I was delighted to be invited to attend ThingsCon 2024 in Amsterdam to demonstrate our Shadowplay installation at the event which took place at the wonderful Volkshotel. Here is a little trailer of everything that happened at ThingsCon:
And, if you’re interested you can watch all the ThingsCon keynotes and various other bits of media here. They even let me speak for 2 minutes to pitch Shadowplay:
Shadowplay on the Road
This was our first go at taking Shadowplay on the road. It was a bit nerve wracking to pack everything up into a Peli case and check it into an aircraft, but, thankfully it all worked out okay. The fact we can travel with the equipment opens up lots of interesting opportunities.
AI VJing with PromptJ
The evening before I was due to set up the Shadowplay demo there was a reception which included a performance from WLDRF. While they were setting up I asked if they would mind if went to get the Shadowplay gear from my room, set it up, and tried it out as a sort of AI-powered VJ system. To my delight they agreed, so, Shadowplay had its debut as an accompaniment for live music - and it was a hit! Matt Webb was there for his keynote, and I believe this message was part of his reaction to Shadowplay in VJ mode!
Sometimes you have the privilege to be a place to witness a new reality being born
— Matt Webb 🌸🌼🌸 (@genmon.fyi) December 13, 2024 at 4:41 AM
Learning to Clarify what is Shadowplay, PromptJ and PromptTank
It has started to dawn on I and Roger Whitham that although we have been referring to everything we are working on with realtime AI visuals as Shadowplay, the naming is a bit confusing. Shadowplay is really the specific installation, which started at TEI 2024 and has evolved since then. Part of that evolution has been the building of a user interface, there have been many iterations of our prototyping, but the first proper version is called PromptJ, and it looks a bit like this:
PromptJ takes inspiration from audio or visual mixing desks. The basic idea is that you have ‘channels’ that you can blend with each other. In the case of PromptJ these channels represent different types of text prompt that are being used to effect the video signal that we feed into the system. We can configure PromptJ to run in an autopilot mode where it essentially shuffles through a big list of predefined prompts. Alternatively we can pilot it live, using predefined prompts, adding prompts, using up to 4 channels all at the same time.
We also have another user interface that is currently in development which is called PromptTank. While PromptJ tries to take inspiration from audio mixers and apply it to working with AI, PromptTank is designed from the ground up with AI in mind. More on PromptTank in the near future.
