Welcome back to AWE Talks, our series that revisits the most engaging content from AWE’s catalog of conference sessions. Now that AWE EU 2023 has concluded, we have a fresh batch of sessions to dive into.
This week, we look at a topic that AI proponents like to speculate about: automating XR creation workflows with AI. But can it be done today? Prolific AR creators Page Piskin and Maria Nova put the theory to the test.
See the summarized takeaways below, along with the full session video. Stay tuned for more video highlights each week and check out the full library on awe.live and AWE’s YouTube Channel.
Speakers
Page Piskin, creator
Maria Nova, creator
Key Takeaways
– A popular topic in XR circles is how AI will soon be able to automate experience creation.
– We're already seeing some generative AI tools that boast text-to-3D model creation.
– But is AI at the stage where it can handle the end-to-end process of creating an AR lens?
– Piskin and Nova set to find out, by cobbling together several disparate tools.
– These included ChatGPT, Midjourney, Blockade Skybox, Photoshop, and Lens Studio.
– In the end, they were able to create a working lens, however it wasn't as seamless as hoped.
– There's no one end-to-end tool and the process requires many hacks along the way.
– Another learning for Piskin and Nova is that it's difficult to generate exactly what they wanted.
– Many of the tools that advertise generative 3D capabilities aren't ready yet, says Piskin.
– For example, too much is left to chance in the generative AI models and what they output.
– The other ironic result is that this AI-fueled process was more time-consuming.
– Altogether, AI tools can augment the workflow but not yet replace it end-to-end.
– All the above signals an opportunity for a platform to federate and streamline the process.
– Some of this will occur naturally as generative AI gets better at 3D creation, says Piskin.
For more color from Piskin and Nova, including a technical deep dive on their process, see the full video below...