Autodesk enables custom AI motion training and editable Gaussian Splats

Autodesk is shifting generative AI from a black-box novelty into a functional pipeline tool by emphasizing data sovereignty. By allowing studios to train models on their own animation styles, Autodesk addresses industry concerns regarding IP rights and creative homogenization. This move, combined with editable Gaussian Splats, signals a move toward hybrid production where captured reality and AI-generated foundations are fully integrated into standard DCC tools rather than remaining as isolated experiments. For the broader ecosystem, this lowers the barrier for mid-sized studios to execute high-volume animation. Watch for the adoption rate of 3DGS in traditional TV pipelines as a benchmark for real-time environment integration.

The push toward data-specific AI training comes as the generative AI market in animation is projected to reach approximately $3.23 billion by late 2026, growing at a CAGR of 36.1% per Research and Markets (July 2026). This growth is largely driven by a demand for hyper-realistic and stylized content across streaming platforms that the traditional manual pipeline cannot meet at current production speeds. Autodesk’s focus on 'Bring Your Own Data' directly mirrors a broader industry pivot toward data privacy and sovereignty, where studios seek to use automation without exposing their internal IP to external foundational models.

Technologically, the inclusion of 3D Gaussian Splats (3DGS) in 3ds Max and Arnold represents the format's move into the 'boring middle' of production—a sign of technical maturity. Per Digital Production (June 2026), the release of Arnold 7.5.2 was a critical milestone, as it added support for the OpenUSD 26.03 Gaussian Splat schema, allowing these captures to be treated as standard point-based data within Universal Scene Description (USD) pipelines. This standardized approach allows environments generated by emerging AI platforms like World Labs to be imported, relit, and rendered alongside traditional 3D assets.

Competition in this space is intensifying as other major software providers integrate similar capabilities. Adobe recently previewed AI features focused on automated masking and cleanup at Adobe Max, while specialized startups like DeepMotion have launched browser-based services that transform text prompts into 3D animations, per Envato (January 2026). Autodesk’s strategy at SIGGRAPH 2026 suggests it intends to defend its position as the industry standard by keeping these advanced AI and capture workflows within its existing, deeply integrated ecosystem rather than letting them migrate to standalone tools.