AI-Supported Previs
This is EnvisualAIzer — an Unreal Engine Prototype leveraging AI-generated environments and scene reconstruction for Virtual Production workflows.
Here are a few of the prototype's features.
Character Preset Placement
Easy drag and drop placement of Mixamo preset characters via gizmos.
- Mixamo Characters
- Gizmos
- Drag and Drop
Previs Camera
Adjustableprevis camera for shot framing and previs saving.
- Previs Camera
- Adjustable Parameters
Animatic Generation
Image-to-Video animatic generation on saved previs scenes via Luma Dreammchine API.
- Luma Dreammachine
- Animatic Generation
- Image-to-Video
- API Integration
About EnvisualAIzer.
EnvisualAIzer is a research prototype started as a Master's thesis project at the Ludwig-Maximilian-University of Munich in collaboration with the University of Television and Film Munich . The project aims to explore the potential of AI in the field of Virtual Production. Designed with a focus on democratizing advanced previsualization techniques, EnvisualAIzer leverages a novel 3D approach that aligns closely with traditional previs practices—offering filmmakers enhanced control and consistency compared to common image and video generation methods. By integrating innovative techniques such as Depth Mesh, 360° Panoramic Meshes, and Gaussian Splatting, the platform creates a more tangible and navigable creative space. Looking ahead, the project intends to expand its capabilities by integrating a comprehensive suite of new features to build a holistic AI-driven pipeline.
EnvisualAIzer's
tech stack.
Unreal Engine
Blockade Labs
Luma Scans
Stable Diffusion
ControlNET Models
Dreammachine API
Coming Soon.
In-Prototype Environment Generation
Generation of new environments directly in the application via Blockade Labs Skybox Mesh API.
- Blockade Labs Skybox
- Environment Generation
- API Integration
In-Prototype 3D Asset Generation
Generation of 3D assets and characters directly in the application via Hyper3D's Rodin API Integration.
- Javascript
- NPM
Viewport Real-Time Style Transfer
Integration of StreamDiffusion Real-Time Syle Transfer on the viewport render.
- StreamDiffusion
- Real-Time Style Transfer
Awards.
"The thesis stands out for its innovative combination of theoretical exploration and practical application that integrates both measurable performance and industry expert feedback."
Jury Comment - AI-HUB@LMU Prize Commitee for the Best Master’s Thesis Project in the Field of AI
Interested in the thesis?
Request to Read Thesis