#community-help

1 messages Β· Page 16 of 1

tepid path
#

Hello, I was trying to make the clip interrogator work but after successfully running "Check GPU" and "Setup", no field to paste the url shows up under "Interrogate". Anyone else experiencing a similar issue or having experienced it in the past?

unknown.png
golden ruin
feral kite
#

Hi guys!
What's the next event if there is one?
Is there any news coming from dall-e for the next weeks/months?

bitter delta
regal lynx
#

You'll have to wait and see! πŸ™‚

feral kite
#

It means yes?

#

About news i also mean like upscaling or dallify for discords..

regal lynx
# feral kite About news i also mean like upscaling or dallify for discords..

We don't know what OpenAI is up to as a company here, so we don't know if/when features will be added or what those features will be. In regards to Dallify, I don't think it will be released as a Discord bot that you can add to your own servers because it's not a product by OAI. Rather, Dallify was built by volunteers. Will it be available on this server again? I do not know the answer to that question

feral kite
digital yarrow
#

I can't seem to get the editor to work. First time using it and I used '?' and that doesnt bring up commands nor do any of the commands work to move or scale an image. Let me know if anyone knows what im doing wrong

regal lynx
#

are you referring to the editor on the labs website?

digital yarrow
#

yup yup!

#

I uploaded an image and just try to combining two images and it doesnt do anything and my 2 uploaded images just stack over each other

regal lynx
digital yarrow
regal lynx
#

when uploading images, you should be able to drag the image over before clicking the checkmark βœ… which locks it in place πŸ€”

rugged wing
#

Does anyone know if Dall'e can be used to generate correct logos without text?

digital yarrow
regal lynx
regal lynx
rugged wing
#

Thank you, I will try

digital yarrow
#

google chrome

regal lynx
#

I'm trying it on Chrome on desktop right now, and I'm not having issues. This tells me that the website is working correctly

rugged wing
#

I was thinking of this prompt : "a simple logo for a lighting company representing a hanging lamp" can someone improve it ?

regal lynx
rugged wing
#

Thank you !

#

If you can anwser to it in the other channel it would be great

feral kite
#

What is curated art?

regal lynx
vague yarrow
#

Does this discord ever host art competitions?

regal lynx
#

we haven't hosted one yet, but we may in the future

vague yarrow
#

That would be very cool

vague yarrow
#

Are the website devs available for suggestions here?

regal lynx
vague yarrow
#

Thankyou!

bold temple
#

I was wondering on how Photopea did their implementation of Dall-E. You can render images in 1024x1024, and you don't need an openai account to use it, but I thought you could only access Dall-E through the official openai website.

regal lynx
bold temple
#

Ohh gotcha, thanks.

midnight lynx
#

Where do I find the frame extending tool?

short salmon
midnight lynx
pearl dew
#

hi there! If we create a digital piece of art qwwith Dalle2, are we able to mention in has been done with Dalle2? we may promote the exhibition mentioning thae art has been done with Dalle2

#

or we should say "done with an AI"?

golden ruin
timid thorn
#

Hii. i want to ask something. How can i remove my history? or will it be automatically removed after several days?

pine void
timid thorn
#

well. it only clears recent tab. not the history.

pine void
#

Oh! You mean your entire History?
As far as I know that's not possible skeletonconfused

timid thorn
#

yeah. I see. Thank you!

pine void
rugged wing
#

When does a celebrity become a historical figure (and is okay to use)? I've used Albert Einstein several times without complaints. Would, say, David Bowie be okay? Malcolm X?

pearl dew
drifting condor
#

GM, serendipitous timing with this update: "We’ve simplified our Terms of Use, and you now have full ownership rights to the images you create with DALLΒ·E"

I was just wondering how to specify licenses when DALL-E components are included in a larger work

does anyone know what specific CC license DALL-E works might fall under given that new update (ie. pretty much anything we want to choose, from all rights reserved to CC0), or does it depend on the rest of the final product?

void talon
#

does this change in rights of the image generations mean the consumer now fully owns the image generation, or that the consumer also owns the image generation, and openai also owns it?

answer wont affect my practice/behavior, mainly just a logistics question

drowsy pewter
#

im new to this kind of thing, do i just put the prompt script into python?

image.png
median oxide
#

Is the python package supposed to be updated? Do I have to manually post the url

golden ruin
rustic marlin
#

I just tried the API, looks like it has a separated pricing on the usage?

#

am I not able to use the credits I have on the API?

drowsy pewter
#

i tried python but for some reason it didn't do anything

rustic marlin
#

you probabily have the same issue

#

Billing hard limit has been reached

#

looks like the API uses you money on the openai account, shared with the othe rmodels such as text

#

and it does not use your credits from the DALLE website

full briar
#

I get an error when I try to run the API
TypeError: openai.createImage is not a function

rustic marlin
# 
const { Configuration, OpenAIApi } = require("openai");
const configuration = new Configuration({
  apiKey: your apikey here,
});
const openai = new OpenAIApi(configuration);
full briar
#

alright let me copy and paste that in to check

#

no Im using javascript

full briar
rustic marlin
#

im using TS too

#

i madding credits to my account to test it

#

Ill get back here when I make it work =P

full briar
#

ok

rustic marlin
#

something is wrong on the billing setitngs lol

#

I've gotBilling hard limit has been reached

#

the usage is at 0.05, I've set it to 5.00

#

nvm, now it works

drowsy pewter
#

wait

#

lugui

rustic marlin
#

Β―_(ツ)_/Β―

drowsy pewter
#

python 3.11 for this

#

or something im guessing

rustic marlin
#

well, it works

#

just needed to add the credit card to my account

terse drift
# 
curl https://api.openai.com/v1/images/generations   -H 'Content-Type: application/json'   -H 'Authorization: Bearer $xx-XXXXXXXXXXXXXXXXXX'   -d '{
    "prompt": "a white siamese cat",
    "n": 1,
    "size": "1024x1024"
  }'

dunno why but its giving me an error :/

  "error": {
    "code": "invalid_api_key",
    "message": "Incorrect API key provided: $**-***********************************************. You can find your API key at https://beta.openai.com.",
    "param": null,
    "type": "invalid_request_error"
  }
}
rustic marlin
#

ok, I think I found the problem

#

a small slip on the documentation =P

#

let me jsut check and I'll tell you

#

yep, that is the problem

image.png
#

ok.. so... at least on the Node lib..
the response has a data object, which contains all the data of the response..
the response itself also contains a data entry, which has the rest of the data

#

so.. replace const image_url = response.data[0].url; with const image_url = response.data.data[0].url;

long lily
rustic marlin
#

yes

long lily
#

how hard is it

#

can you guide me please

rustic marlin
#

I have my bot already

#

just added another command

#

here is the code, for anyone interested

import { TextChannel, Message } from "discord.js";
import { Configuration, OpenAIApi } from "openai";

import config from "../config";
import fetch from 'node-fetch';

export async function run(client, message: Message) {

    if (!config.dalle2.whitelist.includes(message.author.id)) {
        return;
    }

    const textChannel = message.channel as TextChannel;
    textChannel.sendTyping();

    const inputStr = message.content.split(" ").slice(1).join(" ");

    if (!inputStr) {
        message.reply("Escreva mais alguma coisa...");
        return;
    }

    const configuration = new Configuration({
        apiKey: config.dalle2.apikey,
    });

    const openai = new OpenAIApi(configuration);
    await message.react("πŸ”„");

    try {
        const response = await openai.createImage({
            prompt: inputStr,
            n: 1,
            size: "256x256",
        });

        // remove the reaction of the message and add a checkmark
        await message.reactions.removeAll();
        await message.react("βœ…");

        const image_url = response.data.data[0].url;

        // download the image to a buffer
        const imageBuffers = [await fetch(image_url).then(res => res.buffer())];

        // send the image as an attachment
        await message.reply({
            content: inputStr,
            files: imageBuffers,
        });

    } catch (error) {
        console.log(error);
    }
}
#

anyway, it would be very nice if the devs could either fix the documentation to reflect the return of the lib, or the lib to reflect the example return

#

just to be clear, this is wrong:

image.png
#

it should be image_url = response.data.data[0].url;

rustic marlin
drowsy pewter
#

could i test it out myself?

#

to generate a sample

rustic marlin
#

test what?

drowsy pewter
#

the discord bot

rustic marlin
#

nah, sorry, private bot on private guild

drowsy pewter
#

i understand

rustic marlin
#

but I can help you with code if you want

drowsy pewter
#

sorry for asking

drowsy pewter
rustic marlin
#

sorry for the ping @short salmon @split hazel @pine void
maybe we could have a coding channel, since we've got the API now

short salmon
#

πŸ˜‰

rustic marlin
#

nice

pine void
#

Enjoy! :)

regal lynx
#

woo!

novel raven
#

so i've never meddled with coding before. what's the difference between using an API and using the openai website?

rugged wing
#

I've got a question about the new ownership rights for images

#

Does this mean we can resell our images and extend those rights in various ways?

#

For example, before, you could monetize an image in various ways, but now with the new ownership rights, we should be able to do whatever we want with the images, such as making a stock-art pack of Dalle2 images to resell?

polar horizon
#

Literally dying trying to setup the API...

meager ermine
#

do my 5 credits a day i got for the october month disappear this month? (I still have many left)

regal lynx
meager ermine
#

ah okay thank you!

tardy pewter
#

Hello. Is this dalle e 1 or 2?

