go check these out.
This is kind of a funky one to begin with.
So we're not looking for and we provided no direction.
OK, interesting. Yeah.

(49:22):
But I mean, this is probably a really tough one.
Tough man. What are we asking for here?
But yeah, let's go. We'll try the the fat, the high
motion too and see what what happens.
OK, All right. So kind of the same thing across
all four high now. OK, there's definitely a whole

(49:46):
format to be had with. US.
Yeah, yeah. You can.
You can do like I'm already thinking of like people walking
through. But this is right like the and
and I think it starts with the intent.
The intent around these images wasn't with this in mind.
If you already have this in mind, right, you're already
going to have better connective tissue between the images to

(50:07):
begin with. These are just using the same
style code, right? And we're just trying to pick.
But that's, you know, to the point of like creating a
storyboard or a flow that would just kind of come.
If you have that intentionality up front, you can build these to
be a lot closer. Or even if you've got 2 to start
with right, you can build on it from there.
But because I'm just thinking oflike, you know, just like a

(50:27):
standard workflow, like, all right, let's just take if you
have like a person honestly would do this.
It's just like an idea, like youhave a person like walking side
profile, like walking in an environment.
You could like take the, you know, go into editor, remove the
background, generate new environment, keep the person
Exactly. The same you could have like you
could take this girl, right? You could take this girl and

(50:48):
maybe she's walking entire down an entire block and through
different scenes, right? Like just continue.
It's like this infinite walk of,you know, like somebody walking
the New York City streets, you know?
Definitely, that's how. Block after block after block
after block. This is just continuous runway.
It's this. Is that like that actual like

(51:11):
keyframe piece now on the extend?
Makes it a lot more interesting to me.
Yeah, because I didn't think about it when I.
First, yeah. The other thing was we could not
give it an end frame 2 and extend it because we didn't do
that. We went here, now it should.
Now we can just extend it without it.
Yeah, I wonder what's happened to that guy who disappeared.

(51:35):
Yeah, I don't know. Is he gone for good?
Is he coming back again? Is he going to sneak back in
there? He had to get out just in time
for the end frame to conclude tothen jump back in for the
beginning of the next. One, he has his six seconds of
existence in in this reality, and then he's gone, no.
He goes, oops, they need me. They need me one more time.
They need me one more time. But yeah, so that's, that's I'm,

(51:58):
I'm intrigued now. There's now there's time for for
me to play with this because I have so much stuff keyframe wise
from what I was doing with Runway back in the day when it
was like keyframes like the coolest thing ever.
And now I want to go back and test some of the old stuff, even
style code stuff. It's really, you know, super
easy to transition from stuff. Make it a little more Morphe

(52:19):
again, there's endless directionhere.
It's like funneling your funneling your ideas to like
actually get to some point now, which is so much harder.
Creative. Is on full display.
Yeah, you could. Couldn't you like just
hypothetically to you could justdrag now you can't.
You can't drag in style references into the video

(52:41):
prompt. No.
You have to use style codes, right?
I don't know what do you what doyou mean by that?
Well, because I'm, I'm saying like if I wanted to then take in
this, is this from the same thing?
But like there's no way to drag this into an S ref to apply to

(53:01):
future frames, right? To just like dial in the
consistency of the colors. Yeah.
But if you're generating scenes with a style code, you can
obviously hack that. But I'm just thinking out loud
with that. Yeah, because it's you can take
sort of these again, you know? Whoop, I'm back.

(53:26):
I love that. So all right, I mean, I think we
I think we pretty pretty well covered the the video situation.
What's going on here? We'll probably go a.
Lot more, yeah, this is this is just launched.
So we we really haven't had a a real good spin at this.
Would love to hear what you guysare coming up with as well, if
there's anything that we maybe missed for this initial launch.
But I think we're going to we'regoing to play with this a bit

(53:47):
between now and the next time we're on here.
What else you want to talk abouton here?
Did you want to go into any of the the stuff that we're just
seeing on social right now? I know like the Jason code
stuff's just been like popping off, but you know.
Runway Act 2 like. Runway act two, yeah.
If anyone's not familiar with Runway Act 2, it's the advanced

