Stable Video Diffusion

by Stability AI

🟢Active

Stable Video Diffusion is Stability AI’s open generative video model family for creating short image-conditioned clips. The initial SVD and SVD-XT research releases generate 14 or 25 frames from a source image.

VideoFounded by Emad MostaqueVisit website →
🔍 Last checked: 9/3/2026 · Confidence: 82%

Why Abandoned

The official product and model pages remain available, and Stability AI has not announced a shutdown. The original research release is still distributed, though newer Stable Video models have followed it.

💵 Pricing & access

Pricing model
Not documented
Starting at
Not documented
Official site
Visit Stable Video Diffusion

🩺 Health Signals

1d ago
HTTP Status
200
Response
SSL
Confidence
82%
Domain expires
Last content change
Last tweet
API status

📅 Timeline

2023-11-21

Stable Video Diffusion released

Stability AI announced the research preview of Stable Video Diffusion, including SVD and SVD-XT image-to-video models capable of generating 14 and 25 frames, respectively.

2023-11-21

Model weights published

Stability AI released model resources through Hugging Face under a community license, enabling local research and experimentation subject to license terms.

2024-01-17

Stable Video Diffusion 1.1 published

Stability AI published an SVD 1.1 image-to-video checkpoint designed to improve motion and consistency relative to the initial release.

2024-07-24

Stable Video 4D announced

Stability AI introduced Stable Video 4D, extending the Stable Video line toward multi-view video generation from a single input video.

🔄 Alternatives to Stable Video Diffusion

Frequently Asked Questions

What is Stable Video Diffusion?

Stable Video Diffusion is Stability AI’s generative video model family. The original release turns a still image into a short video sequence using latent video diffusion.

Is Stable Video Diffusion free and open source?

Model weights have been made available for download, but use is governed by Stability AI’s license rather than an unrestricted open-source license. Users should review the current license for commercial-use conditions.

Can Stable Video Diffusion generate video from text?

The original SVD and SVD-XT checkpoints are primarily image-to-video models. A user supplies a starting image; text-to-image software can be used first to create that input.

How long are Stable Video Diffusion videos?

The initial SVD model generates 14 frames and SVD-XT generates 25 frames. Actual playback duration depends on the selected frame rate, so outputs are generally short clips.

What hardware is needed to run Stable Video Diffusion?

Local inference generally requires a modern GPU with substantial VRAM. Exact requirements vary by checkpoint, resolution, precision, optimization settings, and the software implementation used.