Minshen's FramePack F1 image-to-video, built using RH official flow, default generates 30-frame videos

Advantages: ① Stability is better compared to WAN (not to mention Hunyuan...)

② Can generate infinitely; as long as the machine hardware meets the requirements, it can continue generating stably, unlike WAN and Hunyuan which typically crash after exceeding 6 seconds (a natural advantage of FramePack)

③ Generation efficiency is faster than WAN, and it does not rely on the resolution of the reference image (WAN and Hunyuan perform better with higher resolution input images)

Disadvantages: ① It is a bit too stable, resulting in dynamic effects that are not very good and somewhat stiff (background dynamics, physical trajectories, etc., are inferior to WAN); it performs better in scenes with less overall dynamics, such as inside a room

② The generated frame rate always feels like it's halved; the generated motion feels very slow, and 30 frames appear as though they are output as 15 frames (though in reality, it is 30 frames output). However, I have already applied frame rate doubling to mitigate this

③ Poor understanding of prompts; the generated actions are always conservative relative to the prompts.


Creative work is not easy. If this has helped you, please save, like, and frequently execute applications or workflows online. Your support is my motivation for creating. Thank you for your support! 

If you have any questions, please leave a comment. I will reply promptly upon seeing it. Feel free to exchange ideas, learn together, and make progress together!