If I were to use their API at the same volume I use Stable Diffusion in my spare time, I’d be spending $2,000 a month. The only thing “open” about this is your wallet.
@@anazi unsure if phones are capable of handling that much. But you can always simplify a model for simpler worse quality results, retrain that and embed inside of your app
Shortly before one of their big reductions in the amount of free renders given each month, I did a huge batch of "generational" faces, starting with one made using Disco Diffusion's face model, then making multiple batches out of each variation from that, down to as many levels of recusion as I was able. It converged toward ghostly 3d cartoon characters, very pixar-looking.
For those who tried variation and didn't work, it's because at 3:45 the path is to call mona in img folder but the mona.png is in the root. Just move the mona.png to root and it will work fine.
At this point i see no reason to give spotlight to Dall-e when Stable Diffusion exists, seems to give much better results with less effort in prompt, has no restriction in what you can query and is free for self hosting or incredibly cheaper when deployed in cloud
OpenAI POV: "For your own safety we can't share the model despite our mission statement saying we want openness in AI. Because nude photo generator 9000 and cleverbot 2 are too dangerous to release to public. Pay us money to use them also we limit what you can use it for"
@@speedfastman so, false advertisement is ok now? I can name a restaurant "Vegan Paradise", only offer meat dishes and when they complaint say "Hey man if you fall for the name that's on you “? Really?
A similar completely free API for stable diffusion already exists. It's called stable horde and it has a free API, although if you contribute computing power you will have priority over other people.
How is "createReadStrem" deifned on the second example? I just get that it is not defined. If it gets defined then it says that "await" cannot be used outside of an async function so not sure how you get that code to work...I'm using Node 18+
hello fire ship can you make a video on how to achieve something like the pinned github repos the way you can drag them and organize them i tried to but its so so hard
const src = './src/image-1.jpg' const result = await openai.createImageVariation( createReadStream(`.img/${src}`), 1, "1024x1024" ); Keep getting an error createReadStreamis not defined, do i need to import that from somewhere?
Nitpick of the day: Thumbnail is a dumbnail, all art is artificial. But then, Tool Assisted Art doesn't work either because almost all art is tool assisted. What about erm... algorithmic art? Yea that works I think.
I thinks that term is reserved for Art made with non-semantic, recursive image creation models (like procedural shapes and rotations). The image generator does use an algorithm, but with semantic understanding of the hyperparameter space supplied by a deep learning model, not a recursive "rule"
@@jeffreyjdesir Yea I suppose designing and rendering a good looking fractal would fit the term better. Eh, "artificial art" does the job in the end, even if it's a redundant term it still gets the point across.
OpenAI unlike its name was gatekeeping AI image generation(Elon Musk is involved, there's no surprise there). SD release was probably a big kick in the butt for them. Dall-E is pretty much irrelevant now.
this is not a tutorial ... it's a glorified documentation dump ... it's fine, but sadly the people that actaully can write an AI could be counted on the palm of a hand, while everyone here even making money saying " yeah, i code AIs, i'm like so smart, 10k in 5 minutes, take that single mother working 2 jobs , flex "
JS is the most used language in the world and the language of web development. I'm a data scientist so i never use JS, but i can't really hate on JS when it's clearly an effective language.
@@Frozander 😂😂😂 yeah to be seriously 200mb for backend framework and another 100mb for frontend just to render a web page ?! the whole linux kernel is under that !
If I were to use their API at the same volume I use Stable Diffusion in my spare time, I’d be spending $2,000 a month. The only thing “open” about this is your wallet.
Couldn't have said it better
Is there a way we can add Stable Diffusion on a React Native App ?
