Wan 2.2 SVI 4-pass image-to-video workflow Chains the NSFW SVI-Camera GGUF Q8 unet pair with the SVI v2 PRO LoRAs across passes, Generates up to 25-second videos on a single high-end GPU.
Models (~40 GB total)
HuggingFace:
UMT5 XXL fp8 text encoder — 6.3 GB
Wan 2.1 VAE — 0.2 GB
SVI v2 PRO HIGH LoRA — 1.1 GB
SVI v2 PRO LOW LoRA — 1.1 GB
lightx2v I2V 480p distill LoRA — 1.4 GB
Wan2.2 Lightning low-noise LoRA — 1.1 GB
CivitAI (NSFW)
SVI Camera HIGH — GGUF Q8 — 14.4 GB
SVI Camera LOW — GGUF Q8 — 14.4 GB
Description
FAQ
Comments (3)
Very tidy workflow. Can you also list the upscale model please? Thanks!
small note - the SVI model links in the description seem to be api urls or something and aren't clickable.
can you fix it to the proper page?
Seems to be a civitai issue with files in general. The workflow has the same issue. Likely not something the uploader can fix.