This person does not exist ...

  • Poser Ambassadors

    Similar tech is making security footage less reliable too. It's no longer enough to have a video of someone caught in the act, as that video can easily be doctored now. There's also videos of various politicians and other public figures saying things they never actually said, at events they were never actually at, but according to those videos they were.

  • The teeth are a real giveaway that they aren't real (no British teeth comments please!) but the teeth don't look real

  • Whoa, he needs to see a doctor now!


  • @amethystpendant the orange hypnotoad's doctor will give him a clean bill of health! ;-) [The things that come out of peoples mouths these days!]

  • Incredible... this generator, in a smaller way inside poser, maybe as an replacement for the faceroom would be awesome!

  • So that's what happens to all those Facebook profile photos... ;o)

  • @3dcheapskate yes, exactly. That is where this thing gets its raw images from.

  • @ghostship I guess they probably used at least one monkey's backside in their dataset of valid human faces then ? ;o)

  • @3dcheapskate it looks like a bit of a black cat's face to me. There are SO MANY cat pictures on FB. lol

  • Facebook is too much work to scrape for image data in regards to such a deep learning project. Bing and Google images are the free and easily accessible image bases for use in this type of work. The respective companies supply the APIs for the scrape and less than a half dozen lines of code will start building your databases. I do it on a Raspberry Pi. An old core2 duo I have builds the CNNs from the image sets.

    The fact that this GAN is available to the general public is of great interest. I have yet to get the spare time to delve into this in detail, but; it may be possible to use the data to create morph target data and thus random characters with textures via Python in Poser. A randomly dialed up character via script is simple and easy. The results not necessarily so good. This might be a better way.

    I could just be making more work than the results warrant. That's why I find this interesting in an intellectual way. A thought exercise that results in a physical script if it is worth my undivided attention for a while as it fits into my current work stream.