2026 04 18: Updated workflow, fixed the bug from yesterday.

2026 04 17: Updated the base model, upgraded the smooth model version to 20, (Version 30 changes the effect of the starting frame of the image, so version 20 is better), if you need the previous AI application, you can use my workflow. You only need to modify the saved nodes and the prompts you want.

LU_Short Series Six-Part Set Part 3: Image-to-Video
High-dynamic image-to-video generation. AI can assist in generating prompts, or you can use your own prompts.
If AI is generating prompts, you only need to input the theme and the number of storyboard frames, and select true: ai prompt.
If you are providing your own prompts, you only need to define the prompts and select false: div prompt.

ps: I generate 1280*720 resolution videos, 3 seconds of video takes about 3 minutes (02:35/03:20).
ps: The default fps is 16. If you need to modify it, use the workflow, or I will release a small application for frame interpolation and HD enhancement later.

LU_Short Series Six-Part Set Part 1: Text-to-Image
https://www.runninghub.cn/post/1987052252043087873
LU_Short Series Six-Part Set Part 2: Image-to-Storyboard
https://www.runninghub.cn/post/1988519232235331585
LU_Short Series Six-Part Set Part 3: Image-to-Video
https://www.runninghub.cn/post/1987043435041460226
LU_Short Series Six-Part Set Part 4: First-and-Last Frame-to-Video
https://www.runninghub.cn/post/1987043353088954370
LU_Short Series Six-Part Set Part 5: Multi-Frame-to-Video
https://www.runninghub.cn/post/1987043344478089218
LU_Short Series Six-Part Set Part 6: Video Synthesis and Audio Synthesis.
https://www.runninghub.cn/post/1987052320276066306