#

Im new to this

rustic marlin
tardy pewter
#

Ah cool okay

#

Hello. I made an image but in the bottom right i see color squares

#

How do i remove that?

golden ruin
#

or crop it

tardy pewter
#

Where is the inspect button?

golden ruin
# tardy pewter Inspect?

no, i mean inspect option on your browser, right click and then click on inspect, you will have source image without watermark there, or you can crop it

Screenshot_2022-11-04_at_00.54.27.png
tardy pewter
winter latch
#

Hi all
Has anyone had any success creating first-person perspective photos?

charred linden
#

It's not good πŸ˜…

winter latch
#

how? I haven't been able to, no matter, what kind of phrasing I use

charred linden
#

IdkπŸ™ƒ

winter latch
#

come on dude, give me an example of the words you used

golden ruin
winter latch
winter latch
golden ruin
# winter latch like this

you can see the prompt here under this photo, what prompt did you use and what did you find impossible? πŸ™‚

winter latch
golden ruin
winter latch
#

"the world through the eyes of a cat", "first-person perspective of a cat", "how a cat sees the world", "the world from a cat's point of view" and many otheers

#

and simply got a photo of a cat's face etc

golden ruin
winter latch
#

so how would you phrase it?

#

also, another question if I may: is there a way to download all these photos at once?

402e60ad90fdb364.png
golden ruin
# winter latch so how would you phrase it?

once again depends what is your focus, if you want to show us a world and then let us know it is from cat's perspective, we have to see the cat, whether it's a reflection in the mirror, water or etc. if you do not wish to have cat in your image, then you can just simply generate anything that includes the world and pretend it's from cat's perspective

winter latch
#

ah yeah I see what you mean

golden ruin
keen hamlet
winter latch
winter latch
#

Is this the right place for questions about the software because I still have a few lol

keen hamlet
sly quest
#

just a question: is not bad make a ai generated image and then trace it to do a drawing?

keen hamlet
# sly quest just a question: is not bad make a ai generated image and then trace it to do a ...

That is a pretty common workflow for artists. How to credit the role of the AI in underpainting isn't really covered in the TOC, unfortunately. My personal opinion from an ethics point of view is that if the art was not solely created by the human artist, the artist has to disclose some collaboration with the AI. Composition ideas and concept art are a major use case for the AI in professional applications.

#

From the Content Policy:

unknown.png
polar horizon
# rustic marlin here is the code, for anyone interested ```TS import { TextChannel, Message } fr... 
const { Configuration, OpenAIApi } = require("openai");
const configuration = new Configuration({
    apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration);

module.exports = 
{
    name: "generate",
    async execute(msg)
    {
        const promt = msg.content.split(" ").slice(1).join(" ")
        const m = await msg.channel.send("Lade Bild..")
        const completion = await openai.createImage({
            model: "text-davinci-002",
            n: 1,
            prompt: promt,
            size: "256x256",
            temperature: 0.6
        });
        if (msg.channel.messages.cache.get(m.id)) m.edit(completion.data.data[0].url)
    }
}

        const completion = await openai.createImage({
                                        ^

TypeError: openai.createImage is not a function
#

any Idea how to fix that?

split hazel
polar horizon
#

oh sorry

#

didnt see

split hazel
#

Np

rustic marlin
polar horizon
#

I did, I fixxed that already. the npm package was broken idk why

noble rune
#

So I saw a video where a person generated new frames of an image using an AI and then someone said it was Dall E 2.
Is this true? And if so, how do I do this?

rustic marlin
#

Dalle does not interpolate video

#

even if you somehow make it into an image sequence

#

there are AIs that do that tho

noble rune
#

hold on

#

I can get the video

#

never mind

#

forgot I can't send videos

hollow locust
#

How can I make an API request from a windows computer? The example request was for linux

regal lynx
hollow locust
#

Thank you

echo lagoon
#

I saw on your blog that the API is available. When will I get the email?

dull basinBOT
#
[❗️] β€” DALLβˆ™E 2 API

DALLΒ·E API Now Available in Public Beta
Developers can now integrate DALLΒ·E directly into their apps and products through our API. More than 3 million people are already using DALLΒ·E to extend their creativity and speed up their workflows, generating over 4 million images a day. Developers can start building with this same technology in a matter of minutes.

regal lynx
echo lagoon
#

And as well, I might need help with setting it up because I know nothing about code

regal lynx
echo lagoon
#

Ok

vivid mortar
cold junco
#

Hi. I just started using Dall-e 2 and I love itπŸ₯³ but I would like to remove the watermark/Singnature on the picture. How can I do do that?

DALLE_2022-11-04_12.00.25.png
echo lagoon
tame comet
#

depends on your web browser, usually you can do it by right clicking -> inspect

cold junco
tame comet
golden ruin
lean bridge
#

How can i create images from text

golden ruin
lean bridge
golden ruin
lean bridge
#

@golden ruin why

lean bridge
golden ruin
# lean bridge Why?

to generate images you have to use either website or do it through api, this is how it is at the moment

lean bridge
golden ruin
lyric igloo
#

Can my OpenAI credits pay for my Dalle api credits? and where can I see clear pricing for the API? Thanks!

verbal crag
lyric igloo
#

Roger thanks major

tardy pewter
#

Hello. I try to download my image on Dalle2 with the remove-watermark method, but it saves it as WEBP? How do I maintain the highest quality and remove watermark? Thank youuu

rustic marlin
#

webp is high quality, just convert to PNG

tardy pewter
rustic marlin
#

well, google for a program called file-converter

#

it adds a context menu entry on the files.. so you can just.. convert it

rustic marlin
tardy pewter
#

Ah thanks!! got it. But 316kb is high quality ye? I expected more

tardy pewter
#

got it im super happy

#

wanna see my first created dall-e? im so happy with it omg

#

lol

#

i will delete it now after u seen it

rustic marlin
#

cool, but I don't think they will let you send it here, but send it on a Gallery channel xD

tardy pewter
#

ye i will delete it after u seen it hahah

#

im crazy proud of it lmao

#

ai art is amazing

rustic marlin
#

it is really good

tardy pewter
#

yayyyyy

#

thanks πŸ˜„

#

where would you recommend to 'showcase' it?

rugged wing
#

Hello, how can I extend pictures on phone? I can't find that option.

tardy pewter
#

now I just upload it on my instagram

rugged wing
#

Hello, how can I extend pictures on phone? I can't find that option, and it I try to do desktop mode I can't cause it's buggy, can someone help?

rustic marlin
rustic marlin
tardy pewter
rugged wing
rustic marlin
#

yea, it wasn't really made for mobile

#

sorry, but i guess you'll need to get on a pc

rugged wing
#

can you do it for me?

#

i give you image to do it?!

rustic marlin
#

sure

tardy pewter
#

hey @ligui what do you think of midjourney and nightcafe etc then? why do you use this over that or do you use all of them

rustic marlin
#

per the rules, Other AI generations where to be at #ai-discussions
they are cool tho, used a few of them, but so far dalle is the best IMO

tardy pewter
rugged wing
#

If someone not on this discord has used my dalle creation without permission and without crediting the dalle program, is there any recourse?

#

It was used commercially.

ocean minnow
#

hi! how would you describe the strengths of dall-e? for example, it is very good at landscapes, graphics and abstract stuff but people and animals are still quite rudimentary

regal lynx
dull basinBOT
#
Rule 11
Do not share content from this server without the creator's consent.

We do not allow crawling, scraping, or otherwise externally posting of art or links posted in this Discord server without explicit consent from the original creator.

boreal gulch
#

Hey all,
the DALL-E API returns "Billing hard limit has been reached" for all my requests. I didn't get a single picture so I didn't hit the limit.

regal lynx
# ocean minnow hi! how would you describe the strengths of dall-e? for example, it is very good...

Personally, I think it's difficult to give Dall-E 2 blanket statements like "it's good at x, but bad at y". Due to how new this technology is, it's hard to see how the model acts as a whole. I've gotten generations of great animals followed with ones of bad landscapes and vice-versa. Does it mean that the model is bad at something? I think it's due to the random nature of how the model creates images

regal lynx
boreal gulch
ocean minnow
echo lagoon
#

By the way, when you use the API, does it take the credit from the account?

regal lynx
regal lynx
echo lagoon
rugged wing
#

I need help! no matter what I do there is always motion blur

nvcdfjt_gikikikikikik569xdeiofujuzi_uikiiikikjm_ikikikikikikikikikikikikik_gggggggggggggggggggggggggggg.png
austere monolith
#

I have noticed a lot of AI image generators support the use as an image 'as inspiration' that can help guide the AI in the right direction. Do you think DALLE2 will ever support this? Wondering if there is some licensing reason why it can't or otherwise

regal lynx
#

If you're talking about making variations of an image, DE2 already supports that feature!

austere monolith
#

I'm not; it's similar though. Other platforms for example allow you make variations using an image as a template but with different prompts. As far as I'm aware DE2 only supports re-generating variations of the exact same image.

