11–20 of 37 posts
Apparently this is only set up on the www subdomain -- trying to access it without www gives me a German 403. I feel like if I paid for 15 minutes, I'd feel cheated sitting there watching the clock tick down while it "starts the AI cluster."
And you are right, we should not include the startup time in the 15min. Will fix that :)
Apparently this is only set up on the www subdomain -- trying to access it without www gives me a German 403. I feel like if I paid for 15 minutes, I'd feel cheated sitting there watching the clock tick down while it "starts the AI cluster."
thanks. Looking into fixing the subdomain issue. And you are right, we should not include the startup time in the 15min. Will fix that :)
I realize that you're probably not targeting mobile, but just in case. :)
Edit to add output example: https://i.imgur.com/yflCAd9.jpg
The resulting image is much darker and more pixelated than my actual front camera. Given the difference between my face type and Nicholas Cage's (I'm fat and bald), I'm surprised it mapped as well as I did.
Notable failures:
1. Edge of face ("ghosted" edges with my actual round face protruding from the faked area)
2. Map mouthing (very difficult to change the faked expression)
3. Eyebrows (they don't seem mapped at all)
Notable successes:
1. Angle (the faked face is consistently angled when I move my face/head)
2. Complexion (no real skin color variance)
Earlier quoted context omitted.
thanks. Looking into fixing the subdomain issue. And you are right, we should not include the startup time in the 15min. Will fix that :)
I tried this on my Pixel XL and it worked surprisingly well, with one exception: the picture froze when I tried to select a different fake type. The page itself was still responsive (timer ticked down, I was able to "select" pictures, could scroll normally), but the final image was frozen and did not change despite waiting a minute or so. I realize that you're probably not targeting mobile, but just in case. :) Edit…
i have a 720p camera and a clear, bright image. i see no difference when choosing between female and male actor. it seems stuck in angelina jolie.
So you tried all of the available actors but it was always stuck on Angelina Jolie? Can you let me know what browser/platform you are using?
Earlier quoted context omitted.
I tried this on my Pixel XL and it worked surprisingly well, with one exception: the picture froze when I tried to select a different fake type. The page itself was still responsive (timer ticked down, I was able to "select" pictures, could scroll normally), but the final image was frozen and did not change despite waiting a minute or so. I realize that you're probably not targeting mobile, but just in case. :) Edit…
Thanks. Currently we are not targeting mobile as there are several problems with it. So the default - before switching - produced actually a working output on your Pixel XL, is that correct?
Chrome 69.0.3497.86
Android 9.0.0; Pixel XL Build/PPR2.180905.006
This is interesting, but what is something like this expected to be used for? Is it for novelty value or is there a practical reason one would want to deepfake themself? (Not a criticism, just legitimately curious.)
A more legitimate one would be for videoconferencing -- if you fake yourself in the best possible shape (say, after some weight loss, plus makeup, plus great hair style, plus wardrobe), you can ALWAYS put your best visual foot forward without the hassle/money. So you could be sitting there bloated after eating a bunch of tacos but in a videoconference, look your best.
Similarly, it could be great for film productions: prepare a library of photos in proper makeup/prosthetics, then deepfake the actual footage, allowing the actor to work normally without the time and money spent redoing them every day.
This is interesting, but what is something like this expected to be used for? Is it for novelty value or is there a practical reason one would want to deepfake themself? (Not a criticism, just legitimately curious.)
I can't believe you're missing the Snapchat-esque use-case. People are already augmenting their faces, this technology is right up the same alley.
Reached its capacity limit