(54:07):
Motion capture. It's a it's an improvement.
You don't know how how aggressive you can get with it.
It definitely, it definitely works and it works really well.
I did, I did like a stupid video, sort of like the reverse
of how it would be. I had Jack Nicholson acting and
like guiding my performance. So it was like it worked really
well in that, in that sense. But you know, I think this Jason

(54:30):
stuff is getting pretty interesting 'cause you see some
of the stuff, it's some of it's pretty close to production
ready. Like depending on what you're
looking for, like that's a. These are really cool.
I think it depends on, you know,it depends on the brand for
sure. You know where like this, this

(54:52):
is fun social media, you know, fodder and and like I I think
maybe where is your struggle is if you're looking for like
product specific level of detailand no morphing, I don't think
we're there, but. You're looking for jumping on
this. Storytelling through the use of
video. Yeah.
What? Yeah.
Yeah. Like this?
Is like a brand right now I'm doing, I'm pumping this on

(55:15):
social media like whatever my product is, I'm finding fifty
ways to just like get impressions and attention
because why not? It's just, this is fun, you
know, But it opens up sort of like that conversation around
Jason, which if anyone's not familiar with Jason, it's just
computer code. Oh, there's some.
There were some cool examples that too I found last night that
were kind of fun, and it was in this thread here.

(55:39):
This is just kind of. That's great.
Hold the door and then there's like, I've seen this, you've
probably seen this one. Yeah.
Tasty burger with the tongue. It's just so.
This one is crazy, dude. This one's disturbing.

(56:01):
Unreal. Whoa bro that one, that was
wild. That one called.
That's got to have like millionsof impressions.
That's that's so social media. What is is that Red Bull?
It's like, I don't know, yeah, maybe.
I think it's just like differentthings as waves coming in, like,
yeah, there, yeah. So I think so there's Monster.
Yeah. So I think that maybe the first

(56:22):
one was Red Bull. That is cool.
This one was this one first I was going to show you this one's
great. Is this someone giving the
middle finger? Yeah, Yep, that's Higgs field.
I know that's a Higgs field pre programmed shot.
Yeah, that's wild. Yeah, that's fine.

(56:43):
Then you just, you know, you getto this point and then you just
hit reverse, then come back and reverse the video straight
together. What is this koala going to do?
Oh wow, even got the thrown the floss.
On Hey, hey, hey, hey. Whoa that's good good after it

(57:05):
yeah I've I found I I I can't remember.
I found this thread last. I was like, there's some there's
some good ones in there. It is good.
Yeah, man, there's you know, there's a lot of this the Jason
stuff that's taking off. Let's see if I.
Can throw throw up my oh you have Sydney Sweeney who the the
second part of this commercial is is VO3 yeah so walks into the

(57:26):
car and they didn't even try to hide it like when she but it's
not. Super, super obvious.
It's like, right? Boop when the switch is into a
different I thought I thought I thought it was when she.
Jolted when the car jolted. Yeah.
And it's just like, all right, Iguess we're doing that now for
American Eagle or whatever company did it.
Then it took off. If you go to, if you go to my

(57:49):
profile real quick, I have the The Transformers 1 was the most
impressive to me because that's a really hard, hard thing to do.
You get a, you get a Scooch by Weevie, the weevie tutorial.
That's a thread up there. Was I did like that one.
That was a great one. Then there was this one where I
was like, yeah, I was like, OK, we can even like with the Jason

(58:13):
prompt, you can control sort of like what happens in sequence,
too. It's like arms come up, body
folds into each other, legs unfold, you know so.
Let's go to your prompt on that.Where was the prompt?
Oh, here it is. OK.
Little more. It's it's super detailed, right?
But you can have it as like one to four set one to two seconds
does this 2 to 4 seconds. Right here 6 to. 8 seconds does

(58:36):
this, yeah, so you can have morecontrol over it.
It's also not. I'm not writing this, I'm asking
Chad GBT to write, then I'm editing it.
So what are you? What are you putting in to get?
I guess is is it a template thatyou're using or are you just
describing all the things that you know for sure you want in

(58:58):
it? So I'm asking you for and I.
First off, I'm giving it an idea.
Or do you want to show it? Do you want to show it?
I don't have the I don't have the.
It's so buried because I did like 700 of them to go find this
specific one. But really what I did was
essentially like, here's my idea.
I want to show a tractor trailerturning into like Optimus Prime

(59:19):
and Google VO3. Hold on one second.
Sorry. Then it's like break it out into
these categories. I want like the colors, I want
the VFXI want, you know, the motion transitions.
I want it timed like a specific way, right.
So think about it like this. Like I'm, you know, like a scene
in Transformers. But like but like this, right?
Like you're not. Putting it all in, that was all.

