Discover and read the best of Twitter Threads about #dalle2

Most recents (24)

Wat als.... Van Gogh een ander leven had geleid? Een draadje vol inspiratie met behulp van kunstrobots.

We beginnen met zijn stillevens. Als hij gewoon in Brabant was gebleven hadden we misschien dit nu prachtig gevonden ๐Ÿงต
Of wat als #VanGogh meer het hongerige type was geweest? Misschien hadden we dan dit schilderij als achtergrond op de telefoon?
Als #VanGogh een partner had gehad, hadden we misschien een deze twee schilderijen in print op de muur hangen.
Read 14 tweets
I'm working on a tool for #promptEngineer #aiart ๐Ÿ‘€
This will blow your mind, you'll be able to store prompts and generations from Dalle2, DreamStudio and MJ (from browser)
If you want early access, signup here:
myai.art
A brief breakdown of currently supported features:
โญ๏ธ Select any text and generate it with SD1.5
(Could this be the reason Getty images stopped submissions of AiArt? ๐Ÿค”)
๐Ÿ”ฅ Right click any image and use that as init_img for img2img generation.
Select how strong you'd like the init image to inspire your generation
Read 10 tweets
More DALL-E art.

This is what it's like walking into a store (or plane) and being one of the only masked people there.

Prompt was 'photorealistic painting of evil rainbow clowns exhaling green smoke everywhere'.

I think this captures the feeling pretty well. #dalle2
It's a special kind of hell to walk into a space you need to be in to obtain a required good, like FOOD, and seeing so many unmasked faces, not knowing which of them might be puffing out neurodegenerative viral particles with every breath.

And getting odd looks for masking.
I can't see the virus, but I know it's there. I know that, statistically, at least one of the people I'm looking at probably has it. And they just don't care. They think the pandemic's over. They're merrily breathing all over each other. Some of them will get sick. Some will die.
Read 4 tweets
I suspect people can tell the difference if an image was generated by #dalle2, #stablediffusion or #midjourney.
Here's from the same prompt "A beautiful rendering of a holographic brain." Can you see the difference?
By default, Midjourney is artistically stunning. StableDiffusion is quite good through combining a mix of techniques (i.e. img2img, txt2img, upscaling, etc). Dall-E is better at composing multiple objects. It's weird that each tool has its own strengths and weaknesses.
Read 4 tweets
FSDL Lecture 7: Foundation Models is now available!

This lecture is ๐Ÿ’ฏ new to the course.

We talk about building on Transformers, GPT-3, CLIP, StableDiffusion, and other foundation models.

Brief thread below.

The brave new world of large models is astonishing.

With scale, these models show emergent capabilities that seem truly magical.

At hundreds of billions of params, many GPUs are needed simply to load the model, and API-based access makes a lot of sense. Image
We start old-school, talking about the importance of embeddings and the concept of fine-tuning models. Image
Read 13 tweets
Last night, I paid money to get a 100 more DallE runs. Honestly, I think if you know what you want it's pretty good value.

Particularly getting into to taking ideas generated on Craiyon and varying / improving them.

Some of these are stunning:
The cover for riverofelectrons.bandcamp.com/album/embodiedโ€ฆ was made by Craiyon and I think it's fantastic. Image
But a but low-res. What can DallE do to vary it?

I think it does some pretty amazing things ... ImageImageImageImage
Read 19 tweets
๐ŸŽ‰New preprint! We used Diffusion Models (same one from #dalle2 and @StableDiffusion ) to generate 3D MRI images of the brain conditioned on several covariates and make 100k synthetic brains openly available
Paper: arxiv.org/abs/2209.07162
Dataset: tinyurl.com/32p4hu7d
1/n
with @jessdafflon @pfdacosta @PTudosiu Virginia Fernandez, Parashkev Nachev, Sebastien Ourselin
@mjorgecardoso
In a collaboration with @NVIDIA Cambridge-1 supercomputer

2/n
We adapted latent diffusion models (arxiv.org/abs/2112.10752 developed by @robrombach and @andi_blatt) to learn to generate 3D high-resolution medical images. Due to its scalability, we were to train these models on images with millions of voxels!

