Foundation models, finetunes, and cloud APIs for AI video generation. Filter by platform, capabilities, and safety level.
by Wan-AI
Alibaba's flagship text-to-video open-weights model. 14B transformer with strong motion coherence, runs on 40GB+ VRAM. State-of-the-art open T2V.
by Wan-AI
Wan 2.2 image-to-video variant. Animates a reference still photo with motion described in the prompt. Same 14B backbone as T2V.
by Wan-AI
Compact 5B-parameter Wan 2.2 variant. Combined T2V+I2V in a single model. Runs on ~20GB VRAM — practical for 24GB consumer GPUs.
This catalog tracks the models behind uncensored AI video generation — open-source image-to-video bases like Wan 2.2 alongside the cloud and API platforms that turn a prompt or a still into motion. Each entry lists the platform, maximum resolution and duration, frame rate, audio support, and safety level, so you can shortlist before committing render time.
Filter by platform to stay inside a single ecosystem (LoRAs and workflows rarely transfer across video bases), by capability when you specifically need image-to-video, video-to-video, or synced audio, and by safety level for an SFW-only or unrestricted subset. Cloud and API models render without local hardware; open-weight models run on your own GPU — check the VRAM and platform notes on each card first.