The model is based on the 14B version.

It supports both Image-to-Video (I2V) and Text-to-Video (T2V) conversions.

For I2V, I’m using wan2.1-i2v-14b-480p-Q5_K_S.gguf, and for T2V, I’m working with wan2.1-t2v-14b-Q5_K_S.gguf. I’m unsure how other models would perform.

The test videos were generated with a strength setting of 1.

I’ve found that Text-to-Video can sometimes struggle with capturing specific angles or perspectives, but it handles the rest of the scene instructions fairly well. The quality of T2V images isn’t the best, though, as I’m using the .gguf model at a low resolution and with fewer steps, so it could be down to my settings. However, Image-to-Video seems to produce good results overall.

Best workflow for this LoRA is embedded into the images above.