JavaScript is not available.

Aug 13, 2022

"Voyage through Time" is my first artpiece using #stablediffusion and I am blown away with the possibilities... We're crossing a threshold where generative AI is no longer just about novel aesthetics, but evolving into an amazing tool to build powerful, human-centered narratives

2:19

480.8K views

288

3,479

11.3K

The only inputs for the above video are these real images + the audio track. Pretty happy with results so far, getting the video to be smooth took me soooooo much work. Thanks to

@j_stelzer

for the LatentBlending idea which was tremendously helpful.

I've been working incredibly hard on a (by now, very complicated) algorithm to do smooth video interpolations between arbitrary input images based on #stablediffusion for our

@eden_art_

project. Here's a first, not-super-great-yet attempt to make that algo audioreactive:

0:46

523 views

After looking at too much #stablediffusion video jitter over the past weeks, here's an unreleased, buttersmooth, good-old StyleGAN audio-reactive video I made about 2 years ago as a test render for

@maxcoopermax

. Full, HD version on my vimeo: vimeo.com/manage/videos/

2:19

958 views

Alexander S

@devdef

Jan 28

Seaside pixels for you. #stablediffusion #warpfusion

0:08

3.6K views

103

Inside the Lab: Taking Atlas From Sim to Scaffold

Jan 22

Here are some of the input stills used to create the interpolations, all credits to

@jbardakos

0:14

262 views

When diffusion of physical ink on paper meets (neural) diffusion in-silico.. Having great fun creating these AI-generated cybernetic isomorphisms using some amazing ink drawings by

@jbardakos

#stablediffusion #real2real

0:14

Ok, so apparently this interface is hcaptcha by

, not Google

Wait... what? Is

training a txt-to-image model using crowdsourced human feedback?

313

Ait, for people late to the party, the slightly-more technical behind-the-scenes video def gives a much better understanding of what's going on here:

youtube.com

How does Atlas recognize and interact with objects? How do we develop new Atlas behaviors? Why is manipulation important for the future of robotics?Join our ...

Victor M

@victormustar

Jan 18

👀 Impressive ML app: using Whisper to transcribe any youtube video and then perform diarization to identify each speaker 😚 hf.co/spaces/vumichi

139

792

Jan 18

Could someone with solid robotics / Boston Dynamics knowledge tell me how much of this demo is fully scripted / hardcoded? I'm pretty sure most of it is, like if you moved the bag 1m and run the same code, it would simply pick up air. The running is hopefully "real"?

Quote Tweet

Sawyer Merritt

@SawyerMerritt

Jan 18

Boston Dynamics shows of their robots latest skills

1:10

21.4M views

Seth Frey

@enfascination

Jan 16

The key 24 words: Organizations are collectives whose members mobilize diverse resources toward a shared goal Institutions are collectives whose members mobilize a shared resource toward diverse goals.

Free App idea: - Take a photo of your fridge - Run object detection to count vegetables and other cooking ingredients - Ask #ChatGPT to spit out some recipes 🎉👩‍🍳🥦🍽️

779

Jan 4

Been working on some audio-reactive #StableDiffusion features for our #Real2Real algorithm. More stuff & code coming soon! Music by the one and only

Quote Tweet

Sound on!!

Something in progress from our lab... Audioreactive #stablediffusion #real2real

0:51

3.8K views

Dec 25, 2022

Been working on this algorithm for a while now: no more prompt engineering, no more style modifiers, just pure and simple 𝒊𝒎𝒂𝒈𝒆-𝒕𝒐-𝒗𝒊𝒅𝒆𝒐. Stay tuned for more.. 😉 #real2real #stablediffusion

Quote Tweet

Eden

@eden_art_

Dec 25, 2022

0:27

New life is growing in our garden... #real2real #Stablediffusion

279

Eden

@eden_art_

Dec 24, 2022

Merry Christmas everyone! #real2real #stablediffusion

0:15

839 views

Dec 13, 2022

You can now essentially take any SD output image and make it look like you! I also see that LoRA finetuning of the text-encoder (which is what I added myself this weekend, is now also in the main repo): github.com/cloneofsimo/lo

0:15

The really cool thing about LoRA is that because its an additive on top of base #stablediffusion, you can use it to interpolate in weightspace! Here I'm interpolating between a base SD 2.1 output and LoRA adjusted SD (trained on a set of imgs of myself) in checkpoint weightspace

0:13

291

A world where all crimes are judged by the same, universal and deterministic AI system sounds dystopian, but at least in #lawyerGPT world, justice would be blind and not for sale to the highest bidder?

Quote Tweet

Erik Schluntz

@ErikSchluntz

Dec 12, 2022

If you cannot afford an attorney, you will be given 10 credits for lawyerGPT.

Show this thread

Naval

@naval

Dec 13, 2022

AI will replace the people who think it will.

718

4,415

28K

The AI Dude

@NaxAlpha

Dec 12, 2022

the context length of GPT-3.5 is ~8000 tokens which is ~12KB. Sounds like the RAM of those old computers.

205

𝐈𝐦𝐚𝐠𝐢𝐧𝐞 𝐎𝐧𝐞 𝐀𝐈 | Inactive

Dec 9, 2022

and we're live 🔥

Quote Tweet

@image_in_1

Dec 9, 2022

Replying to @ai_art_today and @xsteenbrugge

twitter.com/i/spaces/1eaKb

DAILY SPACES | ai art #aiart

Dec 8, 2022

Super excited to be speaking on this Twitter space tomorrow about all things generative AI, exciting times ahead in this space of incredibly fast paced innovation. Join us tomorrow 🔥🥳

Quote Tweet

MΔRS Δi ΔRT TODAY

@ai_art_today

Dec 8, 2022

NEW AI ART INTERVIEW SERIES!

