STABLE DIGEST #12
Artist spotlight - Tom Furse - SDXL 1.0 and more!
THIS WEEK’S ISSUE
Brace yourselves for the most impressive Stable Digest yet!
SDXL 1.0 is here and we can't wait for you to discover all its features! But that’s not all, because we have other toys for you to play with, including our new Platform reboot, SwarmUI, and more!
In this week's Spotlight, we're featuring none other than the maestro, Tom Furse. Join us as we delve into his musical journey and his blossoming love affair with AI.
In our Tools & Tuts section, we're offering insights into ComfyUI alongside a nifty tool that's set to add a dash of harmony to your life.
And of course, our Digest wouldn't be complete without showcasing the stunning creations of our immensely talented community. Dive in and get inspired!
Have suggestions for artists spotlights or community features? We wanna hear from YOU. Email us community@stability.ai
FOR THE COMMUNITY
SDXL 1.0 OPEN-SOURCE RELEASE!
Pop the champagne and turn on the party music because it’s here folks! Thanks to the hard work of everyone at Stability and your invaluable feedback and votes on Discord, we now have the crowning achievement of SDXL 1.0!
This open-source model delivers impeccable image generation with simpler prompts and more accurate understanding, while keeping the process swift and easy, even for those with consumer GPUs.
And the cherry on top? SDXL 1.0 offers fine-tuning capabilities that are as user-friendly as they are powerful. Custom LoRAs, checkpoints, task-specific controls - you name it, SDXL can do it. We're also cooking up some exciting advanced features for image control we which we touched on with Emad and the Applied ML team live on Discord. A record breaking release day stage! (VOD on our YouTube)
Want all the nitty gritty details? Check out our blog and press release!
Play with it now on ClipDrop, Dreamstudio, Discord, Github, HuggingFace, API, and Amazon Sagemaker!
SwarmUI
Ready for an AI power-up? Say hello to StableSwarmUI - the perfect blend of ComfyUI's power, Auto WebUI's ease, and StableStudio's quality.
StableSwarmUI is your new go-to interface for Stable Diffusion. A high-speed, information-dense UI that doesn’t overwhelm you, it utilizes the might of multiple GPUs, whether they're at your home, on rented server space, or scattered across your LAN.
Fancy creating your own workflow? Toggle to ComfyUI mode and go nuts!
Peek under the hood and explore the alpha version packed with interesting features, with more in the pipeline.
Platform Reboot
Huzzah! The shiny new Stability AI platform site is here!
Discover APIs in our Sandbox with ease, delight in crystal clear documentation, and manage your API keys without breaking a sweat!
No more hopping to DreamStudio for API keys management - it's all under one roof now.
So, why wait? Start your seamless journey today!
BY THE COMMUNITY
Ameca draws with SD!
OK, this one was just too fun not to share. If you haven't caught the viral video of Ameca, the Engineered Arts’ AI robot crafting artistic doodles, now's the time!
Powered by StableDiffusion, Ameca's sketches are delightful creations. And yes, every piece ends with an Ameca signature! Don't miss this riveting blend of tech and art!
TOOLS & TUTS
Tutorial Spotlight!
If the complexities of AI art tools have you breaking a sweat, then take a breather! This ComfyUI - Getting Started tutorial is the perfect antidote.
Join Scott Detweiler as he guides you through the intricate webs of ComfyUI, a node-based platform that breathes life into advanced StableDiffusion pipelines for unparalleled AI art. No matter your experience level this tutorial series will empower you to unlock the immense potential of ComfyUI.
So, kick your fears to the curb and dive into the captivating world of AI art with ComfyUI today!
COOL TOOLS!
SUNO
If you're a fan of jaw-dropping innovations and rhythmic vibes, then you're in for a treat! Remember the splash Bark, the text-to-audio transformer model by Suno, made last month?
Well, this clever bit of tech now has a user-friendly version available on Discord. The show-stopper? The sing feature! Simply type in your favorite lyrics, and voila, you've got yourself a tailor-made tune.
Video courtesy of SunoAI website
Give it a whirl and let your voice be heard!
WITH THE COMMUNITY
STABLE SOCIETY DEEP DIVE - Tom Furse
Stay curious
Hey there! Thanks for joining us, we're thrilled to explore the world of Tom Furse! Before we dive into AI, we can't ignore your remarkable journey in the music industry. Could you share some highlights, experiences, and influences that shaped your path over the past 15 years? We're eager to hear about your musical background!
