SOURCE-REVIEWED MODEL PROFILE
Wan 2.2 / speech-to-video
An audio-driven video model combining a reference image, speech and optional scene direction.
Metadata reviewed 2026-10-24 · Original creator: Wan-AI
What it does
- Local video-generation workflow
- Dedicated video runtime required
Local fit
Large GPU workflow; official example lists 80 GB VRAM. The official single-GPU example requires at least 80 GB VRAM. This is a larger-hardware reference, not a validated 12 GB recipe; community offload/quantized workflows require separate review and testing.
Video frames and audio samples are not a text context window.
Evidence and limits
Source reviewed only. No local rendering or avatar quality test performed.
External video runtime; not installed or wired to Chat generation
License and source
Review exact checkpoint, dependencies and input-media rights. No weights or runtime are bundled.
Read the original model card ↗
Reviewed source revision: dab4e9c55bbe4c8c4d03db1c2c98c7f0ac9c454b