Gather round my #aiArtists , and let me show you something:
This is called #stableDiffusion webui
Not to be confused with #dreamStudio, this runs on your machine/python and is even accessible remotely via ipad/iphone!
Features in 🧵 (you don't want to miss this)
It's Jam-packed with features, here are the mains ones:
✨ Use from your phone/ipad
✨ Built in upscaling with ESRGan / CFGGan (quality & face) so you get high-er quality results
✨ Comes with k_lms, k_euler and k_euler_a sampling methods, for low step <> high cohesion results
✨ Create a prompt matrix
Testing several combinations of prompts and artists and styles, for example:
cartoon dog | by greg rutkowski | unity 3d | still frame from rick and morty
Will give you this matrix, and each one of the results individually.
✨ img2img is supported, and there's a loopback, so you can generate 10 images, choose the one you like, use it as init_image with different keywords, and keep going.
There's also basic support for inpainting UI, like in DALLE, but #stablediffusion doesn't seem to do that yet
You thought I was done? Hell no!
There's already support for textual inversion, so you can upload 3 images (say of your cat, or your art) , store them as a "X" and then, you can just use the X in your prompt. Advanced crazy stuff.
All of this, in a collaborative effort, over at Github, discords and reddits, with folks one upping each other, many are using @Gradio and honestly pushing it to the limits (@Gradio can I please have css everywhere??)
IN LESS THAN A WEEK!
How to install and run all this you ask?
This is one of the repos' the UI you see above is my addition to it (hopefully will merge soon)
I don't know if HLKY is on twitter, but props to him on this insane sprint and feature set. github.com/hlky/stable-di…
Enjoyed this? Go checkout the github, or subscribe to my awesome list of #imagesynthesis and #aiart
(this UI link is already there btw, amongst other amazing things) github.com/altryne/awesom…
And now this is accessible through Colab! No need to download or install or to have a GPU!
For all my mac friends, for all my non GPU friends, I give you... the colab version! That's right!
If you have a google account, and can follow VERY BASIC instructions, you can now run this in a few minutes! colab.research.google.com/github/altryne…
Repo: github.com/altryne/sd-web…
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Trying to learn how to animate with #stablediffusion and the @deforum_art Colab notebook by @pharmapsychotic
It's insane how fast things are moving since the release of #SD as open source!
Want to learn to animate? Here's my step by step so far! 🧵
This is using a open source model that related this Monday
A colab that uses img2img to feed the next frame with the previous frame and filling in the blanks and pan/zooming.
And a prompt search engine , with 200 million images and prompts and seeds.
ALL FROM THIS F-ING WEEK
Use #lexica by @sharifshameem as inspiration, find images that have a repeating quality to them, this is going to be important later.
This for example is an amazing prompt to animate (pan from left to right)
There are many other resources like @ai__pub that's worth following (in fact I've added them to my list!)
Some notable examples and the main shoutout is to @pharmapsychotic , maybe the nicest ML engineer around, extremely helpful, has a HUGE tool bad, and his own tools.html list was an inspiration for this 👏
His tools list is awesome! Add to bookmarks! pharmapsychotic.com/tools.html
[THREAD] This feels like a level up on my #homeautomation journey
My @home_assistant custom built ui, on top of a 45$ Fire HD 8 that I got for prime day, mounted with magnets.
Can control and display above 500 entities, including sensors, alarms, appliances, lights ... 1/
... and can even dispatch the Roomba to clean the mess our kids made.
I find @home_assistant to be one of the more incredible open source project. It has connections to anyting with a mac-address pretty much. It can be used as a bridge between apple/google/amazon ecosystems.
I for instance can control smart home devices that don't have #homekit support with my iphone/watch/appleTV and call the same scenes (Good Night house) from Alexa/Siri/Iphone/Watch.
Confession time: I am trolling trump ads on facebook in comments, every time I see that I'm one of the first folks to see the ad. [thread]
Yes, this is not likely to change the outcome of the election, but here's my thinking.
Facebooks mostly older, and right inclined population, don't really understand the difference between a facebook ad and just a post by a group or a friend.
If someone reads a trump ad like a post, then reads the comments. If I'm the first comment, and my trolling triggers their cognitive dissonance a little bit, what bad could it do? Worst case it'll glitch and the alleged supported would actually decide to... change their view?
All my followers with a mac, PMs, Devs, and especially designers, anyone who takes screenshots, this will change your workflow.
The new @CleanShot_app X is amazing replacement to your current screenshot flow + gifs and videos.
It completely plugs in to your existing flow, including taking over your muscle memory shortcuts like cmd+shift+4 and holding CTRL to put the screenshot into clipboard.
It adds so much on top of it, that I think it's completely worth the price, let's explore (I'm going to use it to highlight some of it's features here)
The whole world is learning the meaning of staying home for prolonged periods of time. For a lot of them, it will give them time to familiarize with new technology they haven't needed before. Prediction thread.
I expect the rise of VR (which is sold out everywhere), Streaming services, Food Delivery start ups (especially the autonomous ones).
Meditation apps will start selling more for sure, as most likely apps in general.
Technology stocks will quickly re-bounce, as the whole world will come to rely *even* more on the likes of AWS, Google Cloud and other big tech. Social media will be even more prevalent.