This is a video workflow based on the wan2.2-FUN model

The effect of automatically filling in the start and end frame prompts is better than manually filling them, so an LLM model is needed

Note: The Runninghub LLM API Node node in the workflow needs to be replaced with other LLM model nodes, because the Runninghub LLM API Node is a platform-specific node

Upload the start and end frame images, and it can be used
Usage:
Default 5-second video, the maximum dimension of the longer side should not exceed 1280
The default parameters have been tested multiple times and have relatively good effects

You can click 【Test Effect】 it only takes about 7 minutes throughout, if you feel the effect is satisfactory, you can then download the workflow to deploy locally!