Hi guys. Hi everyone. Welcome back. I'm

Edo. Thank you to be here in my channel

when where we talk about um advanced um

AI generation models. Um today I'm just

recording in English just because I want

to practice uh a little bit. Um and

today we are talking about uh Nvidia RTX

upscaler. A new powerful super

resolution upscaler that I tested in

this during these days many times. I

made many many tests and I want to share

with you my thought about this model and

it's um its capabilities

and I want to test this uh this node

with another powerful open-source um

upscaler um seed war 2 uh that I'm using

um nowadays for upscaling video and

images image. So today it's um a test

day and I want to show you how to

install this node and to face with

problems you can uh you can handle um

and I want to show you the reference and

the problem and um the the uh the notes

about this tool. So first of all you

need your confi of course and you need a

um

update version of confi. Remember uh if

you are using a portable confiu like me

you have to go in this folder update in

your confui folder and in update you

have to doubleclick update confui.bat

that. So you had uh you have um updated

confi and you can work with um and the

first thing you have to do it's go to

extensions and to all extension if you

if you prefer and type RTX

and you will find this node here

confi_vidia_tx

uh node okay Um

this this is

important video

super video

super resolution maybe

no there is there isn't another another

one so you can um

um you you have to choose just this one

you you don't have to work around just

this this node you have to um switch on.

So it is it install everything. You have

to made um a restart and maybe you will

find an error. Maybe I faced this error

many times. So I had to work a lot to

find a resolution and the the resolution

is here.

You have to um to go in this

um in this folder, the confi folder and

you have to go in python embed. Okay,

right here and you have to type cmd. So

you enter in your control and then you

have to uh copy and paste this your this

uh prompt here. This will uh install

everything you need this Nvidia uh VFX

uh through the uh right um uh site.

Okay, right folder. So once you done

this you can use it and you have to of

course uh restart everything and you can

see the complex you know um workflow

of course I'm joking the I I delete this

it's unusable my my test you have to

link a load image node uh a girl I made

with Z image Okay, I used my, you know,

um my workflow. This one

you can find in uh in my

um buy me a coffee repository. It's a

it's a useful tool um that create with

an agent with the quen

that refine your prompt simple prompt

and create something useful for you know

advertising fashion. So uh skin tones

colors lights everything for uh sorry

for image like this. Okay very powerful

in my opinion but it's not the the topic

of the video. So once you have have you

once you have your image you are just to

upload it and to use this RTX double

click RTX super resolution note this one

and we have just just two or you know

three options

that it's a cool feature or not. It it

depends. Now we'll see the the method.

Uh the resize type you have to choose

scale by multiplier. So I'm want to

scale by two.

So we want to to to go up to 5,000

pixels. So maybe a a 5K resolution.

Okay. and the quality that I I I don't

know the the the meaning of of this this

four um level I'm using ultra I'm using

um a 5060 Ti with 16 gig of V RAM so I'm

I'm quite uh sure it works and it work

uh as you can see and and that's it

and that's it

and when you. I want to to show you

this. And when you press run,

okay,

this is the the the speed.

Now I'm recording. Okay.

I just want to show you the the

different the difference. Okay.

Um

maybe we can try with another image if

we want

just to this one. Okay, it's the same.

Okay, by two by four maybe. Okay.

Okay, this is the speed of the

the the process

and I'm facing

5,000 pixels. Okay, 3 seconds.

3 seconds.

uh with recording. Okay, without

recording I'm I'm less than 1 second.

Okay, I had previously 0

49 seconds for an upscaler.

It's

in my opinion amazing. Okay. In terms in

in pure terms of speed,

the same image, the same image with

um seed war 2.

And you have to work a little bit with

this uh you know um with workflow. You

need to choose the the right model for

the details. You need to write choosing

the blocks uh the resolution. You have

to do something with your your hands

and you can reach this. I can reach this

uh resolution

uh with my with my GPU without uh enter

in uh out of memory mode. So this is the

maximum for me. So is less than this one

but however very powerful 4K. Okay. And

this this workflow I don't want to to

show you the process because it's very

heavy with the the the the recording. I

I um I um um upscale this this image in

just a second. 49 seconds.

49 seconds. So

let's say without without recording

1 seconds

we without recording. Sorry this is the

video later without recording 49

seconds.

The difference is incredible.

