Adobe just changed the creative AI world

Adobe just changed the world of AI

creativity. By the end of this video,

I'll explain why. There's a brand new AI

image upreser that has better quality

than any tool we've used up until this

point. And Miniax 2.3 is here and the

quality looks really good, but how does

it compare against VO and clink? Well,

we'll find out in this week's episode of

AI Film News. Thanks for joining. Now,

before we get going, I want to say thank

you so much to the team at Adobe along

with the members of the Curious Refuge

community that said hello at Adobe Max.

We heard so many incredible stories of

students who went through our training

and landed jobs afterwards. I really

appreciate everyone who came by to say

hello. Okay, cool. Let's get to the

news. So, of course, we're not going to

bury the lead. Let's talk about the big

news coming from Adobe Max. There were a

ton of updates from the Adobe team, but

I think the first one that I really want

to focus on is the fact that Adobe has

officially planted their flag when it

comes to artificial intelligence. While

they were a little hesitant to talk

about AI a few years ago, this year's

Adobe Max was almost exclusively talking

about the latest AI tools on their

platform. Now, the most notable update

that I think really has a seismic shift

in the way that we approach our

creativity is the fact that Adobe has

gone allin on being an aggregator that

brings various AI tools together using

Firefly and various tools across the

Adobe platform. You have the ability to

do everything from creating images to

creating videos using some of the very

popular tools on the market. For

example, you can use Topaz, 11 Labs,

Flux Google Ideogram Luma Moon

Valley, OpenAI, Pika, and Runway

entirely inside of Adobe's platform.

Now, I'll do more extensive training in

the very near future on how you can use

Adobe tools for end to-end creativity,

but I do want to show you some of the

newest innovations that we have inside

of the Firefly platform. So, let's hop

in here. You can find a link below this

video. I'm going to be using the mood

board maker inside of Adobe. So, we'll

go ahead and select start mood boarding.

And if you've used other tools like

Miro, it works in a very similar way.

So, you can of course bring in

thirdparty assets if you want. You also

have the ability to create a completely

new board. I'll just go ahead and select

create a board. And of course, we have

the ability to bring in our assets. I'll

go ahead and click X. Now, the cool

thing about using boards is you can kind

of make it anything that you want

depending on your own creative

organizational style. You can lay out

all your images in a straight line. You

can have it be super messy. Everyone's

different and it really just depends on

what type of workflow resonates with

you. So, for example, all you have to do

is go down here to generate an image.

And from here, you can see we have the

ability to select this menu. And now we

have a ton of different image generators

that we can select. And most notably,

there is Firefly Image 5 at the very

top, which is Adobe's brand new image

generator that generates images up to 4

megapixels big. And it's a big quality

improvement over previous versions of

Firefly. That's really important because

again, Adobe's models are copyright

cleared, meaning that your clients can

use the images that are generated from

Firefly and they don't have to worry

about thirdparty IP showing up inside of

the images. But of course, you have the

ability to select other image generators

if you want. Notably, you're not going

to find image generators like MidJourney

or Seeddream inside of this tool. But

I'm going to go ahead and select the

Firefly image 5 preview here. And I have

a quick little prompt here. A cinematic

still of a man in a sci-fi film holding

a magical blue stone. Now, what's also

cool is you have the ability to upload a

reference image. You can, of course,

change the aspect ratio. We'll select

widescreen. And they do have this button

right here that does allow you to go in

and select subprompts. Basically, these

are prompt presets that you can select

to kind of push your generations in a

specific direction. And you can also

select random if you wanted to do

something like that. But I'm not going

to select any of those settings. We're

just going to keep this open. And you'll

also see here that it says it's going to

use zero credits. Adobe has announced

that you get free image generations

using any of the image models, not just

the Adobe models, until December 1st.

