Uncanny! Emotional expression is here via Generative AI image to video. This is getting really close to production quality. And for PreViz, this works. We are soon going be in a situation where the quality is going clash with copyright and commercial use. If the training data is fair use, and the video is from a machine... does anyone own the output? Either way it's stunning how far this is advancing Source: https://lnkd.in/gqe9qGnZ
That is not uncanny what is uncanny is how much it reminds me of the devil
Still a lot work to do, it looks wrong and unnatural... It does not consider that bone and teeth is not flexible, so this looks scary ...
Blackmailing with deep fakes, evidence manipulation, copyright rights and so on
Still uncanny in the uncanny valley sense of the word.
Looks good Bob but is off in certain places and oddly the fact that it is off was comforting to me . I guess I am one of those who don’t know what to do with AI if it gets so close to real stuff
Thanks for sharing my video!
GitHub?
Great case study Bob Duffy
I see Data people . . . . . . Data Liberator | Perpetual Eclecticist | Infinite Learner | Business Intelligence | PMO | Project Controls | Innovation Evangelist | Navigator of Rabbit Holes
4moFacial flexibility to the max! Also however, a flexible skull, in parts, which is probably why this seems slightly wrong/unnatural. The jaw bone and musculature over it seem to be the biggest challenge? When GenAI can relate this without issue, the training wheels will be fully detached… Your point about “free use” derived output is interesting. Are we likely to see licence free output? I suspect not. I suspect more likely there will be some share licence required between the platform providing the GenAI video capability and the entity prompting the output. Whether this be by subscription model, or output by output is a wider discussion. I still think that all AI Video/Image output should be tagged/watermarked to make it clear it was generated and not real. It needn’t be right across the image, but somewhere clearly visible, without distracting.