Updated for Wan 2.2

New version no longer uses a smooth pass.

(There isn't a small 1B version for Wan 2.2 this time.)

This workflow uses primarily GGUF quantized models to reduce vram where possible. The current version runs comfortably on 16GB of vram when using the 4 bit (q4_k_m) models.

Models Needed

Wan 2.2 i2v high noise q4

Wan 2.2 i2v low noise q4

Wan 2.2 ti2v (q8 or q6 recommended)

Goes into models/unet

umt5_xxl_fp8_e4m3fn_scaled

Goes into models/text_encoders

wan 2.1 vae

wan 2.2 vae

Goes into models/vae

4 step lightning loras

Goes into models/loras

high noise

low noise

5b fastwan

Any good upscaler model.

I recommend RealEsrgan_2xPlus

And 2xNomosUni_esrgan

Goes into models/upscale_models

FAQ

Do I need both Wan 2.1 and 2.2 VAEs?
Yes. The 2.2 VAE only works with the 5b model (confusing, I know). Make sure the main section loads the 2.1 VAE, and the upscale section loads the 2.2 VAE.