Thank you for having me! So all the way back in 2005 I started a band called The Horrors with some friends and we were very lucky in that it resonated with the moment and just took off. Within a year I think we were on magazine covers and doing all the things young bands do when they encounter success. Over the next 15 years we released 5 albums, toured extensively the world over and even got to experience the music industry for a hot second in the pre-streaming era. For most of that time I was also doing all kinds of other work in and around music - releasing my own records, remixing, mixing, film scoring, producing, mastering, etc so I’ve had this quite full spectrum approach to exploring all the different ways one can explore music and sound for a living.
In a previous interview, you expressed how AI provided a breath of fresh air by opening up an entirely new medium for artistic exploration. We're curious to know more about your transition into the world of AI visuals. How did it come about and was it a seamless process, or did you face any challenges along the way?
Honestly it was very seamless, very similar to how I felt when I first used an 808 drum machine - “wow ok, this is something I can really sink my teeth into''. I’d seen early examples of text2image online but after hearing an interview with Adverb on the Interdependence podcast about VQGAN+CLIP I just thought ‘I have to try this’. So one morning I woke up and decided that’s what I was going to do that day. Not in a serious way, I just thought it would be a fun thing to try. When I first pulled up VQGAN+CLIP, the whole Colab interface was quite intimidating but I got my head round it fairly quickly and that was that - I was completely obsessed, glued, drunk on the wonder of it all. That was a couple of years ago now and it’s still something I do almost everyday in some capacity.
As we scroll through your Instagram reel, we can't help but be captivated by the surreal and edgy visual aesthetics that are showcased there. Could you share what inspires and influences your draw to these themes in your exploratory work?
I’ve been asking myself the same questions - I suppose when confronted with the near infinite possibilities of large image models it feels too claustrophobic for me to just explore one avenue, so I make notes about things that catch my eye or attention. They might be artists or styles or materials or sometimes just noticing something as subtle as the way a street lamp lights a house at night. Sometimes I’m just trying to find a way to visually communicate an idea or a feeling, and sometimes the work is autobiographical. Inspiration and influence is everywhere - it’s constant - and one of the jobs of artists is to act as the conduit or the receiver for all that information. The world around us right now is surreal, it barely makes sense, and we know it. It’s a deeply unnerving moment in time. So I can either make nice pretty things that feel safe and comforting or I can make something that feels more like how I actually feel - I suppose I do both? I suppose we need both in some ways?
Reel - Tom Furse
One thing I’m almost certainly obsessed with is the pursuit of beauty, and pushing at that edge where beauty is unnerving. I love the fact that Stable Diffusion distorts bodies and faces, that it hallucinates its own anatomy and in doing so offers new and unexpected forms, some of which are quite intoxicating.
Your music blends live instruments and synths, showcasing your expertise in engineering and recording. Have you found any similarities between your technical skills in music and your use of AI art tools? If so, how has this overlap influenced your choice of tools and creative process in AI? Any surprising tips or tricks you've discovered along the way?
Technically DAW music production and image synthesis are very different, but I think if I transferred anything over from my time in music I think it’s probably just the lessons of being an artist, of having a practice, of being consistent in doing the work, knowing that it’s ok to make mistakes and explore, that it’s ok to try on different hats. Maybe one of the biggest lessons is that the creative energy that drives us doesn’t have to just find one outlet, but can be a multi-faceted thing, ready to take you to wherever you’re ready to tread.
You've directed and produced AI music videos and visuals for various artists, showcasing a diverse range of AI techniques. When making projects like this, how do you approach the initial stages of visualization? Could you take us through your process of building concepts and establishing themes that align with the essence of the music?
I’ve been very lucky in that most of the time people just come to me and say ‘we love what you do, will you make us something?’ so from there I just experiment with the first idea I have and go from there. It’s a very organic process, often with some false turns, but I just try to be reactive to the music that’s put in front of me. When I have to plan and present a more solid pitch, I tend to make a storyboard or mood board with images and test animations to give the artist some idea of what I’m thinking. The unpredictable nature of the tech as it stands presents its own problems I think as people have got so used to very very fine levels of control over media. For well over a decade I’ve been using generative music techniques in my work so I’m quite comfortable with the idea of giving up some level of control but this is generally a new idea to people and I think can be a bit of a shock - so I try and make this clear from the very beginning of the project.
On a practical level though I think if you’re familiar with music production you’re probably not going to have a hard time transitioning to video editing and vice versa - it’s mostly just moving blocks of stuff around on a screen.
You have some lovely collections in the NFT Marketplace, from your nostalgic inducing VQGAN animations on FND to your “A Place To Become Lost In” architectural studies. What led you to venture into the world of NFTs, and were you just passing through or do you have any plans for future projects in that space?