@@anazi you can run it on your server and create an api for it. probably someone has already done that
@@katech6020 I though we could use the power of their phones instead of us paying the bills 😆😆
@@anazi unsure if phones are capable of handling that much. But you can always simplify a model for simpler worse quality results, retrain that and embed inside of your app
Shortly before one of their big reductions in the amount of free renders given each month, I did a huge batch of "generational" faces, starting with one made using Disco Diffusion's face model, then making multiple batches out of each variation from that, down to as many levels of recusion as I was able. It converged toward ghostly 3d cartoon characters, very pixar-looking.
let us see 👀
let us see 👀
let us see 👀
using stable diffusino might be quite a bit cheaper, you could host it somewhere where you pay a fraction of dalle for the compute time
For those who tried variation and didn't work, it's because at 3:45 the path is to call mona in img folder but the mona.png is in the root. Just move the mona.png to root and it will work fine.
At this point i see no reason to give spotlight to Dall-e when Stable Diffusion exists, seems to give much better results with less effort in prompt, has no restriction in what you can query and is free for self hosting or incredibly cheaper when deployed in cloud
My thought exactly. Dall-E is irrelevant after Stable Diffusion
There are google collabs with Stable Diffusion. Completely free.
I'm curious what would be the outcome of a feedback loop between this and an image captioning api
Random noise
OpenAI POV: "For your own safety we can't share the model despite our mission statement saying we want openness in AI. Because nude photo generator 9000 and cleverbot 2 are too dangerous to release to public. Pay us money to use them also we limit what you can use it for"
Stable Diffusion: "Fine, I'll do it myself"
You expect a company to run on thin air instead of money?
@@speedfastman Nah, but if you are not going to be open, don't call yourself open. ai
@@munyxxi Hey man if you fall for the name, that's on you.
@@speedfastman so, false advertisement is ok now? I can name a restaurant "Vegan Paradise", only offer meat dishes and when they complaint say "Hey man if you fall for the name that's on you “? Really?
"I am using a picture of myself writing some code" The self burn out of no where got me good LOL
4:12 *Tool fans heart rate suddenly spikes*
A similar completely free API for stable diffusion already exists. It's called stable horde and it has a free API, although if you contribute computing power you will have priority over other people.
Thank you for this!
Amazing video bro
What software do you use to get that cursor effect in the screen recording? would be really helpful
You have no idea how much time i stuck into /g/ (and a certain other board) cloning the openAi leaked repo, but that stuff is insane
How is "createReadStrem" deifned on the second example? I just get that it is not defined. If it gets defined then it says that "await" cannot be used outside of an async function so not sure how you get that code to work...I'm using Node 18+
When saving the file from the fetch, you could have just piped the responses readable stream directly into a fs writestream.
"A picture of myself writing some code" xD
can you upload a image and then use a prompt for a different variations?
"a clown writing JavaScript code" 🤣 my hope is that one day Fireship will megachurch and I'll be a devoted follower. Best channel on YT
Hugging face is what OpenAI wanted to be when it grew up
would it be possible to create a variation with a prompt?
thanks for explaining the mask!
Love your videos
OPENai is now OPEN? wow, who would've thought
Vai ser uma mão na roda para a arte
Is it okay to offer a SaaS like this using OpenAi Dalle-2 API ? There won't be any problem for commercial usage ?
"I'm using a picture of myself writing some code" Hahaha, I spit out my water. Thanks a lot.
I don’t get how you can use await when it’s not inside of an async function
hello fire ship
can you make a video on how to achieve something like the pinned github repos
the way you can drag them and organize them
i tried to but its so so hard
Cool, now I can make tinder pictures where I have friends 😆probably really weird ones though
after seeing these comments please do a video on stable diffuaion or other more open ais
How about little tutorial and comparing other types of image AI and look for a free one. Right now there are multiple solutions.
Stable diffusion is open source so I'm sure someone made a saas of it right
off-topic question: why not fireship?
love the video
Is it really unlimited tho? The API has infinite credit?
I meant unlimited in terms of creative possibilities. You still have to pay for it and it's rate limited.
@@beyondfireship Yeah thanks for reiterating man
the username... well played 🤣
const src = './src/image-1.jpg'
const result = await openai.createImageVariation(
createReadStream(`.img/${src}`),
1,
"1024x1024"
);
Keep getting an error createReadStreamis not defined, do i need to import that from somewhere?
did you find the answer?
