Stable Video Diffusion Review — Stability AI's Foundational I2V Model: SVD, SVD-XT, and the Birth of Open-Source Image-to-Video

Stable Video Diffusion (SVD) by Stability AI was the first widely accessible open-source image-to-video model, released November 2023. Built on SD 2.1's U-Net with temporal attention layers, SVD generates 14 frames; SVD-XT extends to 25 frames (~4 seconds at 6 FPS) at 576×1024. Minimum 8 GB VRAM, comfortable at 24 GB. API deprecated July 2025. Superseded by Wan 2.1 and HunyuanVideo for quality, but still used in low-VRAM workflows. No MCP server. Rating 3/5.