(59:43):
That was all that. But I mean like what?
Yeah, but I mean like what are you?
What are you giving detail wise to then give like what are you
giving context wise to get this detail right?
Like even with even with the seconds?
I mean, I mean, I'm, I'm guessing you weren't thinking
about the seconds were. You well, after I read this was
this was like after like 10 iterations, right?

(01:00:05):
Like this doesn't happen on the first shot.
I was like, oh, let's let's run this.
And then if the first thing thathappened was it was like the the
truck would just like snap transform like there was no sort
of gradual process. So then I'm feeding, I'm
downloading the video, feeding it back into ChatGPT so it can
re you can watch the video and I'm saying I don't want it to

(01:00:27):
snap transition. I wanted to like gradually
transform like this was an earlier one, but then this one
where I was like, all right, let's have the arms unfold
first, then do the torso, then do the legs.
And then it's like, OK, like I got you.
Then it was more so like, OK, let's keep going with that kind
of idea, like let's have the camera push in and sort of like

(01:00:48):
ride up as it comes in or let's have it pull out or let's have
like, you know, change the sounds to more.
But I mean. Piston hisses.
I mean, where? Come on.
I was saying that I was saying that though, I was like, like
match the basically like match the the tone with like what's
happening mechanically. So you know, it's it's mostly
I'm just having a conversation with ChatGPT based on like

(01:01:09):
generation, like change this, change this, what about this
keep this the same change this. So you know, I'm uploading
things. I was I uploaded probably like
15 videos. So this is like, yeah, these
these are iterations stacked with compounding learnings along
the way, kind of what we were talking about with, you know,
we're just talking about with the the mid journey video,

(01:01:29):
right? You're, you're running it first.
You're seeing what you know, what flaws exist.
You're also inverting that. And you're also seeing what
things could are missing and could be added at the same time,
right? Yeah, because you got to, I
mean, you have to look at this as it's not going to be one
shot, one shot deal. Now the Corona party in a

(01:01:50):
bottle, right? Like the one down below this.
That was a one shot, which is why I was so crazy to me that it
just worked. And I was like, that's cool, you
know? Like what?
Now walk me through how you one shot at this.
So Selma, if you go to I don't know if you know Selma on a
right there. I tagged her in it.

(01:02:11):
Selma some Abu Akar right, the right there in the prompt right
above the prompts. So right, But her her profile
wanted to shout her out for the credit because she dropped the
prompt that she did for the IKEAad that she made.
And I just took that gave it to Chad GBT and said, like, let's
do this for, you know, like Corona party in a bottle.

(01:02:33):
Let's take this. Yeah, yeah, yeah, yeah.
Use this maybe same framework like template, Yeah, Jason code
framework. But instead the idea is this.
Yeah, There was also, I think itwas maybe dust in Hollywood,
who's on here that that threw out another prompt structure for
anyone that's that's interested.It's called like Mars or
something like that. Mars.

(01:02:54):
And it's like super detailed in the same way, but just not as
code language. See Dust in Hollywood.
Maybe he put it up on here. Let's see should be called like
Mars prompt structure something like that.
MARSI think it was like there. It is this called Mars prompting

(01:03:21):
modular action rendering syntax just again like a very well
defined structure that you can utilize on ChatGPT because it
gives it it gives it a very goodlike format to work off of and
like covers everything. So there was, you know, I was
trying to recreate some old stuff too.
Like remember the old Budweiser commercial with the frogs?

(01:03:42):
Budweiser. You know, like the.
Yeah, of course. Dude, come on.
Just trying to see if we could do that.
It's really hard for whatever reason in VO for it to get like
3 different things talking at different times, like staggered
and that probably like 20 times.Like I can't waste any more
credits on this. So you got to cut this one off
at the knees before I spend my whole months of credit on on

(01:04:03):
trying to figure out if I can recreate this commercial.
But dude, like this stuff is this right here.
If you're watching, screenshot this.
Yeah, I'm, I'm going to make sure I bookmark that too.
I think that's, that's dope, right?
Because yeah, again, that's that's the way I would think
about it is almost like creatingthese templated structures that
you have and then you apply new ideas to them, right.

