Daydream in the Desert: Open Source Real-Time AI as a Creative Instrument

For decades, the art of the Playa has been something you had to be there to experience, vast installations, ephemeral performances, and moments of collective creativity shared only by those able to spend a week in the desert.

Playa Daydream is our attempt to democratize that experience. In 2024, BMwebcast partnered with Livepeer Studio to bring the essence of Black Rock Desert to a global audience through innovative live streaming.

This year, by weaving real-time AI into the official BMwebcast, we’re opening a door for global audiences not just to watch, but to create alongside the festival. Anyone, anywhere, could now influence and remix the visuals in real time from their phone, turning the Playa into a shared canvas that extends beyond Nevada and into the hands of a worldwide community.

This creative project came together through the work of a small team at Daydream Live 

• Ryan, who adapted and tuned the StreamDiffusion model that powered the real-time visuals.
• Gioele, who designed the frontend and API experience that let participants interact.
• Victor, who built the backend that carried the stream and inputs.
• Amrit, who architected the infrastructure that made the system stable and scalable.

Together, they turned a bold idea - making the Playa participatory for anyone, anywhere, into a functioning system running live at scale. 

"Since 2013, I’ve centered the webcast on real-time engagement and Remote Participation for those unable to travel to Black Rock Desert. The gift of Daydream extended that mission by giving Camp Envy and the global Burner community a collaborative way to modify their views of the Playa in real time. The feedback we received was overwhelmingly positive.This year, integrating AI to reimagine the webcast aligned with the theme Tomorrow, Today and with our artists’ tradition of exploring the relationship between big art, leading-edge technology, and challenging concepts. In the midst of debates about AI’s role in art and society, the webcast offered a space to test its creative potential openly. I see it not just as a new lens, but as a living canvas where Remote Participants can shape what they experience together, as a global community."
- Matthew F. Reyes (@Motorbikematt)
Producer at Burning Man webcast

Why we built Playa Daydream

“The Playa has always been about community, inclusion, and artistic self-expression. But for those unable to make it to the Black Rock Desert due to disability, age, family commitments – the experience has remained largely out of reach. While the festival’s webcast has grown increasingly sophisticated, streaming higher resolutions each year, it still left artistic participation to those physically present.

For over a decade, artists have been creating beautiful works inspired and derived from webcast screencaps - a tradition that has flourished online. Playa Daydream is designed to drive more of that, making it easier than ever for remote participants not just to watch, but to contribute creatively.

Through Daydream’s hosted StreamDiffusion platform, viewers can apply AI renderings in real time to the official BMwebcast, bringing the spirit of participation and self-expression to a global audience.”
- Gieole

The technical and product challenges

Most AI inferences today run on costly, resource-intensive server farms powered by expensive GPUs. Almost no cloud solutions can deliver them in real time. Our biggest technical hurdle was not only running a live AI stream, but doing so affordably.

We were also mindful of today’s broader AI debates around copyright and intellectual property. Many AI models are closed-source, with opaque training data and little accountability.

How we solved it

“This year’s webcast introduces open-source AI woven into the broadcast, transforming the stream into an interactive experience. The technology powering it, called StreamDiffusion, is an open-source tool developed by researchers that enables real-time creative expression using AI. I’ve been working with the Daydream team on a version of StreamDiffusion that gives artists a more granular level of control over the model, letting craftsmanship shine through in the output.”
- Ryan Fosdick 

To solve the cost challenge, the AI inference happened on cost effective consumer GPU such as RTX 4090. Thanks to the Livepeer Network’s distributed AI capabilities, Playa Daydream achieved this with an extremely low carbon footprint.

The broadcast demonstrates just one example: using a reference image to change the style of the output. That’s only a single knob.

There are many more that can be turned, allowing artists to adjust specific visual elements live, as the stream unfolds.For Playa Daydream, we chose an open-source, mobile first approach: running StreamDiffusion without additional fine-tuning, using a preset of IPAdapter images that avoid copyrighted material or specific artistic styles.

We also carefully configured ControlNets to ensure the system remained safe and respectful toward creators.

The result is a new layer of accessibility and participation: a way for the ethos of creativity and inclusion to extend beyond the desert, through open technology that is both transparent and sustainable.

Real-time AI feels alive: audio-reactive visuals, performances that merge human improvisation with machine responsiveness, or broadcasts like this BMwebcast stream where an interactive layer transforms a passive experience into a participatory one. With Daydream, we aim to make these tools accessible beyond labs or studios. Our cloud GPUs do the heavy lifting so that anyone, even without powerful hardware, can experiment in real time.

“Underlying this is a commitment to open source. As an artist, my skepticism toward AI aligns with the anti-authoritarian ethos that open source embodies. In a world drifting toward opaque, centralized AI systems, transparency and openness are essential if these technologies are to remain creative instruments instead of tools of control.”
- Ryan

The Result

• 10M minutes and 102k unique viewers for the stream

The response to Playa Daydream showed us that the project wasn’t just a technical milestone, it was a creative one. Tens of thousands of remote burners logged in and voted on their favourite real-time outputs. For many, this was the first time AI felt less like a tool hidden in the background and more like a medium for collective expression. Remote burners shared how excited they were to actively shape the webcast, and the team reflected on what it meant to see their work come alive at this scale.

From the Team

"The BMwebcast Daydream livestream is just one example of what becomes possible when AI transforms from a black box into a responsive creative instrument. StreamDiffusion is just the beginning, we have more real-time AI tools in development, and as more artists gain access to these technologies, I'm excited to see what new forms of expression emerge from this frontier where human creativity and machine capability meet in real time."
- Ryan

"Exactly one year ago, I made my first commit on what would become the foundation of our StreamDiffusion API service. To see that work now interacting with BMwebcast feels surreal. What began as an experiment in real-time AI streaming is today carrying a piece of the Playa out into the world. The Playa has always been a place I have deeply admired, even though I haven’t had the chance to be there myself. Contributing to this project gave me the opportunity to help others, like me, participate from afar. For remote burners around the globe, the livestream wasn’t just something to watch. Daydream turned it into a shared canvas, alive with collective imagination. For me, this moment represents the first time our technology has truly met the scale and spirit it was designed for. It’s an honor to have played a part in opening that door, and I can’t wait to see where our real-time AI journey takes us next."
- Victor

Playa Daydream began as an experiment in extending BMwebcast’s ethos of participation to anyone, anywhere. What emerged was a shared canvas where human creativity and real-time AI met in the moment. By choosing an open-source path, building on distributed infrastructure, and centering accessibility, the project showed how technology can deepen inclusion and democratize art. 

For the team, seeing their work woven into the cultural fabric of the Playa was both surreal and affirming. For the global community, it was a chance to experience the Playa not just as spectators, but as creators.

This is only the beginning. As more artists and builders gain access to real-time tools, we expect new forms of expression to emerge proof that when technology stays open, transparent, and community-driven, it becomes an instrument for collective imagination.