I’m very into NFTs as a technology and as a way of digital artists being able to make a more direct living from their work - I’m also very appreciative of it as a way of creating a record of one's work in a decentralised way. So when I was first rediscovering this love making visual art again it was great to be able to share that with an audience and earn enough money to pay my bills through that. I had a few quietly successful collections and am hugely appreciative to everyone that supported that work.
There was a series I made called ‘Magic Happens Here’ which was made just pre-Stable Diffusion - it was a combination of lots of different iterative techniques, some ML, some very traditional, to create these wild and incredibly detailed landscapes that I populated with characters and symbology. I’m very proud of that series - it’s got quite a unique feel to it, the combination of the complex workflow and the heavy use of Huemin’s JAX Diffusion definitely yielded some interesting results.
After some time in the NFT space I did have a small self-reckoning - when confronted with a technology that allows you to create infinite imagery and a platform to sell said imagery, there is a temptation to lean into that - creating a whole series in an evening is a completely feasible task. But then at that point I started to ask myself what were my motivations here? I don’t just want AI to become a tool of mass production, of churning out work to meet a wobbly market demand - it’s much more than that to me. So for the moment I wanted to take the time to explore, to search - to really DIG IN and figure out what exactly it was that I wanted to see and what I wanted to say. Giving your subconscious time to process and let things float to the surface is very important I think. We have a tendency to monetise everything very quickly these days, it’s good to not give in to that urge all the time.
Ok, let's dive into the world of Wedgwood, Charles Jefferey AKA Loverboy, and your involvement in the project. Can you share your experience with us, from how it all came about to the concept behind it, your role in the music video, and your creative artwork for them? We're eager to hear the story behind this collaboration!
Well I work a lot with Charles Jeffrey doing the music for his collections and projects - Wedgwood happened to be one of them. I made the music for their collaboration but also through getting to know the Wedgwood team I was invited to reinterpret the iconic ‘Borghese Vase’ design - which I did in a AV piece called ‘EKSTASIS’ - a complex piece involved 55 separate generated elements which were hand coloured and then assembled into an exploded version of the vase form, layered over ritual scenes made with text2video. That piece was shown at the V&A museum in London which is probably the first time my work has been shown in a major arts institution - all because I decided to try out some funny bit of open source code one day.
EKSTASIS and process video
Last year, in an interview with AMI, you mentioned taking a break from music to focus on your visual projects, but expressed excitement about the potential of a new wave of music technology upon your return. How does the current world of AI music technology compare to a year ago? Have you started to delve back into music creation, and if so, what new tools are you exploring?
Well, AI in music hasn’t had the same explosion as visual media, and the reasons behind that I think are mostly legal but I try and stay up to date with what Harmonai is up to, I’ve experimented with training and using RAVE & Musika! models and find that whole strand of research very exciting. There’s going to be a whole new tool set and that almost always means new genres, new styles, new sounds, etc. I’ve felt for a little while that music needed a new paradigm to play with.
As your artistic journey continues to evolve, we're excited to learn more about your future plans.What are your upcoming projects or goals as an artist, and how do you see the evolving landscape of AI playing a role in your future creative endeavors?
Well I can’t say I have a concrete plan, I just try and take every day as it comes. At this point machine learning techniques are a big part of my creative practice so I can’t see that changing anytime soon. Currently I’m attempting to make some kind of film for the R.AV Club / Refraction label - the emergence of text2video is another mountain to climb and explore and doing that successfully I think will take some time and development to get right. But it’s a real joy being able to compose music and create the visuals together almost as one thing. I’m also enjoying collaborating with artists I know like Scarlett Carlos Clarke by training models on their work via fine-tuning - which is EXTREMELY powerful when done right. In general I think I’ll continue to hone these skills for as long as it continues to be an enjoyable part of my life. I also want to start painting more seriously too, and finding ways of realizing my work physically, of bringing it into the world.
Thank you for taking the time to speak with us! Would you like to share any additional thoughts with us and our community of readers, or give shoutouts to anyone who has been an inspiration to you along the way?
My pleasure! I think my general advice or thoughts would be ‘stay curious’ - you never know what’s around the corner. Shout out to the Huemin JAX & Deforum crew who continue to be very supportive and valuable contributors to the space. Also s/o Holly Herndon and Mat Dryhurst for their own contributions through their own work and giving a platform to many inspiring conversations via their Interdependence podcast.
“Stay curious” - Tom Furse
WINNERS OF THE WEEK
And of course it wouldn’t be a Stable Digest without a shoutout to our COW winners!
Emerging from the field, katmoget sends chills down our spines with their Unveiled Masks and cryonicus is capturing our imagination with what could be Stills straight out of an sci-fi blockbuster!
Head on over to the Community Events on the Discord and earn your chance to be highlighted!