(01:04:26):
Yeah. Because if you want the detail,
you got to write a lot like. Granted, you know, you can just
feed stuff into VO and just giveit minimal.
Direct. This is kind of the hard part
really. And even that right?
Like there are plenty of sites that like film makers use
photographer you like. There was one in particular, I
can't think of the name of it, but it's got like real movie

(01:04:47):
scenes on it, you know, and it gives you all this, it gives you
all these details in the caption, right?
What camera was used, who the director was, the VF that you
know, all this stuff. So it's essentially like you
could take that and build out this template.
And then again, maybe you have afew different templates that you
use and then you apply your new idea and then you've gotten that
new stuff, right? And that kind of, that was the
same thing with what we were talking about with just image

(01:05:09):
Gen. you know, and stills and all that, right?
It's just like you get, you needto have like there are certain
things you need to probably be thinking about that give you a
little bit more consistent results, right?
Is like the type of photography,like the type of style it is,
right? The camera angle, if it matters
to you, you know, like all thosethings that we talked about.
But yeah, this is good, this. Is the best way you can go build

(01:05:32):
AGPT off this just a starting point?
I feel like everyone needs to have a starting point because
video is iterative. I don't care what anyone says
like it's iterative. You're going to get 60% of the
time you're going to get awesomestuff 40% of the time it's going
to be trash. I think you're going to have to
like figure out what the problemarea is or just rerun it, but.
Oh, let's go back to let's finish up with the Runway 2 here
real quick because I know you were playing a little bit around

(01:05:55):
that. I haven't played around with it,
but I've seen plenty. Runway Runway to.
Are you are you blown away by it, though?
I would say like my impression is I wasn't it like didn't feel
like a huge jump from from one. I guess I was like, I'm not

(01:06:17):
trying to knock it, but I just. Like full video to video with
mocap. But it wasn't that.
It was more so just like the expressions are better like this
one. If anyone hasn't seen this, we,
you know, we sort of made a little mock of step brothers
here where I placed Drew and I in a scene using runway
references, you know, just just to shout that out here.

(01:06:37):
And then, you know, dubbing overwith what we're doing.
This was the more impressive thing, like having Jack
Nicholson drive my performance. So just giving it a picture of
me and uploading a video of him.But you can see like the the
hand motion, the face motion, the lips, the eyebrows.
You do look like you've had someLooks like you've had some work
done. Yeah, a little plastic, a little

(01:07:00):
smooth, Yeah. But like that's the.
Yeah, that is that's. Incredible man.
That's. Incredible 'cause it was good,
but you wouldn't get in the handmotion in the last one.
I think that's really like what it it did.
Like some of the stuff it adds the hand motion like in our
video, it did the high 5 which in act one it would not have
done the high 5 or have like anyof the hand gestures which I

(01:07:23):
think make things look real. The you know the days of the the
hey gem. You're right, right?
In a robotic voice. Dude I love.
I, I, I thought this one was pretty awesome too.
We're not going to be able to get the volume on it because we
both got our headphones in, but this is worth a watch too.

(01:07:44):
This one's great. Sloth F1 racer Hawking sleep
products right? And it just I, I just wanted to
have some fun. I needed to like dump all of my
my trauma of work from last weekonto into a video.
And I'm like, I'm gonna make this something really dumb.
And I had. Love the idea.
So like random? Yeah, Sloth, Sloth sleep

(01:08:06):
product. I was, you know, I was at an
affiliate marketing conference. A lot of it's supplements.
I'm like, oh, that'd be fun. Like what?
What character could I use for some of these stupid
supplements? And I was like, oh, sloth.
Because he already goes to sleepand it's like, let's make it a
little bit more ridiculous. Let's make him do something fast
as a profession. And then I, I like the, I, of
course, love the, the Trump one.That's great.
That one was too good. That's a whole genre to be

(01:08:27):
unlocked is the impressions. And then like doing it, but you
know, this one for the context here, there's this guy that did
a impression on on on Instagram.And it was.
Who is it? Who?
Call it his name is Nacho Nacho Hombre or something.
Like that. Oh, interesting.
OK. I was thinking like, there's
another guy I really like his Trump.

