Altis’s presentation took a futuristic turn as he demonstrated how AI transforms static View-Master slides into dynamic experiences. He integrated ChatGPT into his scanning software to auto-generate fun facts for video descriptions—like three tidbits about Robin Hood—via its API, adding an educational twist. Using Eleven Labs, he creates human-like narration tailored to each reel’s theme, even cloning voices (e.g., Bruce Lee) for an authentic touch, with plans to demo audio in the afterparty. AI upscaling enhances 50-year-old film clarity, predicting details like facial features based on deep learning, while video diffusion and neural motion animate scenes—think a fountain with rippling water or a flag waving in the breeze. Altis shared a prompt-driven example where he added ducks to a river scene, complete with ripples, and another where he controlled actors’ emotions, toggling between panic and calm in a collision scenario.
However, the tech isn’t flawless. Altis entertained viewers with “bloopers”: cars sliding chaotically in traffic, an extra girl appearing in a flask-drinking scene, or a lion chase turning one character into a mouse. He explained that AI struggles with abstract or fantastical content (like Alice in Wonderland) due to limited training data, requiring precise prompts to avoid costly misfires (each animation costs about 25 cents in credits). He’s also tackling color-shifted reels—where cyan fades to red—using Oldify and training a model with faded and unfaded pairs, showing promising early results with a Superman reel. This segment was a rollercoaster of awe and amusement, highlighting AI’s potential and pitfalls.