Low VRAM wan image to video workflow

This is a simple workflow I used to create a video from a single image using the wan2.1-i2v-14b-480p-Q4_0.gguf model.

The workflow can generate a 6-second video in approximately 30 minutes on a 3050 GPU (4GB).

I’ve added a LoRA node to the process — it worked well on my setup. However, if it’s not efficient on yours, feel free to remove it.

For upscaling, I used an anime-style specialized model from ReXeeD123 on Civitai. Alternatively, you can use the Real-ESRGAN model from Hugging Face: RealESRGAN_x4plus.pth, which generally provides better results.