Dalle-2 Is Now Open to Everyone! - Getting Started Tutorial

Hey, everybody.

Scott Detweiler here.

So, the big news today is that DALL-E 2 is

now open and available

for everybody to try it.

So, if you've been on that waiting list

and you haven't been called,

now's your chance to play with it anyway.

So, unlike Stable Diffusion or Midjourney,

some of the other AI art tools out there,

it has been around a while.

It existed earlier as Craiyon or

DALL-E Mini, which isn't as powerful.

DALL-E is a little bit

more stock photography looking

if you're looking for a description

of what it produces, say Midjourney,

which is more painterly,

Stable Diffusion, which seems

to do pretty much everything.

For example, if you need a picture

of a cat with a hand grenade,

you can get one.

It's exactly that type of tool,

but it has some other really cool

capabilities, and that's the ability to go

in and modify a part of an image and give

it a very basic prompt, and it will

be able to fix that area of the image.

So, I use it more for troubleshooting

in my photography than I

do for AI art generation.

So, let's take a quick look at it so

you can kind of see how to use it.

I'm gonna give you some basics,

and then you can go and play.

Now the DALL-E 2 interface

is pretty simple.

It's just web-based,

and you've got just your prompt.

No other settings really,

so you're basically just

gonna type in what you'd like.

One of the refreshing things about

this system is that it

is very good at English.

So, it understands sentence structure,

unlike Midjourney or Stable Diffusion

where you've kind of got to comma separate

things or space separate things,

and it really doesn't matter

how you put it in there.

This one actually

is very good at sentence structure.

Another thing is that this seems to be

trained more with stock photography,

so you're going to get more

of a photographic look from this.

At least that's how

I've been looking at it.

When I'm looking to make art,

I'm reaching for Midjourney

or Stable Diffusion.

This is more of a problem-solving tool

for me, and a very good one at that,

but it is not one that I use

for generating art from scratch.

So, for example, you can ask

it pretty much for anything.

A bag of potato chips that is on fire.

Notice that every generation inside

of here will cost you a credit.

Up at the top, you have

access to your membership.

You can see how many credits you have.

So, this did not cost us four credits, but

it cost us one for each generation shot.

But the most exciting part

of this tool is not this,

it's actually the outpainting.

So, if you click up here and you go

to Outpainting, it's gonna open an editor,

this is the part that is amazing to me.

We've got two options.

You can either work with an image

and generate pieces of it,

or you can start and put a piece

in the middle and generate outwards.

I'll show you each one.

So, the generation frame in here is 1,024

pixels by 1,024 pixels,

and we can load an image on top of it.

So, I'm gonna load one in here,

and notice that this image

is larger than 1,000 pixels.

So, if I want to, I can have

the generation frame come in.

We can work our way around the image, or

I can scale the image down

so that it's under the 1,024.

In this case, we'll make it

a little bit farther under.

When I click OK, I'll be handed this back.

Now hold down your space bar,

you can actually move around the Canvas.

The interface is a little wonky.

Like I really don't like this white bar

that comes across, because if we zoom in,

for example, at some point the white

bar cuts through the image.

You can also use your control key

and your mouse wheel if you'd

like to scroll this way.

And then you can move the generation frame

by left-clicking inside of it anywhere.

So, we can position this

image where we would like.

Now again, because it understands English

pretty well, we don't have

to overemphasize the prompt here.

Let's type in autumn

forest and hit generate.

This will cost us one generation,

and we'll be given four options.

As you can see, it does a very good job

of matching the style and color

tones of the image for the most part.

Sometimes there's a few that are special.

These are pretty good.

Once you find one you like,

just download it.

Now if we refresh this by clicking

on the name up here,

we can go back up to show editor.

Now we could try it the other way.

For example, let's work

with this image here.

So, you can see it's much larger

than 1,024 pixels, but that's okay,

we can accept it, and now we have our

generation frame, and we can float it.

So, we can move it off to one side,

for example, if we wish

to continue this image.

You can see it, did a pretty nice job of trying to match the context what was given

with an extremely small prompt, which I think is really impressive

Lastly, what you can do is you can choose to use the Eraser

When you use the Eraser, you can take part of the image out, example here,

we remove this window.

Uber generation frame there, click generate again.

But the coolest things you can do with the outpinning tool is put an image in, have it generate the outside and then load the image in again

and again creating what looks like an infinite zoom out, are you zooming out of the image because each time you're asking it to generate the perimeter.

We can start with this image for example. I just need to make sure that I'm less than 1,024 pixels on the long edge. I can position the generation frame where

I'd like it and then continue working.

Again, oftentimes, an image will be

a little bit off, but that's okay.

We get four options,

and we only pay for one.

Now what I would do is I would download

this, and I would re-upload

it and start from here.

So, that's DALL-E 2,

very powerful.

As I said, I use it more

for troubleshooting than I did

for ideation or art generation.

But it's pretty much the most amazing

stock photography generator that it's ever

going to be and a great way to

edit parts of an image that seem like

they're almost impossible to find

something that's gonna fit.

It fits perfectly.

So, I'm very excited.

I'm glad that it's public now.

I think it is, because Stable Diffusion

and Midjourney have both been very

popular, and it's been very difficult

to get access to the system.

So, I'm glad to see that they

have finally opened that up.

So, give it a whirl and let me know

what you think in the comments below.

Everybody take care.

Stay safe, and I'll catch you all next time.