So far, I'm still not happy with results I get from Hunyuan, but I'm happy to say that more options are maturing. LTX/LightTricks Video is another t2v+i2v model that is coming up in the world and lacks censorship. Though, it seems to be terrible at following prompts. I find no prompt (or at least no positive, a negative helps a bit) tends to get the best results for i2v. Though longer videos tend to devolve into surreal slop. It is no Kling, but it is very cost effective and even low enough requirements to run locally for me. CivitAI also supports it now, and it costs 63 buzz for a 5 sec video which is 10x less than the cost for Kling. Another one I want to try is WAN 2.1.
Here are a few clips that I'm reasonably happy with. There's a lot of room for improvement, but it is a start.
Attached to this post and in release archive under:
Releases\2025\012-Mom Balls Titfuck
And Teachers Balls:
Projects\2025\DC-MomBallsTitFuck
At some point, I'll need to work on releasing an image set. And I may make enough videos to compile into a reel or longer video eventually.