- Type
- Generative AI
- Year
- 2026
- Tools
- ComfyUI, Houdini, Apple SHARP, Flux, LTX Video
The project
A fictional fashion show presenting six models and their outfits, inspired by a designer's creations. The goal: smooth videos, with perfect consistency between the clothes and the environment.
Hybrid AI and 3D pipeline
01Gaussian Splatting and SHARP
Creating 3D volumes from the designs with SHARP (Apple), for a real spatial base.
Creating 3D volumes from the designs with SHARP (Apple), for a real spatial base.
02Houdini
Importing the splats to animate cameras and export precise depth passes.
Importing the splats to animate cameras and export precise depth passes.
03First frame
Creating the high-definition reference image with ComfyUI and Flux, to set the style and materials.
Creating the high-definition reference image with ComfyUI and Flux, to set the style and materials.
04LTX Video
Generating the final video from the depth pass and the first frame, for a stable structure and full control of the environment.
Generating the final video from the depth pass and the first frame, for a stable structure and full control of the environment.
Tools and skills
- ComfyUI
- Houdini
- Apple SHARP
- Flux
- LTX Video
Key points
01
Spatial control
3D is used to direct the AI.
02
Stylistic fidelity
Faithful to the designer's textures and cuts.
03
Innovation
Use of recent rendering technologies, including Gaussian Splatting.