(01:08:48):
It's like Austin, NASA or something like that.
Something like that. But it was it was about Lord of
the if he was like Trump in Lordof the Rings, they'd I just want
the ring, you know, I I don't need it, but I want it.
And it was I was like, screw, let's make Trump, you know, a
Lord of the Rings character and just put it on there.
But that's a whole, that's a whole treasure chest to be
unloaded with the impressions that are done on.

(01:09:09):
Maybe we maybe we get into this weevie thing next time.
We should get into that. I think that's, that's a cool
little unlock. We could, we could walk you
through that. Rory's been going pretty hard on
that. I would also selfishly like to
learn a little bit more about that.
I think that's a great little unlock as well when you're
talking about character consistency, product stuff,
right? So I, I don't think we're at the
stage yet where one tool feels like it can do it all, where you

(01:09:31):
don't need other tools yet. We're not there.
So that's just kind of where we're at.
We need to stack some things. And that's why it's so hard to
keep up is you kind of need to know more than a few tools,
right? Like.
That's why Weavey's good, that'swhy we have access to everything
except Major and EP of that. But maybe Vapi, who knows, maybe
they'll have be one of the ones that gets it.
So that'd be cool. But yeah, man, that was it was

(01:09:54):
good catching up. Cathartic, right?
It was it's been a couple weeks.We we were busy and this felt
good and always like right afterwe get back on after we haven't
been on for like a two weeks or something like that.
We're just like, oh man, that's long overdue, needed.
It needed it. We'll.
Be, but I'm glad we I'm glad we also got a mid journey update in
between 2:00. I mean, we didn't have a lot of
time to really think on that, but I think there'll be some

(01:10:18):
things that we can unlock further next time.
Also, I think there's like an opportunity.
This is maybe a little bit late in the podcast if anybody's
still on here, but I think the other opportunity too is like
really refining prompts based onthe video, you know, capability.
Now, I haven't really seen too many people sharing that, right?

(01:10:39):
Like I know on an office hours call, well, I haven't seen
anybody show that on an office hours call.
They were talking about, you know, like somebody had asked
like if they're going to includelike, you know, filmmaking terms
and blah, blah, blah. And so it like, I don't think
those are baked in. I but I haven't tried right
like. It works.
Does it yeah. So I mean like you know there's

(01:10:59):
a whole thing with that that I think is still to be unlocked
and talked about more if you're if you're really serious about
leaning into video for that. I think up until this point,
video with mid journey's kind ofbeen a little bit more of the
fun thing and like you know, lowlo fi in terms of like the lift
right, It's like like maybe likethe smaller projects, maybe like
the quick hitters, you know so far, because we know that it's

(01:11:22):
sort of like the MVP, but. Yeah, it's it does stuff like I
put that Transformers prompt in there and it worked like it did
what I told it to do. It was it looked didn't look as
realistic as the stuff from video, but like still that's
that didn't work in any other tool.
It's been funny trying to throw these Jason prompts.
I've just been testing him and Sora also just to watch the

(01:11:44):
thing just crash and burn It's. Oh yeah.
Like man, what a flop of a tool compared to everything.
Else hey, don't throw Jason in mid journey by the way, you
might be running like 1000. Iterations because I learned.
Guess what? Permutations, baby.
I forgot to take some of them out.
I thought I'd gotten them and then it was like running like 62
variations. I was like, no.

(01:12:05):
I just thought about that. I was like, dude, don't do that.
Don't do that. There it goes.
Yeah, but that was that was one thing that I would I I
definitely I covered of the I covered of the, you know, the
curly braces giant. Little curly braces.
Yeah, whoops. There.
We go, there goes all the fast hours.
But yeah, dude, there's going tobe a lot to talk about over the
next coming weeks. I am certain of it.

(01:12:28):
Good to see you, man. It's been good to see you.
You look handsome. Handsome as ever, you know.
Yes, so. If you're still here, you know
the deal. Like, subscribe, Tell your
friends. Like we'll see you in the next
one. Tell your dog.
You'll see. Tell your mom.
Tell your yeah. Tell your mom I.
Don't know, tell your mom we're,you know, we're wholesome
individuals, fun to watch. Yeah, yeah.

(01:12:52):
But yeah, man, we'll, we'll catch everyone next time.
See you guys later.