3/n
Read 10 tweets
Family

Today I would like to share this series of #dalle2 AI-generated images. (thread for detail and process) Image
The series uses an initial prompt:

"Brilliant 8K portraiture from the film [film title], featuring people wrapped in faded blue linen, standing in the desert, directed by Annie Leibovitz"

to generate starter portraits in Leibovitz style.
The prompt is not a secret incantation or anything, it's just a rough guess at a good starting point (after generating tens of thousands of images with Dalle).
Read 17 tweets
1/ As soon as StableDiffusion landed we dropped everything to build a GENIE! ๐Ÿงžโ€โ™‚๏ธ

๐ŸŽ™ Voice UI
๐Ÿ–ผ AI Art
๐Ÿ˜Ž AR previews
๐Ÿค‘ Instant purchasing
๐Ÿ“ฆ On-demand production

Could AI make every shopperโ€™s wish come true? ๐Ÿค”๐Ÿ‘‡

#ai #aiart #stablediffusion #dalle #dalle2 #vui
2/ Shopping is hard when you want something specific but arenโ€™t sure where to find it. New models from @StabilityAI and @OpenAI can create a detailed image of anything you can imagine, so skip searching โ€” ask for exactly what you want and print it on-demand! ๐Ÿ–จ๐ŸŒฝ
3/ The patterns in our prototype were created using the prompts you hear in the video and the tiling technique we developed. Each frame of animation is a real step of the generation process. Youโ€™re watching StableDiffusion dream in real time ๐Ÿ’ญ๐Ÿ‘€
Read 7 tweets
It is extremely instructive to follow the exponential growth of diffusion models in AI art. At present 3 major players @openAI with Dall-E, @midjourney founded by @DavidSHolz, and @StabilityAI founded by @EMostaque. Each player has taken a different path in their release.
More capable systems exists from Google (i.e. Imagen imagen.research.google ) and Facebook (Make-A-Scene ai.facebook.com/blog/greater-cโ€ฆ), but these aren't available to the public.
I want to discuss how players release their AI products into the wild. #dalle2 has been available since July 20th after opening up a restrictive waitlist 2 months earlier. I had the chance to refine my prompting skills since then:
Read 19 tweets
I really wanted to illustrate long stories with Stable Diffusion! So, hacked together this pipeline:

Long text ->
GPT-3 suggests illustration ideas ->
GPT-3 translates from English to "prompt-English" ->
Stable Diffusion outputs images

Open source ๐Ÿ‘‡๐Ÿฟ
github.com/sharonzhou/lonโ€ฆ
So here's the gist of "promptgen". Without any code.

1. Copy-paste these good "prompt-English" examples, from this file in the repo: github.com/sharonzhou/lonโ€ฆ
2. On a newline, write your prompt in plain English
3. Generate prompt-English suffixes and use it in your fave model!
Read 6 tweets
1/ Using AI to generate fashion

After a bunch of experimentation I finally got DALL-E to work for video by combining it with a few other AI tools

See below for my workflow -

#dalle2 #dalle #AIart #ArtificialIntelligence #digitalfashion #virtualfashion
2/ First I used DALL-E to generate outfits. I did this by erasing parts of my existing outfit and inpainting over it

Btw when I erased the entire outfit, the results didn't look as good. By keeping parts of the original, DALL-E was able to better match color and lighting
3/ But hereโ€™s the challenge. DALL-E works great for individual pictures, but itโ€™s not designed for video. It wonโ€™t give you consistency from frame to frame.