#

Unless this functionality exists and I'm just missing it, but it seems like DE2 does not allow you to modify the prompt for variations of an image

regal lynx
#

Okay, I understand now. Currently, DE2 doesn't allow you to add prompts when generating variations. We don't know if that feature will be added

austere monolith
#

E.g., I've seen this elsewhere but playground AI has this for stable diffusion. I think before they moved DE2 to an add-on this was also integrated with DE2 as well, but it's not via the official openai.com site

#

ok

#

hopefully it will one-day be supported then if that is the case πŸ™‚

obtuse cairn
#

Who all can currently use Dall E- the first one?

austere monolith
dull basinBOT
#
<:dallelogo:1000465618258100294> You are using DALLΒ·E 2

The invite you received from OpenAI is for DALLΒ·E 2. The first version of DALLΒ·E, version 1, was never released to the public.

Read more here: https://openai.com/dall-e-2/

regal lynx
#

@obtuse cairn

regal lynx
austere monolith
#

Hmm ok. I cannot seem to find this for uploads, only outpainting. but maybe it is in there somewhere. Cheers.

obtuse cairn
#

Who is using Dall e (first one)if the public uses Dalle 2?

regal lynx
rustic marlin
rugged wing
austere monolith
#

I mean...I think ultimately the AI doesn't care what you want. It is trained and reacts to a particular set of language rules, so you have to work within that framework

rugged wing
verbal crag
cinder sequoia
#

Ok let’s say I want to sell a Dalle generated image in Shutterstock. The TOS say that I must say that it is created using AI. I understand that. When the watermark is removed should I state that it is generated using Dalle? I have read the tos but it seems unclear to me.

static breach
#

been searching thru messages of ppl also having the issue of createImage not being a function, so i tried using a fresh api key, then i also saw that u need to set the auth before the generation code, but i already do in my code. another thing i saw with a person was that they accidentally used the text gen code, but the code i have is merely copied and pasted from the docs with minor edits for my usage

split hazel
static breach
#

oops didnt realize i switched channels in my search

hoary helm
#

guys, is there any command to make the dall e create a image of a whole sentence? like, a jpg of text

rustic marlin
#

do you mean, literal text?

hoary helm
#

yes

#

literal text

rustic marlin
#

the AI doesn't do text, it can only do scribbles that look like text from far

hoary helm
#

i mean i tried with -word "blacked out" vaporwave-

#

and this was the response

#

pretty accurate but idk

#

so i gotta keep trying and see what happends or...?

rustic marlin
#

you'll probabily not get it

hoary helm
#

oh

rustic marlin
#

the AI doesn't understand the concept of meaningful text

#

it only makes an aproximation of the concept of letters

#

but it cannot understand the proper order of the letters

hoary helm
#

i get it

#

thanks

#

sad that i couldnt make it but now i understand better the IA

#

:))

rustic marlin
#

yea, IMO, not beeing able to get text done is a major problem with the current state of this tech

#

but Im pretty sure that soon there will be a way to do that

#

maybe not DALLE, but some AI will do it

hoary helm
#

yes! we could get really cool fonts with that

rustic marlin
#

indeed, describing styles for fonts would be super cool

regal lava
#

yo

obtuse cairn
#

Can anyone tell me what API means in Dall e?

indigo rune
wraith marlin
#

hello, in the website of Dall e it said:
"We’ll invite 1 million people from our waitlist over the coming weeks. Users can create with DALLΒ·E using free credits that refill every month, and buy additional credits in 115-generation increments for $15."

does that include the API?
so can i only make 115 photos with the Dall e API, because im certain i made more than 115 photos and i can still make them?

rustic marlin
#

the API is part of the OpenAI already existing API (which include other stuff such as models for generating text)

#

the credits you get from the Dalle website registration are not related to the usage limits on your OpenAI account

#

they are separate systems

#

with separate billing

#

so can i only make 115 photos with the Dall e API, because im certain i made more than 115 photos and i can still make them?
the API is not free, you pay for each request unless you are on a new account which still have the initial 20$ for testing, but those credits spire

#

the website gives you 115 user for free, ang 15 more each month

winter latch
#

I tried to create some fan art of Bobbin Threadbare from the PC game Loom to the best of my ability. These are some of what it came up with. I like them, except Bobbin's face in the game is supposed to be pitch-black, with only his eyes showing through the hood. The problem is that when typing in a prompt such as "a black face", or any other variations I tried (such as "blackened face", "pitch-black face", "black-colored face" and more", Dall-E's algorithm censors the output because it thinks I'm referring to blackface. Any way around this?

dc5a1b85c2ce5a64.png
#

This is Bobbin's actual face from the game. It is simply, a black, face lol

unknown.png
rustic marlin
#

remingds me of Noita

#

ot Magika

#

actually, it reminds me of something else...

#

let me just...

#

done

errant hemlock
vague flax
#

what are some of the specifications normally you have to write to get something more accurate?

obtuse cairn
#

How to get Promtober Hall of Fame badge

lyric dock
#

how can i get different variations in the haristyle for a uploaded image?

pallid socket
golden ruin
ornate axle
#

How do I use the image extension tool?

rustic marlin
rustic marlin
#

then, position the area that will be generated in a way that it gets part of the existing image to provide context for the AI, and left some blank space to be filled

#

write a prompt and generate

#

rinse and repeat

vague flax
rustic marlin
#

you are welcome =)

placid mortar
#

why does this cost a credit?

unknown.png
#

cant you just zoom in yourself?

rustic marlin
#

what is this?

placid mortar
#

the outpainting editor

zinc marsh
#

That's not a credit cost--that's a keyboard shortcut you can use (Command + 1 on Mac, or Ctrl + 1 on Windows) to achieve that without even having to open the zoom menu.

placid mortar
#

oh icant

I thought it cost a credit but I think it might have just been lagging behind for a moment and when i happened to press that button to zoom in, then checked my credits it was finally updated

obtuse cairn
#

How can I get the Promtober Hall of Fame badge?

zinc marsh
#

You would have had to have submitted a generation during Promptober that received 30 pumpkin emotes. Unfortunately, that event ended after October 31st.

inner zodiac
#

Is the word "femboy" against the content policy?

regal lynx
rustic marlin
#

as long as you keep it SFW, it is ok

regal lynx
rustic marlin
#

yea, I know, but I dont think this is the case for this one

#

he is probabily asking about furry art.. and this can be really wholesome.. or really NSFW...

inner zodiac
#

I just put femboy, not as a nsfw way. It just doesn't allow the word..

rustic marlin
#

uhh

#

oh well, it is probabily blocked then

inner zodiac
#

Shucks

grizzled stag
#

I know they announced they improved the "safety" features, but I'm definitely getting worse results when using the "animal human chimera" tags. It's spitting out a lot more humans and less chimeras.

Screenshot_20221104-232714_Chrome.jpg
frosty quarry
#

Hi! Is output always maximum 1024 pixels?

rustic marlin
regal lynx
frosty quarry
#

Even outpainting?

rustic marlin
#

The area that you generate is 1024x1024

#

The final result can be bigger

trail shuttle
#

hi guys how do i get dalle 2 provide a proper faces, all faces im getting are super mutated

regal lynx
rugged wing
#

Are you allowed to use your DALLβˆ™E images on other AI image generation websites?

wide skiff
#

Yea, clearly says you can use them even commercially

rugged wing
#

Yeah, I get that, but the whole watermark ordeal is what concerns me. Wondering if I still have to mention that it originally was created on DALLβˆ™E before being altered by another AI.

regal lynx
dull basinBOT
rugged wing
# regal lynx The Terms of Use explain this πŸ™‚

So does it mean I am allowed to use it on other AI websites? Is it where it says "(d) Third Party Services. Any third party software, services, or other products you use in connection with the Services are subject to their own terms, and we are not responsible for third party products."?

Sorry... I have severe ADHD + Dyslexia, I struggle with comprehending long text documents sometimes.

regal lynx
rugged wing
halcyon sapphire
#

I just got the content rules violation for this prompt-
β€˜Portrait of Pakistani teenage girl looking up in New England autumn forest.’
Is it the teenage girl part of that? Is Pakistani more likely to be flagged b/c in an effort to avoid racist content?

golden ruin
golden ruin
winter idol
#

DALLE2 API down for anyone else?

regal lynx
#

Please read over the terms of use

dull basinBOT
regal lynx
#

We don't know when/if that will happen. You can always drop it in #placeholder-channel though! πŸ™‚

normal ether
#

why cant it draw nsfw

short salmon
dull basinBOT
normal ether
rustic marlin
#

oh well, my reply got automod'ed xD

rugged wing
onyx osprey
#

hello quick question here. does anyone knows how to make all images from API public?

#

via the same http calls?

golden ruin
#

your generations are private unless you make them public, so no one has an access to it but you. however there's no delete option at the moment

golden ruin
buoyant dawn
blissful swallow
#

Hello, I am a newcomer here and I hope this is the right place to ask questions. The system won't let me ask for: "a realisting painting of Lesbo on the beach writing poems" telling me that violates the content policy, but why? If only the name of a famous poetess activates the filter then it is a very inefficient and ineffective one.

golden ruin
blissful swallow
#

Ok. thanks I will try

#

I just tried and it works wonders.

golden ruin
wheat lichen
#

Is there a way to cross 2 images?

rustic marlin
cinder sequoia
#

The question was skipped as always

dull basinBOT
#
[❓] β€” Q: Can I remove the watermark?

