Episode 5 went live today, I’d love to hear your reaction. Next weeks episode is a special release giving viewers a new perspective on “ODAR” the secret government agency intercepting and translating the interstellar transmissions. ✌️ https://youtu.be/DdvoCv_76T0

I created the entire action sequence first as a 3D previs using a Clay Render — focusing purely on choreography, camera movement, timing, composition, and scene continuity. No textures. No polished lighting. No distractions. Just the core mechanics of the scene. For action sequences, I find this approach especially useful: if the scene already feels cinematic in a simple clay render, the final visuals only make it stronger. Previs → Clay Render → Final Visual Sedance 2.0.

Sedance 2.5 1080p I tried to create one proper cinematic shot using AI — and in the end, nothing really worked out 😅 I went through tons of generations, different settings, references, and approaches. I even used a Clay Render to give the AI a more accurate understanding of the scene geometry, composition, and camera movement. In the end, I’ve already spent around $200 experimenting, and I still haven’t managed to get the shot I actually wanted. I used the prompt from the attached TXT file together with 4 key visual frames as references for the generation. I ran a lot of tests and iterations, trying different approaches, settings, and workflows. After all that experimentation, I still consider the final result very disappointing and far below my expectations, especially considering the amount of time and money spent.

I created multiple AI-generated tsunami VFX shots to test the limits of AI video generation.

What happens when a human transforms into a wild animal or a terrifying creature? In this cinematic AI VFX experiment, I explore a series of photorealistic human-to-animal and human-to-monster transformations, created using generative AI, motion design and visual-effects techniques.