Stable Video Diffusion
By Stable Video Diffusion · stability.ai
- Free plan
- Open source
- API
- Self-hosted
What Stable Video Diffusion is
Stable Video Diffusion is a model based on Stable Diffusion for generating video from text or images. It supports image-to-video animation and can generate 14 or 25 frames at customizable rates from 3 to 30 frames per second; the product page says videos are created in two minutes or less. Listed model versions include 14-frame, 25-frame, and 1.1. The model is intended for self-hosted use: Stability AI’s instructions describe downloading model weights and running them locally with the reference implementation. The hosted API endpoints were deprecated on July 24, 2025, and API access is unavailable. Stability AI says its foundation models are available through Amazon Bedrock and Amazon SageMaker JumpStart. The free Community License is described for users below $1M in annual revenue and remains subject to model-specific terms. Businesses above that threshold are directed to the Enterprise License, which has custom pricing and includes commercial use and implementation support. Stability AI is based in London, United Kingdom, and setup help is available through support tickets.
Who it suits
It may suit creators and developers who want to generate short videos from text or images and can run a model on their own infrastructure. The licensing terms distinguish users below and above $1M in annual revenue.
Good to know, and what to watch
- Supports both text-to-video and image-to-video
- Generates 14 or 25 frames
- Offers customizable rates from 3 to 30 frames per second
- Self-hosted deployment is available
- Community License is listed as free
- Hosted API endpoints are no longer available
- Self-hosting requires downloading model weights
- Enterprise License has custom pricing
Plans
Every plan and priceEnterprise License
Enterprise self-hosted license
Community License
Community self-hosted license
Free plan
The verdict
Stable Video Diffusion offers self-hosted video generation with image and text inputs, but its hosted API is no longer available. Review the model-specific license terms and the revenue threshold before using it commercially.
What the maker publishes
- API access
- No source
- API status
- The hosted Stable Video Diffusion API endpoints were deprecated on July 24, 2025, and are no longer available. source
- API status
- The hosted Stable Video API was deprecated on July 24, 2025 and its endpoints are unavailable. source
- Capterra praise
- The single Capterra reviewer praises realistic detailed images, the Relight feature and value for money. source
- Commercial threshold
- Commercial use above USD $1M annual revenue requires an Enterprise license. source
- Commercial use
- Yes source
- Company founded
- CB Insights lists Stability AI as founded in 2019. source
- Company headquarters
- LinkedIn lists headquarters in Los Angeles, California. source
- Company legal name
- Crunchbase identifies the maker’s legal name as Stability AI Ltd. source
- Company size
- LinkedIn lists Stability AI in the 51–200 employee range. source
- Data use
- The privacy policy says users can opt out of use of their Inputs and Outputs for model training through Account Settings. source
- Deployment
- Stability AI says Stable Video Diffusion can be deployed on your own infrastructure with a self-hosted license. source
- Developer code
- The public Generative-Models repository contains scripts, configs, requirements, tests and sampling utilities for local use. source
- Frame counts
- The product page says the model can generate 14 or 25 frames. source
- Frame-rate control
- Frame rates can be selected from 3 through 30 frames per second. source
- Frame rates
- The product page says it can generate 14 or 25 frames at customizable frame rates between 3 and 30 frames per second. source
- Free plan
- Yes source
- Hardware guidance
- Clore’s deployment guide lists 16GB VRAM for 14-frame SVD and 24GB for 25-frame SVD-XT, recommending RTX 4090 or A100 hardware. source
- Headquarters
- London, United Kingdom source
- Image-to-video
- Yes source
- Independent output limit
- An independent evaluation describes SVD clips as no longer than about four seconds, with possible low motion, imperfect photorealism, poor text rendering and problematic faces. source
- Independent quality limit
- The same guide says newer models have surpassed SVD in motion quality and offer features such as native audio that SVD lacks. source
- Independent technical limit
- A 2026 technical guide says SVD is image-to-video and does not natively accept text prompts, despite the current official product page advertising text-to-video. source
- Integration
- Stability AI says its foundation models are available on Amazon Bedrock and Amazon SageMaker JumpStart. source
Checked against the maker's pages on 21 Oct 2026.