WAN local mode is supported in full / trial versions, WAN API also in LITE version. In local mode, you first have to click select the folder where to store models and also click ‘Install requirement’. While installing, you should also click on ‘Download models’. Note that models take a huge amount of space, like 100’s of gigabytes and all the different models need to be downloaded separately (like text-to-video vs imahe-to-video. Minimum on 16gb GPU memory needed, much more recommended 🙂 Currently generation speed is quite slow, tens of minutes for 5sec clips, since it’s pretty new model. Updates to local mode will hopefully be available.
For API mode, which also takes several minutes per generation, you can use Alibaba cloud. It’s bit more trickier (and sketchy) to set up, compared to others, because they require credit card to be added, without any pre-paid amount of credits. You can set up spending limit alarm, but I haven’t yet found out how to set up hard limits. So, beware… You need to create your API key in Alibaba Model studio: https://modelstudio.console.alibabacloud.com/?tab=globalset#/efm/api_key
Since WAN 2.2 uses lot’s of memory and generation times are unpredictably long, you need to manually release the used memory after you are no longer generating more videos. Using for example stable diffusion or framepack at the same time WILL NOT work!!! Using API based services however works, while waiting for results: