Lina Inverse getting boinked by Donald Duck could be a thing.
You can 100% already do this, with high quality, locally. Generate an image with PonyXL (and Lora’s if it doesn’t get the characters quite perfect), do img2vid with Hunyuan or this.
The catch is that it’s all unreasonably user unfriendly to set up, and most online UIs are remarkably shit. Basically AI land just needs a few more “integrators” to plug this stuff into each other and package it sanely (as opposed to researchers just throwing papers out or “entrepeneurs” making shome shitty closed cloud UI), some better hardware support, and you’re already there with no new models/tech.
You can 100% already do this, with high quality, locally. Generate an image with PonyXL (and Lora’s if it doesn’t get the characters quite perfect), do img2vid with Hunyuan or this.
The catch is that it’s all unreasonably user unfriendly to set up, and most online UIs are remarkably shit. Basically AI land just needs a few more “integrators” to plug this stuff into each other and package it sanely (as opposed to researchers just throwing papers out or “entrepeneurs” making shome shitty closed cloud UI), some better hardware support, and you’re already there with no new models/tech.