harry7557558/spirula-studio

What it solves

Spirula Studio simplifies the complex process of creating 3D Gaussian Splatting models. It removes the need for fragmented toolchains—such as separate Python, PyTorch, or COLMAP installations—by providing a self-contained binary that handles everything from raw photo/video input to the final textured mesh.

How it works

The software uses a Vulkan-based compute backend (with a legacy CUDA option) to ensure compatibility across NVIDIA, AMD, Intel, and Apple GPUs. It implements a combined strategy based on MCMC, IGS+, and MRNF to produce sharper results with fewer artifacts. To maintain high efficiency, it uses quantized training, allowing it to handle up to 10 million Gaussians within 8GB of VRAM. It also integrates Structure-from-Motion (SfM) for camera positioning, AI masking (via SAM), and specialized support for 360° and fisheye cameras.

Who it’s for

It is designed for users who want to create high-quality 3D reconstructions from images or videos without dealing with complex environment setups or dependency hell, as well as developers working with 3D assets on various hardware platforms.

Highlights

• Cross-Vendor GPU Support: Works on NVIDIA, AMD, Intel, and Apple Silicon via Vulkan.
• All-in-One Pipeline: Built-in SfM, frame extraction, and AI masking eliminate the need for external scripts.
• VRAM Efficiency: Capable of training 10M Gaussians in 8GB VRAM through quantization.
• Specialized Camera Support: Native handling of equirectangular and 360° imagery without needing undistortion.
• Advanced Post-Processing: Includes bilateral grid and PPISP for exposure and white balance correction.