Here was my early experiment. See, no consistency between frames
Read 9 tweets
ืœืงื— ืœื™ ื™ื•ืชืจ ืžื™ื“ื™ ื–ืžืŸ ืœื›ืชื•ื‘ ืืช ื”ืฉืจืฉื•ืจ ื”ื–ื”, ืื‘ืœ ืื—ืช ื•ืœืชืžื™ื“:
"ืจืฉืช ื ื•ื™ืจื•ื ื™ื", ืžื” ื–ื” ื•ืื™ืš ื–ื” ืขื•ื‘ื“.
ื‘ื’ืœืœ (ื‘ื–ื›ื•ืช!) @StavKislev ืจื•ื‘ ื”ืขื•ืงื‘ื™ื ืฉืœื™ ืž #ืคื™ื“ืจืคื•ืื”, ืื– ื›ืœ ื”ืื ืœื•ื’ื™ื•ืช ื›ืืŸ ื™ื”ื™ื• ืžื”ืชื—ื•ื.
#ืคื™ื“ืื˜ื” , #ืคื™ื“ื˜ืง ๐Ÿงต
(ื›ืจื’ื™ืœ - ืื™ืŸ ืฆื•ืจืš ื‘ื™ื“ืข ื‘ืžืชืžื˜ื™ืงื” ืื• ื‘ืžื—ืฉื‘ื™ื. ืื• ื‘ืจืคื•ืื”.) Image
(1/16) ื ืชื—ื™ืœ ืจื’ืข ื‘ืฉืืœื” ืคื™ืœื•ืกื•ืคื™ืช - ืžื” ื”ืžืฉืžืขื•ืช ืฉืœ ื”ื‘ื™ื˜ื•ื™ "ืœื™ื™ืฉื ื™ื“ืข"? ืžื” ื”ืชื•ืฆืจ ืฉืœ "ืœืœืžื•ื“"?
ืœืžื™ื“ื” ื”ื™ื ื›ืžื• ืžืคืขืœ.
ืืœ ืžืคืขืœ ื›ืชืจ ื ื›ื ืกื™ื ื‘ืœื•ืงื™ื ืฉืœ ืคืœืกื˜ื™ืง, ื•ื™ื•ืฆืื™ื ื›ื™ืกืื•ืช ืžื”ื•ื“ืจื™ื ื•ืžื•ื›ื ื™ื ืœืงื˜ื˜ื”. ืžืฉื”ื• ืงื•ืจื” ืฉื ื‘ืืžืฆืข ื‘ืžืคืขืœ ืฉืœ ื›ืชืจ.
ื›ืฉืจื•ืคื ืžืื‘ื—ืŸ ืžื—ืœื”, ื”ื•ื ืžื›ื ื™ืก ืืœ ืชื•ืš ื”ืžื•ื— Image
(2/16) ืืช ื›ืœ ืžื“ื“ื™ ื”ื—ื•ืœื” - ื•ืคื•ืœื˜ ื”ื—ื•ืฆื” "ืขื’ื‘ืช"!
ืจืฉืช ื ื•ื™ืจื•ื ื™ื ื”ื™ื ื ื™ืกื™ื•ืŸ ืœื—ืงื•ืช ืืช ืžื” ืฉืงื•ืจื” ืฉื ื‘ืืžืฆืข, ื‘ืฉืคื” ืฉืžื—ืฉื‘ ืžื‘ื™ืŸ.
ื”ืจืฉืช ืžืงื‘ืœืช ื”ืžื•ืŸ ืžืกืคืจื™ื, ื•ืžื•ืฆื™ืื” ื”ื—ื•ืฆื” ื”ืžื•ืŸ (ืื• ืžืขื˜) ืžืกืคืจื™ื. ื›ืŸ, ื–ื” ืขื“ ื›ื“ื™ ื›ืš ืคืฉื•ื˜. ื™ืฉ ืฉืœื•ืฉื” ืขื ื™ื™ื ื™ื ืžืขื ื™ื™ื ื™ื:
ืžื” ื”ื ื”ืžืกืคืจื™ื ื”ืืœื”, ื•ืื™ืš ื™ื•ืฆืจื™ื ืื•ืชื ื•ืžื” ืžืฉืžืขื•ืชื. Image
Read 17 tweets
We are living in the future. For someone like me who basically failed art class in grade school, #StableDiffusion can now augment my incompetent artistic skills

Thank you @StabilityAI @TwitterCreators @Twitter Image
@StabilityAI @TwitterCreators @Twitter To be honest, it wasn't clear to me how this was going to end up when I started, but wanted to see how much you can nudge #StableDiffusion

Here's @Starbucks Image
@StabilityAI @TwitterCreators @Twitter @Starbucks And here's @Apple

This one required a tag team with #DALLE2 @OpenAI Image
Read 12 tweets
init images in #stablediffusion explained! โœจ๐Ÿ‘‡๐Ÿผ Image
the use of init images consists of a simple trick that allows us to start our generations from an image.
init images give us a lot of control over our generations.

we can choose where to place each element and what style, colors, or pose it should have. Image
Read 23 tweets
Alfred Hitchcock serving tea to the MGM lion:

1) how its started (original real photo)
2) how it went in #stablediffusion
3) how it went in #dalle2
not fair on the #dalle image i choose for quality, but it got the composition right. here is a better render but not the right composition:
would welcome other people to try on #dreamstudio, #midjourney, or #dalle
Read 4 tweets
AI fashion show using dall-e to generate hundreds of outfits of the day after tomorrow. An interesting way to brainstorm costume and fashion design ideas.

Collab with @shyamagolden

@OpenAI #dalle2 #dalle #aiart
Some of the fits that didnโ€™t make the cut
Used @runwayml to help roto all these dall e outfits
Read 5 tweets
Exciting news: #Parti and #Imagen teamed up to create a hybrid system with Parti creating 256x256 images which then recieve Imagen super resolution to produce 1024x1024 pixels! See the diagram below for how it works.

See thread for more info and new images with this system!
As @savvyRL noted in this great thread on #dalle2, #imagen and #parti, Parti had a small super resolution model. This definitely reduced Parti's visual impact. As Imagen is a cascaded model, it was a natural pairing to use Imagen's SR on Parti's outputs!

This required experimentation and tweaking: big props to Imagen's @mo_norouzi, @Chitwan_Saharia and @wchan212 for working on this with Parti's @alex_y_ku and @JHYUXM.