And they're also giving you free video

generations using Adobe Firefly until

December 1st as well, which is pretty

cool. And as always, I'm going to go

ahead and click generate a few more

times. I like having more images because

I just get to iterate faster and pick

and choose the best generation for my

project. So, we have a few different

images here. We have this first one,

which looks okay. It's not exactly a

blue orb. Number two, it's okay. Number

three, I think is pretty cool. And then

we have number four. And I did not

prompt for him to be wearing a metal

bra, so I'm not entirely sure what's

happening there. But whenever you find

an image that you want, you can go ahead

and select place on canvas and it will

place it on the canvas here. Now I have

the image here that I've selected. And

what's also very cool is you have the

ability to go in and edit and change

these images using settings that you may

be familiar with in other image editing

software like MidJourney or Flux. So I

just want to note that because you do

have the ability to vary your image. you

can go in and edit it. And you also do

have the ability when you edit to edit

using tools like Nano Banana or Flux

Context, which allows you to be a bit

more conversational in the way in which

you change the image. And to put that in

comparison, here's the image that we

generated from Firefly image 5. And here

is the same generation from MidJourney.

You can see Midjourney pushes into a

much more cinematic result. The result

from Firefly does seem a bit more stock.

The lighting does come across as, you

know, stock video, if you will. But I do

think that it's a big improvement over

Adobe's previous image generator. But

for our example, I want to go down here

and select generate video. And you can

see we have a tab that looks very

similar to other video tools on the

market. And again, you have the ability

to select various video models depending

on the specific model that you want. So

you have tools like Mary by Moon Valley,

Pika, Luma, Ray 3, Runway, Google Vo

3.1. A lot of the really high performing

video models are directly here inside of

Adobe. Now, there are a couple notable

exemptions. You're not going to find a

lot of the Chinese models like Miniax or

Cling inside of Adobe, but we have

Google 3.1, which I believe is the most

intelligent video generator on this

list. So, I'm going to go ahead and

select that one. So, to upload our

image, all you have to do is select the

first frame button, or you can go to

this little dropper, and we'll select

our image from the canvas and select use

as first frame. So, we have first frame

here, and then you can, of course, do

last frame as well if you're trying to

be very specific about how your video

progresses. Because this is Google V3.1,

we're going to make sure we have audio

selected. And it's going to note that

this is going to use 400 credits and go

ahead and hit generate. And after about

a minute, it created this video here,

which is really dynamic. It looks really

cool. I maybe I'd prompt out the like

weird blue flames that pop up here, but

I like the idea of this guy kind of

holding this orb and presenting it to

something. And so I was like, okay, how

can we like do a twoot sequence here? So

guy presents orb maybe to like an alien

ruler or something like that. And so we

actually were able to use Firefly boards

to sequence this out. And so we

generated a new shot here using Google

Nano Banana. And then we have this shot

of him approaching with the orb.

and the alien grabs it. And what's

pretty cool about this video is you can

feel a lot of the weight in the alien

grabbing the orb from the guy's hand.

So, I think it did a good job here. And

I should note that you could totally

expand this. You know, it could become a

huge board with all of your shots. And

what I love is you have the ability to

loop the video so you can see all of the

clips next to each other. So, it's not

technically a video editing timeline,

but it allows you to kind of see how the

sequence of events will unfold, which I

think is just really helpful whenever

you are working on a creative project.

Adobe also announced a few other video

updates that you should know about.

