|
|
|||
| forums: groups: | |||
|
Video producers selling scenes to people who want to know it really happened.
AI fetish providers selling the ability to create content to people who just want the visuals.
AI technology is advancing fast but I still don't see it being a viable alternative any time soon. Messy is just to hard to do right, the difficulty way surpasses clean images and videos.
Commercial models just aren't trained on big enough wam datasets whether that is video or images. you may get some some reasonable success but it can't generate what it hasn't been trained on.
The guardrails on commercial models is currently set up to restrict adult, NSFW and fetish material. SORA and GPT is well aware of these situations and will block them. I can't see the future but even with relaxed adult restrictions coming on images I can't see that pushing out to video.
Open models may have more long term future but they're a generation or 2 behind Sora2 and Veo3. If and when local hardware and models can compete with sora2, who is going to train them? Out of the box, open models are based on much smaller datasets. I can count on one hand the people who have trained a WAM model and even few with any success. It's never going to be the case of uploading 5 messy images and out pops a universal model. 