Yes. Avoiding deception is our primary intention. We're not enforcing that outputs are attributed to OpenAI / DALLβ€’E 2. We'll update the content policy to make this clearer in the coming days!

You can read more about FAQs here.

split hazel
#

@cinder sequoia Outputs are not required to be attributed to OpenAI / Dall-E specifically

rustic marlin
# cinder sequoia The question was skipped as always

hey, there are a lot of questions, sometimes it gets buried
you can indeed use it commercially and there is no problem removing the watermark
you may not mislead others about the nature of the work and the AI involvement

cinder sequoia
#

Thanks for the reply

rustic marlin
#

@cyan spruce

#dev-chat message
sounds to me like a browser issue, which browser are you on?

cyan spruce
#

iOS

rustic marlin
#

are you on an old phone?

cyan spruce
#

No, iPhone XR

#

Unless that’s too old

rustic marlin
#

this is a weird bug, that may be caused by some specific version of your browser
this shouldn't be happening anyway for whatever phone or browser you are, I'll notify the OpenAI Team

cyan spruce
#

Thanks so much

rustic marlin
regal lynx
misty vault
#

what is the best way to get an image that looks like it was taken from far away? because using "distant, wide, and far away" have done little to change the perceived distance from the origin to the subject

regal lynx
rustic marlin
#

it is tilt and shift lens

misty vault
rustic marlin
#

try "wide shot"

#

because using "distant, wide, and far away" have done little to change the perceived distance
maybe it is just a matter with the rest of the prompt

#

sometimes, the style you are trying to make has more weight than this words

#

everything on the prompt changes how the AI makes the composition, so, some times, a lot of stuff is ignored or overrided

cyan spruce
cyan spruce
rustic marlin
#

try using the applet it suggests you to install when you open the website, it is a better interface for mobile

rugged wing
#

is there a way to remove watermarks?

regal lynx
#

yes, you can use the inspect element on your browser to download versions of images without the watermark

rugged wing
#

oh that's incredible, thank you!

atomic robin
#

Today the credits they gave us in the artist financing program expire, what time or what?

cedar mural
#

is it possible to generate images with the same character/style repetitively? if i've generated an image of a person, can i feed another prompt to get the same person & style of image as if i was creating illustrations for a book for example. if so, any tips would be greatly appreciated.

regal lynx
zinc marsh
#

I would say it's possible to a very limited extent--you can e.g. erase the background behind the person and subsequently inpaint a new one, thus having the same character in multiple places--but no, it's not possible for DALLE to "remember" and re-generate the same person in multiple poses/from multiple angles/etc., which is probably what you're envisioning.

cedar mural
dull basinBOT
#
[❓] β€” What is Inpainting?

At it's simplest, Inpainting is a way to replace parts of an image using DALLβˆ™E's edit feature.

(There's more you can do with inpainting, this is the most basic and commonly used. Check the #edit-inpainting channel for more ways to inpaint.)

Step 1

Upload an image or select one that you've generated with DALLβˆ™E.

Step 2

Press Edit, and erase the portions of the image you want to replace.

Step 3

Enter a prompt describing the image as a whole (in the example, the prompt changed from "A cat sitting on a bench" to "A dog sitting on a bench") and press Generate.

zinc marsh
#

Now, in your case you're going to want the original person as intact as possible, so what would honestly be better would be to use an external image editor to delete the background instead of doing it manually with the eraser tool on the Editor. (I think most basic programs like Photoshop, GIMP, or even Paint 3D (though obviously use something better if you have it) have a lasso or "magic select" tool you can use to select the character, cut it, delete the rest of the background, and then paste the person back in. That way you have a transparent background DALLE can subsequently fill in.

1.PNG 2.PNG
cedar mural
#

thanks so much - this is really useful!

zinc marsh
#

No problem. Best of luck.

delicate isle
#

who holds the rights of the images the creators of the ai or the person who enterd the prompt

regal lynx
dull basinBOT
atomic robin
#

is there a way to download all the images of the history? because I created a big collection, so is impossible to me download 1 by 1

regal lynx
keen hamlet
# cedar mural is it possible to generate images with the same character/style repetitively? if...

Currently there is no way to create "an asset" in DALLE and store that asset (character or object) for ongoing use in multiple images, like an asset in a game development system. Even if you specify many details of the appearance, there are some random aspects to the generation process. This results in variations that are very difficult to control. You can get a somewhat consistent art look for a series by developing a set of reproducible art style tags, then apply those style cues to various changing subjects using a consistent prompt structure: [subject][style].

cold flame
#

I'm trying to extend an image with outpainting, but it's not really continuing the landscape, it's just mirroring it, all the other generations look similar. Am I doing something wrong?

image.png
rustic marlin
#

also, avout doing 50% or less context, because this sometimes causes the AI to assume the image should be mirrored

cold flame
#

alright, thanks!

#

moving the generation frame seems to have fixed it

cedar mural
rugged wing
#

so I made this awhile back with the generation frame feature

#

and I put in the prompt "guy wearing yellow shirt and black pants, cartoon style"

#

but is that an effective prompt? Im confused on how to actually use the feature effectivly

#

can I just type in "yellow shirt, black pants" or does it need me to type the context and then what I want from it

#

original image for context

NePyIyf1.png
regal lynx
rugged wing
#

is there any reason I wouldn't be able to use dall-e to make images and sell them as NFTs? from what I've been reading it says commercial use is allowed

fading robin
quaint marsh
#

Why can't I generate a picture of "Vladimir Putin with slippers cleaning a truck using a toy car"?

fading robin
quaint marsh
#

oh

#

ok

#

thanks

#

wait

#

so why did "albert einstein fighting godzilla" work?

fading robin
#

i thing dalle2 got it like political "Political: politicians, ballot-boxes, protests, or other content that may be used to influence the political process or to campaign." i think

quaint marsh
#

oh

#

ok

#

thanks

#

btw, is it okay to have both europe and asia roles if i cound as euroasian?

fading robin
#

ye i think

rugged wing
rustic marlin
rustic marlin
rustic marlin
fallen lake
#

Hey guys sry to bother but I can’t seem to be able to prompt anything anymore. I get a message saying that bot is unable to process… any idea ? πŸ™

keen hamlet
fallen lake
#

No I’m sure is something related to the account but I dunno what exactly. I forgot and then had to reset the password.now it acts weird 😦

keen hamlet
#

You may need to send a message to support@oai.com if it is an account problem.

fallen lake
#

Thanks πŸ™

split hazel
#

Please keep questions in this channel related to DALLβˆ™E

primal canopy
#

Hey guys, question - is outpainting part of DALLE API ?

split hazel
primal canopy
#

The doc is not super clear on how we can extend an image horizontally/vertically with masks

dull basinBOT
#
[❓] β€” What is Outpainting?

Outpainting is the process of adding more content to an existing image with DALLβˆ™E. Padded empty space is added around the image, and DALLβˆ™E fills it in!

(There's much more you can do with outpainting, this is the most basic and commonly used. Check the #inpaint-outpaint channel for more ways to outpaint.)

Step 1

Generate or select an image and use the !pad command in the #dall-e-bot channel.

Step 2

Upload your image to DALLβˆ™E and press 'Edit'. Do not forget to erase the watermark.

Step 3

Generate it with a prompt that describes what you want to see in the new image as a whole.

primal canopy
#

Here it shows to use a mask but not clear how this mask should be if I wanted to extend like 20% of the image on top and 20% on the bottom for example

#

To give context, I'm trying to build something that changes images aspect-ratio

#

so if I have a square I want a portrait, I would extend the image on the top and the bottom

split hazel
# primal canopy Here it shows to use a mask but not clear how this mask should be if I wanted to...

I'm not sure how best to describe it in words, so I'll point you towards this old outpainting tutorial from before it was an official feature, and we had to do it manually via Photoshop.
#βŒβ”ƒtips-n-tricks message

This is how "outpainting" works, it's simply the inpainting process plus stitching images together afterwards. The openai.Image.create_edit process is the inpainting process, so to achieve outpainting, you'll need to apply the same fundamental techniques as shown in the video, but with the API.

@rustic marlin will likely explain it much better than me, and I might be getting something wrong as I'm not very familiar with this part of the API.

primal canopy
#

Oh so if I understand correctly, you need to :

  • Crop "out" part of the image and leave a square with X% of the image in and Y% of the image in transparent
  • Apply inpainting to the new image
  • Stitch back the image together ?
#

that's really complex

split hazel
#

Correct that's how it works. There may be very efficient methods of doing it programmatically but I wouldn't be able to help with that myself.