Notably, inside Adobe Premiere Pro,

there's a brand new tool that allows you

to quickly create mask around your

characters, which basically eliminate

the need for rotoscoping for simple

projects or if you work on social videos

where you, you know, cut out your

subject, it's going to be a lot easier

to create those directly inside of

Premiere Pro. They also came out with an

update inside of Adobe Lightroom that

allows you to type in a prompt and find

the images that you're looking for. It

also has the ability to pick and choose

the best images from your shoot and, you

know, can just help you save time

whenever you're pulling selects or

working with a client. They also

integrated Topaz into their tools which

allow you to upres the images and videos

directly inside of Adobe's tools. Adobe

also announced a brand new feature that

allows you to upload a video and it will

automatically generate a soundtrack for

that video that is the specific duration

of your clip. So, let's hop in here and

take a look at it. So, all you have to

do is click the link below this video

and we'll select generate music. So, for

our example, I'm going to bring in an

animation example that I created for a

tutorial earlier this week. It's

basically about a minute long and it

just is, you know, some animals ordering

food in a cafe. And we'll go ahead and

drag and drop that into the video

section here. And you can see

automatically it said we're going to

create a melancholy atmospheric song

with ambient drone style for a dystopian

reflection. It's understanding that this

entire film takes place on a rainy day,

which is uh pretty cool. And of course,

you do have the ability to go in and

edit the style. You can also select

different carrots here depending on, you

know, the project that you're working

on. I'm going to change the energy to

medium, the tempo set to medium, and

then it's 68 seconds long, which is how

long the video is. And let's go ahead

and hit generate. Okay, let's listen to

our first track here.

[Music]

Sounds like a uh a haunted house. Let's

uh do number two.

Okay, a little spooky. And let's listen

to number three.

[Music]

All right. You can see uh it's taking a

very dramatic tone in the generation.

So, as you can see, it will

automatically create music for your

project. Is that music better than

handcurating music from a tool like a

stock music library? I don't think so at

this point, but it is cool that we're

beginning to see automatic soundtracks

inside our editing tools. They also

announced a new tool called Project

Moonlight that basically allows you to

connect your social media account and it

will create content recommendations,

which is pretty cool. very excited to

test out that brand new feature. There's

also a really interesting announcement

called Project Graph where Adobe has

essentially created Comfy UI directly

inside of their platform. So all of the

node trees and workflows that you are

probably familiar with if you use

company UI or more recently runway

workflows, you now have the ability to

do a very similar thing using Adobe's

project graph. And if you work on

projects where you have consistent

outputs that you need again and again,

like social media projects or film shots

that have a very specific aesthetic or

character or style, I think that having

some sort of workflow like this could be

really powerful. They also announced a

brand new mobile version of the Premiere

Pro app, which looks a lot like Cap Cut.

There's a lot of features in there for

allowing you to edit your videos even

faster. And there's a lot of the AI

tools that you've come to know and love,

like the sound noise remover feature

that you'll find using Adobe Podcast. I

use it all the time, and it's really

helpful if you have a lot of media that

you shot on your phone because a lot of

times, of course, you don't have a good

microphone. So, having the ability to

enhance your audio directly inside of a

mobile application is pretty cool. On

the business side, they also announced a

brand new initiative called Foundry,

which allows large companies and brands

to essentially create custom models that

can run inside of Adobe's platform. The

reason why this is important is because

it connects studios and brands with

Adobe's machine learning researchers to

essentially create the best models for

their workflows. A lot of times brands

create a lot of images with specific

products or in a specific style. So

rather than using a complex Comfy UI

workflow or traditional Laura training,

you can work with Adobe's team to create

a custom model specifically for your

brand. And finally, on the business

side, they also announced Gen Studio

that allows brands to train a model not

only on their image data but also data

across their entire company.

