If you don't mind me asking, what software or platform are you using?
For the primary video clips, I'm using WAN 2.5. I also sometimes use Hailuo v2. Both work pretty well, although be prepared to do dozens of generations before you find one you keep.
I often use Hailou to generate still images from character references.
You can find uncensored (at least for the moment) hosting of WAN and Hailou's video models at
https://wavespeed.ai/. Unfortunately, WaveSpeed doesn't host Halluo's still image models, and Halluo's own hosting of them is *absurdly* censored.
I also use a lot of traditional video editing tools (Premiere, etc.) to do the actual assembly, add titles and sound, etc.