#

I suggest taking a look at #dev-chat for more about the API and outpainting robothumbs_up

primal canopy
#

Thanks for the help in any case ! I was hoping it was something simpler but you can't always get what you want haha

primal canopy
#

@split hazel do you happen to know what a mask look like ? Is it white + transparent area (where generation should happen) ?

rustic marlin
#

Im here for I have ben summoned =P

rustic marlin
#

the api only handles square images

#

the outpainting feature is in fact just a way to implemente the so called inpaint, they both are ways to use the edit method

#

you need to make a few edit requests, then, composite the images back together

#

but in no situation you would be able to upload the entire thing (unless you actually size the image down to fit on 1024x1024)

primal canopy
#

I see, thanks a lot !

tulip galleon
#

Is there a better way to generate a logo? I am part of a gaming clan called n00bs. When I generate images it does a great job but spells "nobs" 🀣

image.png
rustic marlin
#

sometimes it gets small words right

#

but text in general is not supported

tulip galleon
eternal fulcrum
#

thats nice that it worked somewhere, but since all generators are different Dalle isnt really focused on text generation

tulip galleon
rustic marlin
#

hey, please, don't send content from other AIs, keep it to the #ai-discussions when talking about other systems

tulip galleon
#

I've deleted it.

indigo rune
#

If you don’t mind losing resolution you can zoom out too

#

Also you should try photopea it has dalle support

dull basinBOT
#
[❓] β€” Generating texts

DALLΒ·E 2 is not engineered to properly generate text. While it may get a few simple words right now and then, that is the exception rather than the rule. If you need to add text to your image, consider using an external image-editing software.

keen hamlet
kindred edge
#

is there a glossary out there of all the different art styles you can use?

regal lynx
kindred edge
#

thank you!!! sorry i missed that lol

sinful igloo
#

Idk how to say it in English so ill give an example

If someone grabs an art that I created, that person needs the creator consent?

regal lynx
sinful igloo
regal lynx
sinful igloo
#

Ty

kindred edge
#

can i edit an image to a different art style? like say i have a photo of a person, can i prompt dall-e to remake it in a pop art style?

regal lynx
sonic escarp
#

is there a way to download my canvas without the dalle mark for later uploading again?

regal lynx
sonic escarp
#

that changes file size to small

regal lynx
#

πŸ€” I assumed it would be the same image with the same dimensions. If you're outpainting the image, you can always erase the watermark object the editor

sonic escarp
#

nope tried in the square image b4 its a thumbnail in comparison @regal lynx

sonic escarp
regal lynx
#

using the inspect feature on Chrome, I was able to get both of these images from the website. Both images are 1024 x 1024 pixels. Interestingly, I could only download the image without the logo as a .webp format 🫀. With that being said, I can download it from Discord as a PNG file. Strange workaround, but I guess it works...?

unknown.png DALLE_2022-11-10_22.54.10_-_a_cute_dinosaur_holographic_sticker.png
zinc marsh
#

You can actually just rename the downloaded .webp files to .png on your desktop, and all of the file information will still allow the images to render and function properly. (That does still require one to manually stitch the images together afterward, though.)

sonic escarp
#

on the canvas I went into the developer setting and it took away all my outpaintings @regal lynx

#

so it doesn't work for outpaintings glad I saved it at least 😭

sonic escarp
#

manual stitch?

zinc marsh
#

Yes. Since DALLE saves every single inpaint and outpaint one does in your history, you can actually download the individual generations that make up your outpaint individually--without the watermark, if one uses the Inspect tab trick (and if one uses it for individual generations, they do turn out full-size)--and then manually recombine them together in Photoshop or Paint 3D. That's actually what I use to make my outpaints these days--that way I can take the parts of each frame that I like and manually stitch them together to make the full image.

sonic escarp
#

Doesn't that tend to have more discrepancies since when dalle does it, it will alter the image a bit when inpainting/outpainting or do I have my info wrong?

#

I find it does alter my image a bit on the gens

zinc marsh
#

Yes, if you try to combine two separate images over anything other than the border you originally generated them off of, they won't mesh in the slightest. But if you have e.g. distinct areas that are all self-contained (like the nine areas I erased in the example below), then one can simply pick and choose what looks best in each one and paste them in. (Sorry if I didn't do the best job of explaining it.)

image.png
rugged wing
#

Does DALL-E support prompt modifiers (e.g. to pay more attention to some parts of the prompt, use different aspect ratios, negative prompts, etc.)? If so, are these documented somewhere? (I've read the FAQs and the DALL-E 2 Prompt Book, but don't see anything in either of those.)

rustic marlin
dusk hound
#

How do we help Dall e to identify specific things or characters for future prompts?

rustic marlin
#

on edit, always describe the entire thing you want on the frame. for example, lets say you are generating a ful body character.
you could start by the head, but then, when you get to the chest / lower body, you'll get better results if you describe that you actually want a image of the character's lower body, wearing some kind of shirt, holding something. .etc..

#

if you don't do that, you may end up with something like.. another head being generated

rugged wing
#

Is dalify free to run?

regal lynx
#

@rugged wing No, Dallify is linked to the Dall-E 2 API. You can find pricing here (under Image models): https://openai.com/api/pricing/

OpenAI
Pricing

OpenAI is an AI research and deployment company. Our mission is to ensure that artificial general intelligence benefits all of humanity.

rugged wing
#

Thank you

gaunt saffron
#

Could anyone please show me how to generate variations from a picture based on a prompt

#

It seems that I can't do it

regal lynx
rugged wing
#

How the heckerdoodles do I get Dale to figure out text? Like I can get it to sort of do one word text but multiple words is difficult

dull basinBOT
#
[❓] β€” Generating texts

DALLΒ·E 2 is not engineered to properly generate text. While it may get a few simple words right now and then, that is the exception rather than the rule. If you need to add text to your image, consider using an external image-editing software.

unreal monolith
rustic marlin
#

way simpler would be so inpaint it to make it empty and photoshop the text

rugged wing
rugged wing
#

the prompt was for "Old Rock"

unreal monolith
rustic marlin
#

but small words may look correct

vale parrot
#

is there a way to have a bigger generation frame ? I see people doing amazing thing with a big image as an output and I'm wondering how

dull basinBOT
#
[❓] β€” What is Outpainting?

Outpainting is the process of adding more content to an existing image with DALLβˆ™E. Padded empty space is added around the image, and DALLβˆ™E fills it in!

(There's much more you can do with outpainting, this is the most basic and commonly used. Check the #inpaint-outpaint channel for more ways to outpaint.)

Step 1

Generate or select an image and use the !pad command in the #dall-e-bot channel.

Step 2

Upload your image to DALLβˆ™E and press 'Edit'. Do not forget to erase the watermark.

Step 3

Generate it with a prompt that describes what you want to see in the new image as a whole.

golden ruin
#

you can increase the frame with outpainting tool, therwise your generations will be 1:1, 1024 x 1024 always

regal lynx
tiny swan
#

If I have an image that has a few transparent spots, and I want to use Dalle 2 to color in those spots but leave everything else unedited, can I use Dalle 2 that way?

rustic marlin
#

yep

#

thats exactly how inpaint works

#

it fills the transparent areas

#

even if you erase it yourself before uploading it

tiny swan
rustic marlin
#

i think that the website inflates the border a bit

#

hmmm

tiny swan
# rustic marlin i think that the website inflates the border a bit

On the bottom, I have an image I'm working with. You can see the transparent bit. It's also clear that the area is still transparent when I upload it up onto OpenAI.
Then on the top is one of the variations it came up with. It worked outside the limits of the transparent space.

DALLE_2022-11-12_15.08.00_-_Kishibe_Rohan_from_the_Jojos_Bizarre_Adventure_OVA__Thus_Spoke_Kishibe_Rohan__The_Run_.png Untitled292_20221112110459_copy_2.png
zinc marsh
#

I strongly suspect the underlying problem is that DALLE's not great with context, and in its mind the transparent space can be potentially filled with absolutely anything. In this case, I'm guessing DALLE-2 saw some transparent space and a prompt along the lines of "an anime person running," and dutifully tried to fill in the transparent space with an anime person running, instead of realizing the surrounding image was the anime person running. I've actually had the exact same thing happen to me before when I gave it the top image below and asked for "tail of a mermaid in a hallway" in hopes it would fill in the background, only to have it utterly ignore the tail I gave it and return the bottom image instead.
In your case, I'd recommend erasing far more of the image (so the AI has more space to work with, since it will often give up on adding details if it doesn't think it has enough space for them; I've had that before as well when trying to inpaint wings on a person) and have the AI just regenerate the entire legs themselves. If you want, you could try a prompt that explicitly mentions the woman's legs/shoes to draw more focus to it (e.g. "legs of an anime woman on a treadmill" instead of just "anime woman running on a treadmill") to draw more focus to them and maybe reduce the chance of getting a full-body inpaint like you received, but I'm honestly not sure if that would help or just lead DALLE to getting more confused.