Essentially, the value proposition is

you can create different types of

designs and assets from a simple prompt

and it'll do everything from social post

to website banners and streamline the

entire process. They also announced a

brand new AI assistant inside Adobe

Express. This is really cool because

basically if you have an idea for a

design project that you're wanting to

create, you can type in a prompt and it

will give you recommendations on the

type of design that you could use for

that project. And then you can kind of

go down the creative rabbit hole and

give feedback. You can change the

colors, you can change the assets and

this is entirely from a conversational

experience. So, you don't have to click,

you don't have to learn tools, you're

just having a conversation and changing

the design based on your own creative

taste. And what's also funny is the

announcement that got a huge applause at

Adobe Max is a new AI tool that will go

in and automatically name your layers

inside of Photoshop. I can't wait till

they come out with this inside of other

tools like Premiere Pro and After

Effects. And you know, it'll just help

organize your projects and make you seem

a bit more organized than you are

whenever you share your projects with

other people. I think one of the most

notable things that I saw from Adobe Max

is a survey they put together with

16,000 content creators and they said

that basically 76% of those people have

used AI tools to expand their business

or brand. I really had no idea that AI

was so popular across the creative

world. Obviously, in this new creative

era, there has been friction, but it

seems like the vast majority of creative

professionals are using AI in their

day-to-day workflow. And from that

survey, they also found that 38% of the

creatives said that cost of AI models is

the biggest barrier keeping them from

using the tools. And that kind of

transitions us to our conclusion about

all of these updates from Adobe. I think

the fact that such a large company is

positioning themselves as an aggregator

and the fact that they already have so

many creative tools out there that

professionals use day in and day out

like Premiere Pro or Adobe Photoshop. It

really says that the future of

creativity is going to come from these

aggregators. We've talked a lot about it

in our student office hours over on

Curious Refuge along with this channel.

The fact is aggregators are the future

of creativity. The era where you would

have one model on one platform, I think,

is slowly starting to come to a close.

And having a really robust all-in-one

solution that can bring in different

models into a single place, I think is

the path forward. And the fact that

Adobe has such a robust network of

creative tools all in one place already

makes them a leader in the space. But

the big caveat to that is exactly what

the people from the survey communicated

cost. So what does it cost to run these

tools inside of Adobe's platform? Well,

if you want to generate a video using a

tool like Luma Ray 3 directly inside of

Firefly, it's going to cost you 500

credits. Now, to put that in

perspective, whenever you pay for their

$10 a month plan, you get 2,000 credits.

So, essentially, it cost you $2.50 to

create one video clip using Luma Ray 3.

And so if you're working on a short film

project and let's just say it takes you

five generations to get the exact shot

that you're looking for, which would be

pretty impressive because usually it

takes more, and then your short film has

200 shots, it's going to take about

a,000 clips to get that short film put

together. That's going to cost over

$2,000. So there's definitely going to

be a need to reduce the price. The best

dollar fordoll deal that is currently

available on the market is the unlimited

Google VO3.1 fast plan, which is $200,

but you do get unlimited generations,

which I use all the time. Enrollment is

now open for the November session of our

courses here at Curious Refuge. If you

wanted to learn alongside the world's

biggest studios and the most creative

people on the planet, we would love to

have you inside of our program. We had

multiple people from Adobe Max come up

to us and tell us that Curious Refuge

has changed their life and allowed them

to land world-class jobs. We would love

to have you inside of the programming.

And we also launched our brand new AI

screenwriting course which is available

over on the website. Be sure to check

out the courses and let us know if you

have any questions. The team at Miniax

released version 2.3 which says it has

improved physics and is just all around

a better video model. It generates the

video clips in 1080p. Let me show you

how to use it. So, I'm here on the Hilu

AI website and you can see at the very

top here, it's very easy to use. There's

a prompt box. So, we can type in a quick

little prompt. We'll say handheld shot.

a troll walks around a field looking for

something. And of course, let's go ahead

and upload an image to create the start

frame for the shot. So, I have this shot

here. I generated this using MidJourney.

And we'll go ahead and bring that into

the platform. And when you're ready, go

ahead and click create video. And after

a few minutes, it generated this video

clip here, which honestly is one of the

more impressive AI video clips I've seen

in a while. There's a ton of weight in

the character steps. You can see the

details of the hairs all around. It

looks really, really good. Now, there

are a few things that I do want to note.

For example, if you really zoom into the

character and look at the texturing in

their skin, there is a bit of distortion

going on. So, it's not crystal clear. I

do think you'd probably want to use a

third party tool to upres or kind of fix

some of those compression artifacts that

are coming through. But, I think it did

an amazing job. Now, of course, let's

compare that against Clling and VO to

kind of see what the difference is. So,

here's the result from Clling. So,

again, Clling did a really good job.

It's very comparable. I'd say the camera

lens qualities are a little more

realistic from Cling, but altogether

it's not too far off. Both are really

good. And then let's take a look at the

generation from VO 3.1. Okay, so as you

can see inside of VO, the physics and

the performance from the character are

the best out of the other tools. The

problem is Google VO 3.1 at this point

is natively creating the video in 720p.

