AI in VFX
summary
Wētā FX describes concrete uses: a bolted-on live depth-map rig streams depth for on-set virtual characters and proves useful in directors’ feedback, a custom 3D facial pipeline and new strain system plus machine-assisted transfer improve animation fidelity, and a recurrent residual denoiser used on Avatar 2 frames pulls more temporal info to cut render cost. Rights get enforced by putting weights and training data into the asset system, requiring per-show models and some bespoke components; data authenticity remains an unsolved problem.
AI-written summary · about
chapters (10)
- 0:00Introduction to AI in VFX
- 1:49Rights management challenges
- 5:21Creative control with AI
- 8:08Empowering artists with tools
- 11:00Live depth comp in virtual production
- 12:263D facial modeling improvements
- 13:30Denoising for faster renders
- 14:58Compute and iteration trade-offs
- 16:02Data authenticity concerns
- 17:27Keeping the fun in VFX
related talks
36mAI in Typography: The Impact on Design and Font Development
27mAI for social good
32mAI Video: The Future of Imagination and Production
33mAI Images: Redefining Quality with Innovation
19mAI video is Already App-ening
30mAI in the Cloud
38mAI in action: How to use practically in iGaming beyond the hype?
27mAI Pipeline for 3D Content: From Idea to 3D Generation