Regardless, If you're trying to just do something simple like get different-colored pants on the original image, the easiest way would probably be to hue-shift the original colored image manually instead of trying to replicate it with DALLE. (Though if you're taking a black-and-white image and trying to add color to it, which is what I suspect you're doing, I suppose that isn't exactly possible.)

unknown.png unknown.png
rustic marlin
#

this button

image.png
astral lotus
#

How can I chat in dallify chat

split hazel
astral lotus
#

When will they get activate

split hazel
rustic marlin
#

uhh, could you describe in more detail?

young saddle
#

yo is there a way to upload a photo to dalle and turn it into the style of a painting?

regal lynx
young saddle
#

so there isnt a way to generate a specific kind of variation?

rustic marlin
#

Dalle doesn't do style transfer

#

you can name a style on the prompt

lucid ravine
#

NightCafe has style transfer.

rustic marlin
#

combinign AIs usually have good results..
do the style transfer on another one, then upload to Dalle to inpaint it and Dalle will keep the image style

indigo rune
stray dove
#

Are you using an online SMS service to register? For account security and for your privacy, I do not recommend this because your account can be easily compromised and all your SMS can be intercepted. (I'd recommend an admin delete this post above for you.)

rugged wing
#

Are generations in the β€œcreepy pasta” or β€œSCP” style allowed?

regal lynx
#

@rugged wing As long as they don't break the content policy, yes

dull basinBOT
regal lynx
#

Just be mindful when it states that you can't generate images that are
Shocking: bodily fluids, obscene gestures, or other profane subjects that may shock or disgust. (This includes blood)

split hazel
#

You can't yet as the editor is still in beta, I recommend not closing the editor tab if you plan to continue later

#

Are you using a browser or extension that puts inactive tabs to sleep to save memory?

#

There's probably a way to exclude the editor tab from that process to prevent it from refreshing and losing your progress

unkempt knoll
#

a teratoma with a face, photorealistic

image.png
#

is it possible to get a refund of the credit i spent on that one? it is not to specification in any way.

split hazel
unkempt knoll
rugged wing
#

Hi. How can I create images with dall e 2 here in discord?

regal lynx
zinc marsh
#

There’s not currently a formal setting to do that, but (for individual generations) you can still obtain the non-watermarked version by finding the link to the raw image in the page’s Inspect tab. (Most browsersβ€”or at least I know Chrome doesβ€”have an β€œInspect element” option on the bottom of the right-click menu, that you can use after right-clicking the generation to pull up the link to the non-watermarked version.)

golden ruin
#

hi, what exactly isn't working, maybe i can help?

zinc marsh
#

If it helps, what I personally do when outpainting is download all of the relevant original images individually without the watermarks (using the Inspect tab trick), stitch them together in another image editor to see what combinations looks best, and then only upload to the editor when I need to generate a new piece of the image. It’s hacky and requires an annoying amount of calculations to make sure each image is getting placed in the exact right spot, but it also actually handles both of the problems you’ve mentioned so far: all of the outpaints are saved locally on my computer, so I don’t have to worry about the editor tab possibly refreshing and deleting my work, and I also have all of the images fully assembled without the watermark already, and thus I don’t have to download it from the editor and subsequently have the watermark automatically pasted in the corner.

regal lynx
rugged wing
regal lynx
rugged wing
wraith warren
#

When creating edits does getting realistic results take a bit of luck or is there a good way to edit?

#

For example, I'm trying to make a picture of two robins. I have corrected some of their features but getting their feet right is seeming really hard

regal lynx
torn vapor
#

Anyone know why I can’t download generated images off browser?

regal lynx
torn vapor
#

@regal lynx Safari MacBook

regal lynx
#

hmmm, that's rather strange. What happens when you try to download an image?

torn vapor
regal lynx
#

can you see it in your history to download?

torn vapor
#

Ahhhhh nvm

#

You have to go back to the original

regal lynx
torn vapor
#

Thanks 🀝

rugged wing
#

Hi
I would like to know if this is possible to generate images based on french text ?

rugged wing
#

It seems that yes

golden ruin
heady fulcrum
#

can I ask what datasets is used in dalle?

regal lynx
rustic marlin
#

i mean.. all of the AIs available are, but I think that the Dalle dataset is the biggest one, while some like SD are more hand picked
I would love to see some discussion about that on #ai-discussions

heady fulcrum
rustic marlin
#

Sure, but im just a volunteer, im not from OpenAI, I don't know any more that what is publicly known

heady fulcrum
#

Cuz I'm just a student, that is having a thesis rn. and we're building an application similar to dalle. is there a way that you can give us insights?

rustic marlin
#

sure, btw, if you are planning on making some code, check #dev-chat feel free to ping me for that =)

heady fulcrum
#

any suggestions?

rustic marlin
#

what do you men by that?

#

do they expect you to attempt to use AI to generate """bad""" content?

regal lynx
#

how can someone dislike the dataset when they don't know what's in it πŸ€”

heady fulcrum
rustic marlin
#

what is the problem? it is free uses '-'

#

thats an absolute win

heady fulcrum
#

they don't like us to spend money to in using openai

heady fulcrum
rustic marlin
#

ohh

#

well, im pretty sure yo ucan get enough material out of it with those 18$

#

each image costs 0.02, with 18$, it is a total of 900 generated images for free

#

after that.. yea, you have to pay

heady fulcrum
#

but thats not what the panelist want on us.

rustic marlin
#

im not sure if Im understanding

#

what are you planning to do?

#

do you think you'll need more than 900 images for your thesis?

heady fulcrum
rustic marlin
#

uhh, with "limited" are you talking about the restrictions on the subject of the content generation defined by the content policy? or the restrictions on how much you can use it for free?

heady fulcrum
wraith burrow
#

Hi, is there a problem with editing images? I set a generation frame and a prompt, submit, then it loads, and appears to give me 4 options in the generation frame, but each one is just empty, ie its just a box around a section of the original image

rustic marlin
heady fulcrum
#

how much you can use it.

rustic marlin
heady fulcrum
#

I think the panelist want Unlimited generation of image lol

rustic marlin
#

assuming you use that only to generate 1024x1024 images, you can get a total of 900 images

#

you can generate smaller images and it will cost less

#

if you don't use those 18$ in 3 months, it expires

rustic marlin
wraith burrow
rustic marlin
#

next best option is to get a nice GPU and run some of the Open Source software that does this, but then, it has nothing to do with OpenAI

rustic marlin
heady fulcrum
#

a good specs of computer?

wraith burrow
rustic marlin
#

ohh, yea, thats is exactly it xD

wraith burrow
#

So i couldn't upload a photo and then make it night time or foggy?

#

thanks btw

lucid ravine
#

You could mask out the sky and prompt it for fog

#

It might work.

rustic marlin
dull basinBOT
#
<:dallecredits:1002181155220631562> Q: How do DALLΒ·E Credits Work?

β€’ You get 50 free credits your first month.
β€’ 15 free credits will replenish every month after that, on the same day of the month.
β€’ For example, if you signed up on August 3rd, your free credits will refill on September 3rd.
β€’ If you joined on the 29th, 30th, or 31st of every month, your free credits will refill on the 28th of every month.
β€’ Free credits don’t roll over to the next month, so they expire after month after they were granted – but you’ll get 15 new free credits.
β€’ Paid credits will expire after 12 months from purchase.

regal lynx
#

⬆️ Currently, this is it

#

They're automatically applied to a new account

golden ruin
#

You can't see them when you click on your profile pic?

verbal crag
#

Check how many generations you’ve made in your history sidebar. Each generation = 1 credit.

#

Have you bought any credits recently?

#

Would be hard to track then, you would need to count manually. New accounts should be receiving 50 DALL-E credits, if you think you did not receive yours, please forward your concern to support@openai.com.

spare shoal
#

Can the AI do T-Shirt designs?

#

or how do you do stickers?

golden ruin
# spare shoal Can the AI do T-Shirt designs?

I didn't really get the question, you can turn your generations into designs if you want to. if you mean mockups, like t-shirt mockups, it can give you mockups based on your prompts, but they won't serve as psd mockups. regarding stickers, depends the style you're aiming to. The simple word "sticker" in your prompt can trigger and generate what we regard as stickers. try simple prompt like "a sticker of a cute little dragon" or something...

spare shoal
#

ok thanks

golden ruin
spare shoal
#

Awwwwwww

half saddle
#

I haven't been able to get a full length person either

golden ruin
# half saddle I'm new to DALL-E 2 and AI art in general but I can't find this answer anywhere....

hi, from my understanding unless you specify the position of your subject it gives you cropped in order to preserve the style and continuity if you try to outpaint it, it has to have some base to built future content on. That particular prompt that i came up with was with outpaint in mind since Mucha is generally known for his long vertical posters, so it isnt really surprise that it crops.

half saddle
#

ok thank you. I have tried several other prompts without Mucha (this was the first) and it does same thing. I took it to outpaint but I could only fill out the top or the bottom. But I do understand what your saying about using that name in my prompt.

golden ruin
half saddle
#

Do you know where I can search for those "medium words" to keep within frame? I have spent most of the day trying to find some prompt words for that but found nothing. Thanks again

golden ruin
# half saddle Do you know where I can search for those "medium words" to keep within frame? I...

#πŸ“šβ”ƒresources and #1021130377026351105 have some tips, also you can check this doc https://docs.google.com/document/d/11WlzjBT0xRpQhP9tFMtxzd0q6ANIdHPUBkMV-YB043U/edit#heading=h.9l4yey2tbzos. there might not be one general word that can affect the prompt same way. sometimes even the simplest solution like ending your prompt with "on a white background" can help

Google Docs
DALLΒ·E 2 Prompt Engineering Guide

DALLΒ·E 2 Prompt Engineering Guide (created by rendo1#6021 & luc#0002) Welcome to the Prompt Engineering Google Doc! This document will go over all general modifiers used in prompts with further information, examples in real-life, and examples of how DALLΒ·E 2 interprets the phrase. This is a f...

half saddle
#

Thanks for all your help.

dry maple
#

hi! I've seen people do selfies in a cyberpunk style and I was wondering how I could do this with Dall-e, I've seen people do it with MJ like this

unknown.png
regal lynx
dry maple
#

awww really?

