Sister blog of Physicists of the Caribbean in which I babble about non-astronomy stuff, because everyone needs a hobby

Thursday 23 August 2018

Near-effortless realistic fake videos

Flat out amazing. And terrifying. Direct YouTube link : https://www.youtube.com/watch?v=GrP_aOSXt5U&feature=youtu.be


vid2vid is a "Pytorch implementation for high-resolution (e.g., 2048x1024) photorealistic video-to-video translation. It can be used for turning semantic label maps into photo-realistic videos, synthesizing people talking from edge maps, or generating human motions from poses."

https://github.com/NVIDIA/vid2vid

7 comments:

  1. You might almost be tempted to say that: truth, indeed, is no longer truth. Anything, any thing, any event or scenario can be representationally and compellingly fabricated from this technical inflection point onwards, as a consequence of this and related tools and methods. These kinds of tools and representational techniques have been in the pipeline for a very long time: artistically, culturally and cognitively. The question which strikes me is: what is the longer-term emergent complexity and cognitive, cultural impact of a world in which truth-in-representation can so clearly be abstracted away from reality, in which we can expect these technologies to be harnessed and exploited by all and (self-interested) sundry ? Ceci ne pas un paradoxe de menteur...https://plus.google.com/photos/112924411476447820619/albums/6592868489601469233/6592868488588100258

    ReplyDelete
  2. But basically it's the same as with photo manipulation, and would you call it terrific?

    ReplyDelete
  3. Well there's photo manipulation that takes hours of painstaking labour by trained experts, which is just painting with another tool, and then there's realtime photorealistic video manipulation by laymen.

    ReplyDelete
  4. There is something else here, as well. One fundamental mistake of cultural, artefactual or (even) technological analysis is to fail to acknowledge that an image, like a theory, an ideology or any other representational entity is not a static thing. Images and their associated technologies are a microcosmic slice of that culturally and historically-specific gestalt which they embody; ways of seeing are very closely related to ways of understanding and consequently, to ways of being.

    Not only this, though. Images, methods and frameworks of representational systems and their complex matrices of interdependent concepts and technical procedures are not merely representative of where a specific culture or technology "is at" at a particular time and in any isolated or static sense. They also, and profoundly, recursively re-enter, re-inform and reconstruct that culture and epistemological, semiotic system and logic from which they emerge. The complex and inflationary logical possibility-spaces of future systems-states are seeded by these technologies.

    The emergence of these ways of seeing, of re-seeing and remaking the world indicate a profoundly, deeply and dynamically reflexive representational matrix of mapping, re-mapping and indefinitely extensible psychological and cognitive systems of world-construction.

    In one sense, this is just another gee-whiz technical method. In another it reflects a deep and unfurling representational complexity, reflective of what is going on between our collective ears, as a culture and as a civilisation.

    ReplyDelete
  5. Kazimierz Kurz People who aren't experts or highly trained professionals.

    ReplyDelete
  6. Rhys Taylor AAA! Now I understand - You probably mean PROGRAMMERS ;-)

    ReplyDelete

Due to a small but consistent influx of spam, comments will now be checked before publishing. Only egregious spam/illegal/racist crap will be disapproved, everything else will be published.

Review : Human Kind

I suppose I really should review Bregman's Human Kind : A Hopeful History , though I'm not sure I want to. This was a deeply frustra...