Stable Video Diffusion
Open-weights image-to-video — turn any still image into a short clip.
About
Stable Video Diffusion, released November 21, 2023, was Stability AI's first serious move into generative video: an open-weights image-to-video model family with two variants, SVD (14 frames) and SVD-XT (25 frames), outputting short 2-4 second clips at 576x1024 and 3-30 fps. It reuses the SD 2.1 image encoder with a temporally-aware decoder, and Stability AI positioned it as a research preview — explicitly not for commercial use — to seed an open video ecosystem the way Stable Diffusion did for images. In practice it was slow on consumer hardware (Ars Technica measured ~30 minutes for 14 frames on an RTX 3060) and early outputs leaned toward pans, zooms, and animating smoke or fire rather than moving people. Historically important as the base half the open video world fine-tuned, but superseded for actual use by everything that came after.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Stability AI-reported: preferred over leading closed models in user preference studies at release (provider-reported)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: Released Nov 21, 2023 per Stability AI GitHub release notes; SVD (14 frames) + SVD-XT (25 frames), 576x1024.; Open weights, research preview; historically important base for open video fine-tuning.; Stability AI-reported user preference wins over closed models at release.; Limitations: research-only license (non-commercial); slow on consumer GPUs; superseded by later models..