Parti's beaver librarians (fig 19,
arxiv.org/abs/2206.10789) now enjoy greater detail! (Note combined watermark.)
Read 15 tweets
Some thoughts on #StableDiffusion vs #DALLE2 vs #midjourney

They are actually all different (complementary?).

First this whole field was catalysed by @OpenAI releasing CLIP publicly last year (!) which @RiversHaveWings, @advadnoun and others built on openai.com/blog/clip/
This allowed image generation to be guided, the output of which was surprising and led to a burst of developer and artist creativity from @ClaireSilver12 and many others.

Proud to say we supported & funded most of the open tools in this space.

It is now good enough, fast enough
So #dalle2 is a model and a service.

It is focused on a certain usage subset that will broaden.

Inpainting is itโ€™s best feature but by default it is random and best used for ideation and more corporate usage, hence itโ€™s clear training on licensed stock images
Read 13 tweets
Ayer me llegรณ la invitaciรณn que solicitรฉ para probar la IA DALL-E 2. Probรฉ los 50 crรฉditos que me dieron y algunos resultados son increรญbles. Os enseรฑo los mรกs interesantes. #dalle2 #dalle2art
"Warriors dressed a magical armor figting against fire dragon in a vulcan, digital art". Aquรญ use mal el tiempo verbal para indicar que estuviesen vistiendo una armadura mรกgica, aรบn asรญ fueron buenos resultados. Salvo la รบltima, las demรกs son geniales. ImageImageImageImage
"A monkey sell fruits in sea city, digital art". Estas fueron exactamente lo que querรญa. #dalle2 #dalle2art ImageImageImageImage
Read 29 tweets
SEO friends! ๐ŸŽบ New content๐ŸŽบ E-Commerce 404 Pages: How To Make Them Smarter In SEO๐ŸŽˆ๐Ÿ›’

โœ… Create a multimodal search engine w/ @JinaAI_
โœ… Integrate the #API to create a custom 404
โœ… Use #dalle2 and #gpt3 for layout and copy

Links to the post/colab/@streamlit follow ๐Ÿงตโ†“
โœ๏ธ Blog Post โ†’ wordlift.io/blog/en/e-commโ€ฆ
๐ŸคŒ 404 Error Page โ†’ product-finder.wordlift.io/the%20hat%20I%โ€ฆ (test whatever you like by changing this url)
๐Ÿ Colab โ†’ wor.ai/smart-404
๐ŸŽˆ Streamlit Playground (by @JinaAI_) โ†’ nowrun.jina.ai/?host=grpcs://โ€ฆ

2/
In addition to the recommended products listing, I also created the layout for the page using DALLยทE 2 and the message copywriting w/ GPT-3 ๐Ÿค“

Biases aside ("fashion" <> "woman"), it's a good starting point and something to keep experimenting with.

Happy SEO-automation!

3/3
Read 3 tweets
Playing around with AI generated art by @OpenAI's DALL-E. The results are stunning. #dalle2
First up: The Toronto skyline by Claude Monet Image
In the style of Vincent van Gogh's Starry Night: Image
"The Monkeypox virus, as if corporate branded from the Art Deco era" Image
Read 5 tweets
So maybe you're familiar with the weird world of DALLยทE, but how much do you know about #Midjourney, the OTHER super-duper text-to-image AI?

Well, between you and me, it's gotten REALLY GOOD recently โ€“ here are some stunning examples that show why you should give it a whirl ๐Ÿ‘‡๐Ÿงต
DISCLAIMER: all images have been plucked from the stunning community gallery that Midjourney offers paid users - of which more in a bit!
PART ONE: SHE HAS THE *RANGE*

@Midjourney perhaps has a reputation for creating amazing but very AI-y images like these. Super awesome, but with the v3 model, MJ is capable of so much more...
Read 21 tweets
Petite promenade parmi les moutons รฉlectriques de #DALLE2

Par curiositรฉ j'ai choisi un sujet lรฉger :
Comment le machine learning illustre ... les #digitalhumanities

Mais avant cela, pour le plaisir :
voici donc un shiba inu lisant l'avenir dans des frites.
Oui. Vraiment. a fortune-telling shiba inu reading your fate in french frie
le sujet est donc :
"in the semi-darkness of the late afternoon, digital humanities scholars reading open science on a holographic screen in an 18th century library with a cup of coffee sitting on the varnished wooden desk , digital art"
#DALLE2 #humanitesnumeriques

2/n
ร  part les visages (point faible habituel du ML, peut รชtre mรชme bridรฉ sur #DALLE2 ?), c'est plutรดt convaincant ...

Essayons maintenant en pixel art

3/n
Read 20 tweets

Related hashtags

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!