So you definitely would have to upres it

in that case and of course you're going

to lose out on quality. So it's

definitely a balance between using

miniax, using cling or using Google VO.

I think that once VO allows you to

create videos natively in 1080p and

beyond, they will ultimately win the

video war because the physics and the

performance is simply better and of

course the audio quality is much better

inside of VO as well. So I wanted to do

a couple other tests here. So let's

start out with a VFX shot here. So, I

have this castle and this wall is coming

down here and uh this is the generation

from Miniax, which I should note does

not have sound because Miniax does not

allow you to create videos with sound

effects. Yeah, it looks pretty good. I

wouldn't say it's photo realistic, but

it's pretty darn close. I think it's

almost on par with professional VFX that

you would expect to see in a like a

mid-tier budget inside of a Hollywood

production. So, not too bad considering

that was the very first generation that

we got from Miniax. Here's the same

result from clean. Again, it's pretty

good. I think it's a little too hyper

contrasted, not quite as realistic. I

like inside of the Miniax generation how

the smoke kind of shared in the uh color

grade from the video. So, I think the

compositing inside of Cling was not the

best for this generation. And then let's

see the same result from Google Vio.

Okay, Google VO went really over the top

uh with the dirt explosion there. I

really like the physics a lot, but I

definitely would want to prompt that a

few more times, I think, to get the

generation that we're looking for. And

then for our final example, I want to

show you why ultimately I think VO is

the best tool for day-to-day AI film

making work. Basically, I have this

soldier here and I want him to be kind

of, you know, upset and say, "What is

your name?" And uh this is the result

that we got from Miniax. He's not saying

what is your name. You couldn't even lip

dub that because it would not match. So

that's not what we're looking for. Uh

here's the same result from cling.

>> Okay. Performance looks real, but

obviously he's saying nonsense words, so

that's not what we want. And then

finally, here's the generation from Vio.

What is your name? Is that

>> okay? Uh, he had a bit of a stutter

there, but you can see I think the

character performance is the best from

VO. And the fact that he actually says

the words that we're looking for is

pretty great. The problem that you may

run into in VO is it will automatically

put sound effects and music behind your

character. Sometimes, no matter how good

your prompting is, you can't remove

that. And so you may want to use a tool

that separates background noise and

music from the voice like the voice

separator inside 11 Labs if you're

working on a film project. So with all

of this being said, is Miniax 2.3 the

very best AI video tool on the market? I

don't think that's the case, but it

could be very helpful in select

workflows. I do think I would primarily

use Google VO, especially Google VO Fast

for a lot of my film projects, but if

I'm not getting a specific shot the way

I want, I would of course hop over into

other tools like Clling or Miniax just

to see what we get and compare the

results. The team at Magnific released

Precision 2, which is their AI image

upreser. It's a big improvement over the

previous version of Magnific, but I'm

curious about how it stacks up against

other AI image upresers. So, let's hop

into the tool to see how it works and

then compare it against other tools. So,

to use the precision upscaler, all you

have to do is go to magnific upscaler.

Make sure you have precision selected

and that you are working in version two.

There's a few different versions of

version two and you can totally pick and

choose the right one for your specific

project. I'm going to go ahead and

select Sublime just to get the maximum

quality possible. Now, for our input

image, I have this image that I

generated inside of MidJourney. I'll go

ahead and drag and drop that into the

input image section. And you can do a

scale factor that's way bigger than your

base image. I'm just going to stick with

times 4 because that image would end up

being very large already. I'm going to

turn off sharpening because I don't like

to increase the sharpness of my images.

