Updated: Aug 1, 2025
base modelThis GGUF file is a direct conversion of Wan-AI/Wan2.2-I2V-A14B / https://huggingface.co/bullerwins/Wan2.2-I2V-A14B-GGUF/tree/main
There's a 14B I2V (Image To Video), in a high+low noise pair
You're expected to run the high noise as a base and the low noise as a refiner
In the Image To Video params:
Set the regular Video Model to the high noise model,
and set the advanced Video Swap Model to the low noise model,
and leave Video Swap Percent at
0.5
Reference CFG range is
3.5
For both 14B types:
FPS is either
24
or16
, ideal value not entirely clear. The Wan 2.1 base model was 16.Sigma shift may be worth experimenting with. The default is 8, but a wide range of values are functional.
You can use Lightx2v or other causvid-likes on the Wan 2.2 14B (not on the 5B)
For I2V, this seems to "just work" - but expect hallucinations as this is not optimised for this checkpoint.
Added some basic Workflow (ComfyUI) and a Preset (SwarmUI): https://civitai.com/models/1823089?modelVersionId=2063163