Would be cool to use it to restore images somehow.
is axios error = true
can anyone help me deal with that
rip my art career
If this was 2020 you could've created bunch of shitty images and sell them as NFT 🤣
RIP Bob
Hey I'm new to programming, can anyone please tell me how he is auto generating line of code is a single click?
He is reversing the video, where he starts from the full code and removes lines in groups.
@@yurqua oh never thought of it, thanks
4:00 unlike humans, which are original
🤯🤯🤯
Just discovered microsoft image creator for AI image generator.
I am waiting for midjourney api. It is so much better.
I'm stuck at that "give me money" part please help fire
Can you do a tutorial on blobs and array buffers
Is it really unlimited?
unlimited what? paying? you can use infinite times and pay infinite times as the video mentions
@@YuriG03042 I was referring to dall-e credits, some apis would use those credits in lieu of paying, I'll keep using the ui version for now.
Nitpick of the day:
Thumbnail is a dumbnail, all art is artificial.
But then, Tool Assisted Art doesn't work either because almost all art is tool assisted.
What about erm... algorithmic art? Yea that works I think.
I thinks that term is reserved for Art made with non-semantic, recursive image creation models (like procedural shapes and rotations). The image generator does use an algorithm, but with semantic understanding of the hyperparameter space supplied by a deep learning model, not a recursive "rule"
@@jeffreyjdesir Yea I suppose designing and rendering a good looking fractal would fit the term better.
Eh, "artificial art" does the job in the end, even if it's a redundant term it still gets the point across.
@@BladeTrain3r amen to that - soon the artificial in "AI" will have the same redundancy...cause we will be but monkeys compared to GP(T-1000)
How i get what i pays for
It's literally flow of electricity
Wooooooo
This type of AI will be the death knell of illustrators.
open ai should change its name to closeai
Tool fans developers go crazy xD
"open"ai
partnered with Microsoft and now no longer a non-profit. they only just didn't change the name yet
hot off the press
I think might make a video about how use a library to make something like this yourself😮
It's not free 'Dall-e API' .
OpenAI unlike its name was gatekeeping AI image generation(Elon Musk is involved, there's no surprise there). SD release was probably a big kick in the butt for them. Dall-E is pretty much irrelevant now.
Careful with lawsuits... others' work was used to train these models
Elon owns everything
Kinda misleading calling it “OpenAI” while it has nothing to do with Open source.
this is not a tutorial ... it's a glorified documentation dump ... it's fine, but sadly the people that actaully can write an AI could be counted on the palm of a hand, while everyone here even making money saying " yeah, i code AIs, i'm like so smart, 10k in 5 minutes, take that single mother working 2 jobs , flex "
Art is by definition artificial and this makes everything much more weird.
#stopusingAI #stopArtthievery
mid
jorney is far ahead
1st
Erster
It's basically a tutorial about how to bankrupt artists. Nice
I can't be the only that gets garbage images from these models 90% of the time
you know nothing Jon Snow...
all this sounds like chinese to me.... i need a noob guide from the start
I see Elon musk I click
js suck
i am considering switching paths because of that
the web is just bloat
Oh boy that sounds familiar. Do you watch Luke Smith by chance.
JS is the most used language in the world and the language of web development.
I'm a data scientist so i never use JS, but i can't really hate on JS when it's clearly an effective language.
@@Frozander
😂😂😂 yeah
to be seriously 200mb for backend framework and another 100mb for frontend just to render a web page ?!
the whole linux kernel is under that !
@@bobaGogo the problem not only withe language but also with the framework
it makes it even more bloat
Switch to Rust
I see Elon musk, I throw up
VIM is superior to everything, prove me wrong. Protip: you can't.
Why would I need to prove you wrong? The burden of proof is on you 😂
@@-7n Not an argument >:->
@@socialkruption no... it is...
@@-7n no ;-)