But for smart grain, we'll keep it at

we'll say about 4%. And go ahead and

click upscale. So after about 30

seconds, we have an image here. So

here's the before. And then here is the

after. You can see we have a lot more

detail on our character. So before was

very soft. It's very pixelated. And then

after, there's a ton of detail in her

skin. You can see just little bumps,

little cracks in the lips. There's a bit

more of a natural sharpening around the

glasses. The weave of her hat looks much

more realistic. And so it did a good job

at making this overall scene look

really, really good. Her hair looks

very, very good. The way the bokeh kind

of slowly fades the hair into the

background looks realistic. Her

shoulder, you can actually see the film

grain and the texture from the shirt.

So, it really did an amazing job. Now,

of course, we have to compare that

against other AI image upscalers on the

market. Last week, we talked about a new

tool called Crystal Upscaler, and this

was the result from Crystal, which again

looks pretty good. There's a lot of

detail in the skin. You can see there's,

you know, the wrinkles, uh, in the chin.

There's a lot of realism in the hat

here. Her glasses have like dust on the

lenses, which looks pretty cool. And

again, it did a really, really good job.

We used Topaz Gigapixel to upres the

image. And you can see her lips look

very plasticky. The skin looks plasticky

as well. So, Gigapixel did not do an

amazing job here. And then finally, we

upresed the image directly inside of

MidJourney. And you can see the skin

texturing looks super super fake. So,

with all of that being said, I think

that the new precision upscaler inside

of Magnific is the best image upscaler

on the market. It seems like it's doing

a photo realalistic job. And if you need

an image to be really really big,

especially if you're working on a

project that's going to be printed out

or put on a large canvas like a

billboard, then I think Magnific is the

best tool for you. From a cost

perspective, it cost about 17 cents a

generation to upres your images. If

you're using that X4 model, that's

compared to 40 cents if you're using the

Crystal Upscaler that we talked about

last week. So, I do think Magnific is

just generally your best bet. And you

can compare that against Gigapixel,

which is $17 a month. So, the tool's

cheaper. You can generate up to 10

images at a time. Altogether, I say

Magnificol

for image uprising. There are a ton of

AI film events popping up around the

world. We have a Curious Refuge meetup

in Denver on November the 1st. We host a

digital office hours every single Monday

where you can get your AI film, video,

and creative questions answered by our

team of experts. That's on Mondays at

9:00 a.m. Pacific time. And we also have

a Curious Refuge meetup on November 5th

in Palm Beach. Be sure to check out our

film events page over on our website to

see not only Curious Refuge events, but

also the greater AI filmmaking

community. And that brings us to our AI

films of the week. We have a ton of

films that we want to shout out in this

week's episode, but I want to give a

huge hat tip to the winners of our

Halloween competition that we put

together in partnership with Epidemic

Sound and Leonardo. Our judges had a

really hard time picking the winner of

the competition, but at the end of the

day, story was king and we were really

impressed with all of the submissions.

So for our films of the week, I want to

shout out three projects that were

created by our studio Promise in

partnership with Adobe for Adobe Max.

The first film is a project called My

Friend Zeff by Dave Clark. This was a

hybrid project that took traditional

film making and AI film making and fused

them together. Some of the shots inside

of this project look incredibly high

budget and it's one of the best quality

examples of a hybrid AI workflow that

I've seen to date. The next film that I

want to shout out is called Kaira by

Meta Puppet. Meta did an incredible job

with bringing a building back from

vintage photographs that he took and he

crafted this really awesome compelling

human story that's also really funny and

uh really showcases his directing and

editorial skills. So fantastic job on

that project. And then finally we have

Nagori by Guom here at Promise. It's a

really beautiful 2D animation aesthetic.

The story is very compelling. that the

voice over, even though it was AI

generated, is really, really nice. So, I

highly recommend checking it out. We're

also going to do an exclusive

behind-the-scenes breakdown with these

filmmakers to share with you a specific

workflow that you can follow to create

similar films of your own in the very

near future. Thank you so much for

watching this week's episode of AI Film

News. As always, please like and

subscribe here on the platform to get

the latest AI news and tutorials

directly on YouTube. And you can

subscribe over on our website to get

AILM news sent to your email inbox every

single week. I hope you have an amazing

week. Best of luck on your projects.