The most recognized open video-generation models you can run yourself
The open text-to-video and image-to-video models you can download and run — Wan2.2, HunyuanVideo, LTX-Video, CogVideoX and more. Ranked by recognition, each with the honest VRAM-to-run, license, and rent-a-GPU cost (video is heavy, so this matters). We list the recognized open video models with transparent costs; the quality call is yours.
How this is ranked: Objective modality filter, ordered by notability/recognition. We report real VRAM/license facts and honest rent costs (video models often need 24 GB+), never a quality/benchmark verdict. Note license restrictions explicitly (e.g. HunyuanVideo is non-commercial).
1. Wan2.2 T2V-A14B — Text → video, Alibaba (Wan-AI) · ~24 GB VRAM · commercial OK