On a wet afternoon final November at Seattle Middle, I walked right into a darkened studio and located a movie crew huddled round a miniature set, manipulating motion figures representing software program builders, a hero-protagonist ghost, and a monster representing AI slop.
Now that is completely different, I bear in mind considering to myself.
It was a stop-motion animation shoot for Kiro, an AI-powered software program growth instrument from Amazon Internet Companies. Once I discovered that this crew was utilizing AI to create the video in ways in which wouldn’t have been attainable prior to now — or would have taken much more time utilizing conventional methods alone — I knew I needed to sit down and speak with them.
This week on the GeekWire Podcast, I’m joined by Brice Budke (president) and Zeek Earl (government inventive director), who run two Seattle studios: Shep, a video company that works with tech corporations like Amazon and Microsoft, and Packrat, a inventive studio that makes a speciality of miniature worlds, handmade units, and retro inventive tasks.
Right here’s one of many movies that Packrat constituted of the AWS Kiro shoot, utilizing AI to assist fill within the gaps between begin and finish frames that they had been capturing once I visited:
You may also know Packrat’s work from the epic and broadly watched 2025 Seahawks schedule launch video, which gained a Gold Clio. In addition they made “Prospect,” an indie sci-fi movie that premiered at SXSW in 2018, starring Pedro Pascal and Sophie Thatcher.
Brice and Zeek focus on how AI is reworking their work, together with what’s working, what’s not, and the way they’re navigating the modifications on each the inventive and enterprise sides of their studios.
Pay attention under, and hold studying for key takeaways. Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you hear.
AI doesn’t essentially lower your expenses. Getting usable output from AI video instruments continues to be a numbers recreation. “To get one usable shot in a last piece, you may want 1,000 generations,” Zeek says, referring to the method of prompting the mannequin to create a brand new variation. The time and experience required to wrangle these instruments means there’s not all the time a transparent value financial savings over conventional manufacturing.
Style and instinct matter greater than ever. As AI handles extra of the analytical and task-oriented work, the human worth shifts to concepts, perspective, and judgment. “Finally someone has to have a perspective and style and a perspective to place in,” Brice says. “If you happen to take that out, you’re simply up for the slot machine, random likelihood generator.”
Model requirements create distinctive challenges for AI. In contrast to the random AI-generated content material flooding the online, model work calls for consistency. Kiro has a definite character, and Zeek couldn’t have AI generate approximations that felt completely different in each body.
The Kiro mission was a breakthrough partially as a result of it mixed handmade manufacturing design with AI-assisted animation, giving him inventive management somewhat than forcing him to work across the limitations. “Doing the Kiro mission was the primary time I felt accountable for AI,” he mentioned.
AI defaults to the generic. Generative AI “is all the time pulling you in the direction of the center,” Zeek explains. “It’s all the time pulling you in the direction of issues which are by-product.” The distinctiveness of handmade work comes from the buildup of human choices, one thing AI can’t replicate.
The sheer tempo of change is a large problem. The AI video panorama is transferring so quick that workflows grow to be outdated virtually instantly, and it’s arduous to maintain up with the capabilities of typically accessible instruments. “It’s chaos,” Brice says. Each time a brand new mission comes up, they need to reassess which instruments is likely to be proper for the job, given the evolution.
Working with infinite instruments is psychologically brutal. Generative AI affords infinite potentialities, which implies you may by no means be totally glad. “As a inventive individual, while you’re working with an infinite instrument, there’s all the time the chance that one thing higher might come,” Zeek says. “On the finish of a day of doing AI, you may simply really feel bodily horrible as a result of you should have by no means really reached a stage of satisfaction along with your work.”
Each inventive goes via the levels of grief with AI. Zeek says he’s moved previous feeling threatened or unhappy in regards to the modifications. “I’m actually attempting to concentrate on what it may possibly do to carry out visions that had been beforehand unattainable,” he says. His hope and dream: that decrease manufacturing prices will allow extra bizarre, fascinating human creativity, not much less.
The longer term belongs to concepts, not busy work. Brice sees a shift in human focus towards producing concepts, somewhat than merely producing output. If AI can deal with extra of the analytical and task-oriented work, he says, it frees up area to concentrate on concepts, benefiting from your individual instinct and style, areas the place people nonetheless have the sting.
Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you hear.
Audio enhancing and manufacturing by Curt Milton.