#

sucks :(

fast fossil
#

How can I join dalle 2? Every time I wanna enter it just get me to dall e 1

dull basinBOT
#
<:dallelogo:1000465618258100294> You are using DALLΒ·E 2

The invite you received from OpenAI is for DALLΒ·E 2. The first version of DALLΒ·E, version 1, was never released to the public.

Read more here: https://openai.com/dall-e-2/

regal lynx
#

@fast fossil ⬆️

fast fossil
#

And so, how can I make that erase a part of the image and put another thing?

regal lynx
#

It's all Dall-E 2 πŸ™‚

#

The original was never released to the public

lucid ravine
#

then type in a prompt and generate.

fast fossil
#

Ooohhhh, okey, now I understand. Thank you Sloth and Dinosaur!

winged ridge
#

I've been trying to outpaint a portrait image into a landscape
but I'd like to reference the grass from the right side when making the left.

image.png
golden ruin
winged ridge
#

grassy, hill, blue sky, white clouds

#

Here's the final image

DALLE_2022-11-16_09.56.27_-_blue_sky.png
#

here's teh original

fahmi-fauzi-tower-commission-sm.jpg
rustic marlin
indigo rune
#

That looks amazing

gentle ocean
#

Hey, if I use an image generated with DallE as a music cover, does the DallE logo have to be visible in the lower right corner or can I use the image without the logo? Thx

regal lynx
dull basinBOT
winged ridge
#

was the answer yes or no?
I didn't find it after searching for corner or logo.

rugged wing
#

i need help

#

i changed phone

#

cuz my phone is broken

#

but i think dall e didnt asked me the phone number

#

verification

rustic marlin
rustic marlin
#

are you not able to login?

winged ridge
rugged wing
#

sry for th bad english

rustic marlin
#

if you can login, it should be fine then

rugged wing
#

oh ok

#

ty

#

so there are no problem if i didnt verified the phone number?

dry stirrup
#

I'm wondering if others had this issue (and hopefully a solution πŸ‘€ ) for these artifacts I often get when inpainting.
I'm trying to use DALLE-2 to fix texture seams in the image it originally created, but on most attempts the image it draws ends up getting overly saturated (almost has a burnt look to it).

Screenshot_2022-11-16_154408.png
rustic marlin
#

by my use experience, the more you inpaint, more it will deviate from your original goal

#

the best is to erase as little as possible to give the largest amount of context possible to the AI to avoid such artifacts

lilac hazel
#

Anyone an expert at this thing? I keep failing at what I want to make, guessing my input is just unclear, but unsure how to clean it up..

rugged wing
#

Give me an example if a prompt you used

#

And I will try to help

lilac hazel
#

Trying to create a picture of a large nose, in the colors of the rainbow, smelling smoke that is also in the colors of the rainbow, and it's all supposed to be kind of impressionist, using van gogh as a prompt.
I only manage to get one part correct, but when trying to combine them it goes all haywire.

rugged wing
#

You can try using inpainting, or other photo editing software.

lilac hazel
#

Yeah, that was my solution as well, but it would be fun to make this thing work πŸ˜„

pseudo gate
#

anybody from the team around here ? I completed the feedback form for the creators and was supposed to be alotted some extra credit for that but I didn't receive it.

regal lynx
fast fossil
#

Is it possible to change the a face of someone without changing the person?

#

For example, if you erase the person's face and ask dalle to put a sad face, then it puts a sad face of someone else, not from the se person

regal lynx
#

I'm not 100% sure, but I don't see why it wouldn't be possible πŸ€”

rustic marlin
#

you can probabily get what you want with external image editing tools in combination with dalle

fast fossil
regal lynx
fast fossil
#

A person I upload

regal lynx
#

I don't think you'll have much luck changing the facial expression due to how subtle and complex they are

fast fossil
rustic marlin
#

the problem is that a "sad face" isn't a matter of just the position of mouth and eyes, you would need to inpaint the entire face

#

what you would be able to do is to take a photo, devete everything except the face.. and generate around it

indigo rune
#

Try the trick by using half the image for the person and leave the other half transparent and tell it to generate the same person with a different expression

finite coral
#

how do I get Dall-e to render a photo that's not up close? I want my sisters holding their dog but one with their entire bodies and not something up close

zinc marsh
#

If I can ask, are you attempting to extend an existing photo, or generate a completely new image from scratch?

plucky smelt
#

hey everyone, is there a desktop application for dall e ?

zinc marsh
# plucky smelt hey everyone, is there a desktop application for dall e ?

So far the only methods of using DALLE that OpenAI has publicly released are the website and their API. Given the API is open-source now, someone might have made a desktop application themselves and released it, but I don't know any off of the top of my head--the best place to check for that would be #1037561385070112779 or simply Googling for one. (Though I would obviously recommend checking the source code of whatever applications you end up finding if you can, just to be safe.)

finite coral
#

Im just wanting to generate 1 photo

#

sometimes it does it sometimes not so trying to figure out how to make it happen, force the photo to be zoomed out and not up close

golden ruin
finite coral
#

Thank you

#

so much to read but been trying, but thank you

normal ether
#

why can you not generate blood? it does limit its capabilities when it comes to texture design.

golden ruin
golden ruin
lost valley
#

Hei!
I need help, what style should I use to have such picture

regal lynx
vivid mortar
lost valley
#

Ok

#

Thank

#

And where can I find how to use "add generation frame"?

vivid mortar
#

If you are on a PC, it will look something like this

image.png
vivid mortar
lost valley
#

Oh

#

Great

#

Thank you, I realized how to use it

eternal dagger
#

two eggs toast and coffee,pop art style,roy lichstistein

rustic marlin
agile sequoia
#

I just wanted to make sure you can use an image of your own face now right?

regal lynx
coral ore
#

Since dalle's api is publicly leased now, is there a place to see everyone providing dalle's imaging services?

regal lynx
rugged wing
#

I’m trying to nail a specific style by a popular old French artist Francois desprez specifically his drolatiques pantagruel work. My prompt specifically mentions his name and book and it’s not hitting the mark how can I phrase it properly so that it replicates well

golden ruin
#

what you can do is to find the art direction he was into, then try to frame your prompt in a way that gets closer to the era rather than artists(since it doesn't know who he is), check #πŸ“šβ”ƒresources and #1021130377026351105 to get some ideas

last helm
#

Are there any plans to make public images shared via the web interface discoverable? AFAIK the only way to find these images now is if someone sends you the link

zinc marsh
#

You can view some server members' collections in #1036148240434860033, but in general I don't think OpenAI has announced anything akin to that yet--to my knowledge, at least, given I don't work for the company in the slightest. (You could try dropping that suggestion in #placeholder-channel, though, I suppose.)

rugged wing
rugged wing
#

Can DALLE be used commercially?

#

A friend and I are curious. The plan was to create a card game, where every artwork is 100% AI Generated.

zinc marsh
#

Yes. The primary catch is that you have to make it explicitly clear that the images were AI-generated by DALLE, though. (I believe the Terms of Use on the OpenAI website go into more detail.)

rugged wing
#

Thank you

dull basinBOT
#
[❓] β€” Q: Can I use DALLΒ·E for commercial uses, including NFTs and freelancing?

Yes.

golden ruin
wheat willow
#

Question: I'd like to generate pics similar to those on this website. What specifications should I add apart the theme?

golden ruin
rustic marlin
#

maybe not every

#

but I'm sure that every image that shows up on the loading screen does

wheat willow
#

Sorry, tired

rustic marlin
#

the AI must be trolling me

image.png
jovial bloom
#

Hey guys, is there any way to make AI generate landscape images instead of square?

zinc marsh
#

The AI itself can only generate 1024x1024 images each time, but you can use outpainting to add more generations to the sides of your image and thus extend it horizontally. It isn't cheap credits-wise, but it does allow you to create landscape images. (See https://labs.openai.com/editor .)

DALLΒ·E

Experiment with DALLΒ·E, an AI system by OpenAI

dull basinBOT
#
[❓] β€” What is Outpainting?

Outpainting is the process of adding more content to an existing image with DALLβˆ™E. Padded empty space is added around the image, and DALLβˆ™E fills it in!

(There's much more you can do with outpainting, this is the most basic and commonly used. Check the #inpaint-outpaint channel for more ways to outpaint.)

Step 1

Generate or select an image and use the !pad command in the #dall-e-bot channel.

Step 2

Upload your image to DALLβˆ™E and press 'Edit'. Do not forget to erase the watermark.

