Stable Video Diffusion Image-to-Video XT
Open Weights
Stability AI
Stable Video Diffusion (SVD) Image-to-Video XT is Stability AI's latent diffusion model that takes a still image as a conditioning frame and generates a short video from it. The XT checkpoint is finetuned to produce 25 frames at 576x1024 resolution, using a finetuned f8-decoder for temporal consistency; a frame-wise decoder is also provided. The model cannot be controlled through text and produces clips of about four seconds or less.
Strengths
- Open weights published on Hugging Face, runnable and finetunable locally
- Generates 25 frames at 576x1024 from a single conditioning image
- Finetuned f8-decoder improves temporal consistency across the clip
- An alternative frame-wise decoder is shipped alongside the temporal decoder
- Commercial use permitted at no cost below the licence's USD $1,000,000 annual revenue threshold
- Image-to-video needs no prompt engineering: one still frame is the whole input
Caveats
- Vendor states the generated videos are rather short (<= 4sec) and the model does not achieve perfect photorealism
- The model cannot be controlled through text, and cannot render legible text in the output
- Vendor states the model may generate videos without motion, or very slow camera pans, and that faces and people in general may not be generated properly
- The Community Licence terminates once a user or its affiliates exceed USD $1,000,000 in annual revenue; an enterprise licence is then required at Stability AI's sole discretion
- The model card publishes no numeric benchmark table, only a human-preference user study versus GEN-2 and PikaLabs, so no benchmark scores are recorded here
- Vendor
- Stability AI
- License
- Stability AI Community License
- Modalities
- video
Capabilities
Vision
Audio
Video
Tool Use
Resources
Pricing
Weights are downloadable from Hugging Face under the Stability AI Community License. The licence terminates for any user or affiliate group generating more than USD $1,000,000 in annual revenue, who must instead request an enterprise licence from Stability AI.