@xsteenbrugge Xander Steenbrugge TWITTER SPACES @ai_art_today SET REMINDER

THIS FRIDAY 7amCST #aiart #aiartist twitter.com/i/spaces/1eaKb

Dec 8, 2022

Was about time I finetuned #stablediffusion on some imgs of myself haha, much fun! As always, avoiding overfitting is a subtle art. For this training run I used BLIB to get a somewhat matching prompt for every training img leading to better regularization. Still room to improve!

Prob Nothing 555

@ProbNothingDAO

Dec 7, 2022

We've been, individually, fellow supporters of

@justintrimble

@braindrops_art

collections from the start. So today, we wanted to show some support to one of the most creative and mesmerzing collection ever made: DreamScape by

. 3 have been collected so far.

OneJPEGman

@one_jpeg_man

Dec 5, 2022

GM - never forget about your frens Disordered friends by

178

Had another awesome cuda-versary today! 🎉🥳

Gene Kogan

@genekogan

Nov 29, 2022

DALL-E = Windows Midjourney = Mac Stable Diffusion = Linux

445

KREA AI

@krea_ai

Nov 23, 2022

The Prompter is back! 💥 Versatile Diffusion is a game changer. 🚂 train a DreamBooth model in less than 5 minutes. 👀 big updates from Krea.

theprompter.substack.com

One annoying thing about Generative AI right now is that new models are constantly emerging, and each one requires you to "relearn how to prompt" it. Prompts that looked amazing in SD v1.5 don't in v2.0. This constant relearning feels like a struggle we can prob improve on🤔

Quote Tweet

Emad

@EMostaque

Nov 24, 2022

Some user notes on V2!

V2 prompts different and will take a while for folk to get used to. V2 is trained on two models, a generator model and a image-to-text model (CLIP). V1 used the CLIP-L14 model open sourced by @OpenAI which was great, but nobody knew what was in it 1/ twitter.com/StabilityAI/st…

Show this thread

106

Nov 22, 2022

Just some late night fun, practicing my prompt engineering skills..🧑‍🎨 Using generative AI often feels like mental therapy: imagine arbitrary fantastical scenes and then get creative with prompts to try and make them come to life. Painting with the mind / AI.. #stablediffusion

149

Nov 19, 2022

So, this is what a basic math class loos like for large language models 🧮👨‍🏫 I have exactly the same approach with

@OpenAI

's copilot as a coding assistant: it can't solve complex tasks one-shot, but if I subdivide the problem into small steps, it does remarkably well 👍

Quote Tweet

Hattie Zhou

@oh_that_hat

Nov 17, 2022

Why algorithmic reasoning? It gives OOD generalization. Why in-context learning? It allows for specialized skill acquisition without losing the generality of the underlying model. Inspired by how children are taught similar skills, we evaluate LLM in 4 learning stages

(2/11)

Show this thread

Nov 14, 2022

Prompt engineering for LLMs is an undervalued artform. Visual styling tricks for txt2img models get all the attention these days, but I bet the landscape might be even richer in LLMs since they're so open ended. Has anyone tried compiling a list of tricks like this?

Quote Tweet

jmill

@jmilldotdev

Nov 14, 2022

Potentially cool use of GPT-3 logprobs for idea generation (and look inside its

) Using this prompt engineering trick of adding some numeric signifier to your ideas like '(Team rating: x/5)', you can get GPT-3 to 'score' its generations

Show this thread

Nov 11, 2022

Soon, Google Search as we know it now will be 100% obsolete. "Neural search engines" will 𝙩𝙝𝙞𝙣𝙠 to give us direct answers rather than explicitly querying a database of indexed webpages. LLM training is equivalent to the construction of large, implicit knowledge graphs.

Quote Tweet

Metaphor

@metaphorsystems

Nov 10, 2022

metaphor.systems is now publicly available! Metaphor is a search engine based on generative AI, the same sorts of techniques behind DALL-E 2 and GPT-3 1/

Show this thread

626

Thanks

for picking up "Four seasons" 🙏 I'm currently working on an exciting new project that unlocks really interesting types of interpolation with StableDiffusion, more news coming soon 🥳 Watch

Quote Tweet

My first #stablediffusion NFTs are four seasons by @xsteenbrugge.

Oct 30, 2022

For anyone using ffmpeg on a regular basis, you need this in your life, this is 🔥🔥

Quote Tweet

zack

@zack_overflow

Oct 29, 2022

I got fed up writing FFmpeg commands. Its unintuitive, and can get ugly and complicated quickly A couple of months ago, I built this GUI to help me write complex ffmpeg filtergraphs that we have been using internally and decided to launch today. ffmpeg.guide

Show this thread

Meta AI

@MetaAI

Oct 25, 2022

Today we’re sharing some of our latest progress on AI-powered hypercompression of audio. Our researchers achieved ~10x compression rate vs. MP3 at 64 kbps with no quality loss — the first time this has been done with 48 kHz sampled stereo audio ➡️ bit.ly/3FfgPCb 1/5

0:16

88.4K views

481

1,733

Stable Diffusion

@StableDiffusion

Oct 14, 2022

Over the 7 weeks since Stable Diffusion's release, we've seen many amazing open-source contributions from the community. A lot of them have come in the form of awesome Google Colab notebooks! 🔥 Here is a thread of 14 awesome notebooks we've seen from the community ↓

793

3,802