Lumen Pro II. Hands-Free. Finally Free.
Gen AI 2023 Demos
The following examples of research and testing early methods to integrate Generative AI into the video production workflow.
Gen AI 2026 Demo
I created this tech reveal promo as an example for the latest gen AI video workflow using Nano Banana Pro and Veo.
Lumen Pro II, a fictional high-end smartglasses that aim to replace your smartphone and finally work hands-free on the go.
Google Flow, Veo 3.1 + Nano Banana Pro
I designed the device using Nano Banana Pro. After a few detailed prompts and refinements, a reference for the gadget as envisioned was produced. A lightweight, stylish, frameless design. With an AI processing unit on the side and a photon charging pad. Depth sensors and cameras are neatly positioned above the nose area. I used Veo 3.1 to generate footage for the device and the actor using it, with a special focus on persistence, relying on prompt iterations, guidance and corrections. The video starts with two precursor prototype devices, then switches to the final device reveal. Our actor and final device are consistent between all the shots. Editing and animating the text supers was completed in After Effects.
Finishing touches were added using After Effects. Audio was generated using Veo 3.1.
Restyling with an AI temporal script. This clip mixes drone footage, an animated transition, and my website logo to create the showreel's outro logo. Generated using Stable Diffusion on local GPU.
Seeking stability versus embracing flicker in AI video generation.
In this clip, I explore the latest text input, temporal, and restyling techniques in Stable Diffusion and deal with the early challenges of Temporal stability in AI video generation. Generated using Stable Diffusion on local GPU.
Benefits of Stable Diffusion versus Adobe Generative AI.
Baalbek's legendary quarry stones were meant for the world's most colossal temple podium. In this example, I used Stable Diffusion to reconstruct the ancient temple on its massive podium and visualize the Temple of Jupiter in its past glory using AI. This method offers benefits over Adobe's generative fill, precise image control, high resolution, and flexible workflow. With the original six pillars, quick Photoshop cloning, and iterative prompting, I guided Stable Diffusion to reconstruct the temple, avoiding the generative model's typical hallucination challenges when dealing with famous ancient locations. After the prompts, references, and variables are set up, you can generate endless lighting and shading variations.
The temple, podium, and scenery were generated using Stable Diffusion on the local GPU.