Step 3

Generate it with a prompt that describes what you want to see in the new image as a whole.

worthy matrix
#

Hey all im new here . Whqts is the best way to get photos the most realistic

regal lynx
regal lynx
worthy matrix
#

Im trying to get a realistic photo of hamburger on a worldcup trophy on soccer field .. but it wasent that good

#

Thats what i got

regal lynx
worthy matrix
#

Yes thank you

supple sedge
worthy matrix
#

Yeah and ot osnt the trophy of the worldcup

supple sedge
#

xD

supple sedge
worthy matrix
#

Il try that thank you

supple sedge
#

yw. lmk how it went

worthy matrix
#

@supple sedge i dk how to do it .. itried and no success

supple sedge
#

1 sec

#

on phone

Screenshot_20221119_201823_com.brave.browser_edit_962265108119311.jpg
#

don't mind my 1 credit im broke asf πŸ₯²

worthy matrix
#

@supple sedge i did that but how i comine it with what i want

supple sedge
#

ill send u a vid

worthy matrix
#

Are all the photos i get are commercial use? I want to edit them for posts i make

dull basinBOT
#
Commercial use of DALLΒ·E
[❓] β€” Q: Can I sell images I create with DALLΒ·E?

Subject to the Content Policy and Terms, you own the images you create with DALLΒ·E, including the right to reprint, sell, and merchandise – regardless of whether an image was generated through a free or paid credit.

dull garden
#

Is there a way to take a cutout of a person and put it into different backgrounds?

I recently got some professional photos done for work, and for some of them they used a white background. I’d like to be able to cut myself out and put it in front of different backgrounds, like an office building, in front of an audience, at the helm of a pirate ship, what have you. How can I do this?

short salmon
dull garden
#

@short salmon so i tried that.. it kinda works.. but it seems to really want to make my ears bigger

#

it's kinda hilarious what it does with my ears tbh.. it makes them huge.. also makes my hair go crazy

lethal hill
#

is it ok to post api images? i know these generations won't have a dall-e watermark

split hazel
upbeat dust
#

OK. Inpainting. I am at a complete loss. I have basically never been able to get it to do anything like what I want. Usually I have an image of a landscape, and I'm hoping to add something - say a wind turbine. And it's just never ever even close. Any tips? Am I using it wrong?

#

I've tried prompts like "windfarm", or "three wind turbines", or "windmills", or "wind turbines in red sea"

image.png
#

& I've tried rerolling again and again

#

This is the best I've got after maybe 10 goes:

image.png
lucid ravine
indigo rune
#

Try erasing more of the image

golden ruin
somber drift
#

How do I generate more of my picture, I'm in the place in Dall-e where you edit stuff, and I just learned about Dall-e yesterday. Can someone tell me how?

rustic marlin
#

you just need to upload your image on the website, when it asks if yo uwant to crop it, don't crop

regal lynx
pliant shuttle
#

how do i save my images i gotten from my python file

golden ruin
jovial shadow
#

hi everyone, I am wondering why some AI art upsizes (in photoshop) to 300dpi pretty decently, and some does not, is that just due to the type of art style the is ultimately used?

#

might also just be more obvious in any kind of line art or thing with a face

golden ruin
jovial shadow
# golden ruin hi, can I ask for the reference? depends on the style I guess

Right now I am so new - and realized i may not be in the correct community channel - my only reference is just what I am seeing in Photoshop. this for example after I resampled to 300, seemed great, and is only at 15.3% zoom, as I zoom up I completely cannot see the entire pic anymore but all the quality looks good. almost ready to print to art paper.
Other stuff i have done on dall-e2 does not seem to fair as well after resample to 300dpi

golden ruin
#

mainly the generations here are 1024x1024 up to 96dpi

jovial shadow
# golden ruin this was generated by dalle2?

it is hard getting used to art in PS zoomed in at those levels too... you have to get used to doing art on tiny portions of the artboard, without seeing the whole painting sometimes. ...

dandelion was Mid

this one for example is Dalle- but would expect to need to do more work due to art style - bird faces etc

overall I still really need to learn to setup art for print in photoshop - but def dif channel πŸ˜„

image.png
#

I would venture - being well versed in Dalle-2.. at this moment.. what I saw in the midJ trial was impressive, possibly better with abstract art even

arctic topaz
jovial shadow
crude wedge
#

Ok

lucid ravine
#

remember, if you're printing that DPI is part of the equation.

#

And can often be misleading.

#

A 1024x1024 picture at 72 DPI can still be printed at 300 DPI as a 3x3 image.

#

300 pixels per inch is what you should focus on for upscaling, and then if you have the software resample.

mild gyro
#

I'm getting a content warning for "top down view of a hand spread out on some grass" I can get around it, but curious why this is getting flagged

jovial shadow
#

Printer arrives Nov 28th!! so excited

regal lynx
broken burrow
#

Will DALE2 have the capability in the future to use your own start image and blend with its current algorithm?

regal lynx
broken burrow
#

@regal lynx thanks πŸ™‹πŸ½β€β™€οΈ will do

torn shell
#

question: dalle quality has declined since before the release. Has there been any plans to restore it now that there are better alternatives for a cheaper price?

golden ruin
torn shell
silent osprey
#

Hey hey, I've got a question regarding the storage of uploaded images for inpainting. How are these stored and are these stored safely? Are the files accessible by other people except me? Any info regarding this topic is welcome! Thanks!

rustic marlin
#

everything you generate on the labs website is stored by OpenAI, this includes the original image and the inpaint results

#

you can see it on the history tab

#

your history is not public

#

you can select individual images to share, which generates a perma link to them
and also do the same with a collection, which is not public unless you use the share feature on it

#

even for what is public, there is no way to fetch a list of public pictures, you must have the link

silent osprey
rustic marlin
#

the editor does not actualy upload the image

#

it only uploads the frame when you click generate

#

in fact, it doesn't even upload the whole image when it is bigger than the frame

#

it works mostly client side, the UI just composites the API results

bitter delta
#

since December is just around the corner, will there be a promptober kind of event for Christmas?

bitter delta
regal lynx
bitter delta
#

πŸ‘€

unkempt pier
#

I just got access to Microsoft Designer and I want to ask this:
Is the Dalle2 image generation feature in Microsoft Designer limited? (As in credits etc)

regal lynx
mellow dragon
#

What does the "pay as you go" means, displayed on a "pricing" page?

regal lynx
#

This is entirely up to you! Sometimes, I add "phenomenal color" to my prompts to get more "artsy" images. When I try for realism, I don't add anything about color in prompts

golden ruin
#

@jolly lake depends what you want and what you're trying to generate

regal lynx
#

Yes, it will know color and colour

#

When in doubt, add both!

mellow dragon
regal lynx
#

@mellow dragon Every 1024 x 1024 image you make with the API is $0.02 . The pay as you go option will draw that money as you generate images rather than putting $15 into it and receiving credits in the account. Whether you do anything with the image or not, it still costs $0.02 as it used compute resources. The people in #dev-chat will know more about how it works than me!

#

The labs website takes 1 credit for 4 1024 x 1024 images

regal lynx
regal lynx
#

No worries!

gleaming tusk
#

Gm! I'm having issues with the eraser- it keeps giving me the box/generation frame- is this something new or a bug?

rustic marlin
#

i think it is a bug, there where a few times I got stuck on this screen
I downloaded the image and refreshed the page

gleaming tusk
rustic marlin
#

ohhhhh, nvm

#

it is working as intended

#

i tought you where stuck unable to change the tool

#

no, this is the new image editor

#

if you move the generation frame you will be able to do out painting

finite coral
#

guys how do I add a word to my AI picture?

dull basinBOT
#
[❓] β€” Generating texts

DALLΒ·E 2 is not engineered to properly generate text. While it may get a few simple words right now and then, that is the exception rather than the rule. If you need to add text to your image, consider using an external image-editing software.

finite coral
#

Ok

#

thank you

tiny talon
#

on the image generation part.. is there a nsfw switch?

#

like something which prevents the generation of nsfw images

finite coral
#

I dont think you can do anything remotely nude or sexual

#

not even kissing is allowed

verbal crag
tiny talon
tiny talon
raven karma
#

Apologies if this has been asked a billion times already - but is there any ability to fine-tune dall-e based off your own model of training images?

short salmon
raven karma
rustic marlin
#

there is a createModeration method on the API which you can pass a text prompt and it tells you, without charging, if it woulkd be flagged

#

I do'nt think there is a simmilar one for the image in the case of eddits

mental eagle
#

Can I generate a popular Disney characters with Dall E2? foxThink

mellow dragon
lucid ravine
tiny talon
rustic marlin
tiny talon
rustic marlin
#

here we help with the OpenAI API

#

but, if you are using DiscordJS, what you want is channel.nsfw

#

im pretty sure it works on a channel partial, so you don't need to fetch it

tiny talon
rustic marlin
#

you should not make NSFW prompts at all

#

if you do it will reply with an error for content policy violation

tiny talon
#

That's the answer I needed. Thanks!

rustic marlin
#

there is a free endpoint on the API createModeration that will reply if the prompt violates the content policy or not

#

using it is free

#

don't actually try to generate potentially NSFW prompts, use this check before

tiny talon
#

Ahh that's neat. Is there a py snippet I could look into which includes this check?

rustic marlin
#

also, be careful that just because createModeration says it is ok and the AI actually generate it that doesn't mean it is ok with the content policy

#

that just pass on the automated filter

tiny talon
#

Hmm okay. But still having this in my code would help

tiny talon
rustic marlin
tiny talon
#

I will if I encounter problems. Thanks!

rustic marlin
#

you are welcome =)