I I I would say it's magic. Okay. But

now we have to be precise and

professional.

These two upscaled image are the same.

No. And I want to show you why.

This

is the

um right image I start with. So you can

see 2,000 pixels. Okay. to 2,7700

and and etc pixel 2k 3k you know 2k

image

this is the first image I create with

the image okay 2k cool I think it's fine

I think it's lovely

this is the ratio

as you can see

of the 4K resolution with RTX

And we we see something here. We see

something here. We see something here.

We see something here. Not so good in my

opinion.

Cause if we see

war, this is the right ratio.

It's a little bit lower than RTX because

RTX is 5,000 pixel. This is 4,000

pixels. But as you can see, the

difference in terms of, you know,

details

that Sid Voir 2 add to your image. It is

in my opinion very impressive.

It's impressive.

We barely see this uh this this nose

ring here.

We see here we see all this texture in

the lips on the lips.

We see a mess

here.

So

we have to be very, you know, honest in

this um in this test. I think it's

incredible in terms of speed, pure

speed. But in terms of resolution,

I don't know if I would use it for a

single image, uh, a single, you know,

hero image. Okay,

it's it's powerful.

Okay. So, we can

we can bring all to you know 4K to 2K.

Okay. From 4K to 2K right here.

And okay, we have uh you know a new

a new pixels. But

uh it's another way to look at these

images.

You have to look at the right

resolution. And this is not the right

resolution. So um

we need to focus to the you know the uh

the the right way to to look at this

this uh models. I think it's quite quite

good in terms of speed but for images I

think it's not worth the the

uh the the the

you know the matter but but

this is about image

very different thing if we are looking

at the videos

because

right here I have this, you know, uh,

workflow

for RTX upscale and you can find right

here example workflow right right here.

This is the workflow for the video for

RTX

RTX. And as you can see, we have a 1K

video,

very simple, 5 seconds

with a different focus on camera. And

this is the double

resolution.

Okay, I want to show you

a full full video, full screen.

This is the 2K video.

I think it's quite quite impressive. As

you can see, this is the the resolution

1,400 2,600

pixels for

starting from a video

700 by, you know, 1,300.

So we doubled the the resolution with a

very cool

um very cool uh res result.

And please listen carefully.

We have upscaled this video this five

seconds video in

I write it down

19 seconds.

19

Okay.

5 seconds in 19 5 seconds of video

upscale in 19 seconds locally.

Remember the number 19. Okay.

Uh this is the workflow for the video

upscale with you know um seed war.

And [sighs] as you can see right there

we have minor issues with the texture.

The model it's very sensitive and it

make crisps every single frame and when

the frame when when the frame are in

loop or in sequence the result it's

you know not so so polish like this one.

There are no artifacts

are no crisp business like like the base

video like the base video. This is 1K.

This is 2K.

It's the same, right? The same. This is

2K with C2 Vir. And we have a

flickering. We have this texture. We

have something very bad. But

the important thing is with seed voir 2

I

render this

video upscale in 29

minutes.

This is uh you know uh a game changer.

19 seconds. 19 19 seconds.

I will I would I want to say it in

Italian.

Half an hour here.

So once again,

I would use RTX RTX video upscaler for

everything. No, but for video, yes.

Yes, of course. Definitely.

It's it's, you know, I'm astonished

by this result.

19 seconds.

half an hour

there is a leap a quantum leap in terms

of of quality in terms of render time in

terms of everything in terms of

proficient way to use AI.

So my advice if you want to use a great

image upscaler open source I have no

doubt

this one seed 2 it's perfect

it's very very very perfect for your

images if you want to upscale your video

you have to choose RTX video from Nvidia

there is no there is no match

Not not in in a second. Okay. 19 seconds

for a 5-second video. Half an hour for

for the same same video.

And for today, it's all guys. Thank you.

Thank you to be to be here with me. Uh,

thank you for your comments, for your

feedback, for everything you need to to

to know. I try to uh to answer to your

question every time I could. And I hope

to see you next video. I want to show

you as much as I can of this kind of uh

incredible

videos. And I want to thank again uh

Stefano Flor uh that um

that sent me this this incredible tool

uh to try uh to experiment with. Graci

Stefano Solito um he's my pusher in

terms of uh uh you know uh candy for AI

and this tool is a very it's a

incredible candy. Um, once again, thank

you and I hope to see you next video.

For today, it's all so