Stability-AI/stable-virtual-camera

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

1652 stars · 125 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Stable Virtual Camera (SEVA) is a generalist diffusion model for novel view synthesis that generates 3D-consistent views of a scene from any number of input images and target camera trajectories. It is released as a Python codebase with pretrained 1.3B-parameter checkpoints and a Gradio demo for image-to-video generation with camera control.

Use cases

• generate novel views of a scene from a single image
• create orbit or spiral camera trajectory videos from photos
• synthesize 3D-consistent video from image inputs with diffusion models
• interpolate between two views of a scene
• explore 3D scenes without explicit 3D reconstruction
• run novel view synthesis research benchmarks

When to choose

• you need 3D-consistent novel views or camera-trajectory videos from one or few images
• you want a pretrained generalist NVS diffusion model without building explicit 3D representations
• you need keyframe-based camera trajectory control for creative video generation

When to avoid

• you need real-time rendering or guaranteed geometric accuracy from true 3D reconstruction
• you lack a GPU or cannot accept a non-commercial-style custom license (NOASSERTION)
• your task is standard text-to-video generation without camera control

Facets

library · maturity active

image-processing video-processing machine-learning deep-learning computer-vision artificial-intelligence graphics deep-learning python diffusion-model novel-view-synthesis image-to-video 3d-consistency camera-trajectory generative-ai linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="Stability-AI/stable-virtual-camera")