MrDeepFakes Forums
  • New and improved dark forum theme!
  • Guests can now comment on videos on the tube.
   
  •  Previous
  • 1
  • 2
  • 3
  • 4(current)
  • 5
  • 6
  • ...
  • 14
  • Next 
dpfksDeepFaceLab 2.0 - General Overview and Discussion
#31
try archi: df instead of dfhd.
and check in taskmanager how much ram it uses and how much ram and how much swap
und my 2080ti i can use just something about 9gb vram of the 11gb...
#32
(01-26-2020, 06:13 PM)aXu Wrote: You are not allowed to view links. Register or Login to view.try archi: df instead of dfhd.
and check in taskmanager how much ram it uses and how much ram and how much swap
und my 2080ti i can use just something about 9gb vram of the 11gb...

I tried everything and nope still the same. 
it use around 3 4 gb of ram.
got 32 gb of ram and i changed the swap size to initial 12 gb - max 36 gb
in your case when it only use 9 out 11 gb of vram. does it say the same message as me ? 
or did you see it elsewhere.
thank you again.
#33
Long time lurker that is becoming a power user of this software.  Very interested in multi GPU performance and this new super resolution algorithm.  I can't find any information on the internet related to "super resolution upscaler "RankSRGAN" is replaced by "FaceEnhancer""  

Anyone know where FaceEnhancer came from or what it's architecture is?  Would love to read more about that.
#34
then u have to start small. default settings, df instead of dfhd and set batch to 4 by res128.
that shoud work. or u send me a pm and i help u over teamviewer. think u do something wrong
#35
(01-26-2020, 07:51 PM)aXu Wrote: You are not allowed to view links. Register or Login to view.then u have to start small. default settings, df instead of dfhd and set batch to 4 by res128.
that shoud work. or u send me a pm and i help u over teamviewer. think u do something wrong

I dont have any probleme making DFL work. i already did some projects. Sorry if i didnt make it clear.
it's just a shame that i can't fully exploit my graphic card, and push the resolution or others a bit higher.
but thank you.
#36
trained as usual but the merge simply doesn't have any overlay. it just shows the usual dest video. (?) is there a naming convention ? i named the model actor1_actor2 with that underscore.  Dodgy - GOT IT - i don't know if it's known but the model CANNOT be named with an underscore. if it's named as said , actor1_actor2, it will be named automatically actor1_actor2_SAEHD , that will prevent the merger from finding it.
#37
Quick question for the author.
Is it planned to increase the size of the extracted faces for future versions?
#38
Dont even think about it...
Just a low percent of the users use res 256, IT needs to much resources. An increase in res would not ben happen
#39
(01-26-2020, 10:45 PM)aXu Wrote: You are not allowed to view links. Register or Login to view.Dont even think about it...
Just a low percent of the users use res 256, IT needs to much resources. An increase in res would not ben happen

Well, it could be an option.
I wouldn't mind to train 5x times the time for a close up.
#40
[quote pid='15450' dateline='1579912002']
I wasn't saying VLC is a virus... I was saying you shouldn't be paying for h265, reffering to what you wrote: "Maybe a codec download will fix that?  It's asking for money in Windows.  No biggie..."
[/quote]

Totally...  I wasn't saying that you were saying that Smile
I found a free version of the h265 codec in the microsoft store, but by default, these newest versions of windows 10 tries to get 99 cents from you to buy it.  There's a video illustrating that here... You are not allowed to view links. Register or Login to view.

But that's the least of the issue here.  I have no idea why, but I just tried with another project in V2.0...    If I send a finished product to my google drive and then save it to my iphone like I've done so may times in the past with version 1 videos, it doesn't even appear in the album. Has anyone else gotten a finished video on to their iphone successfully yet?  I just tested it with an older V1.0 video from last week just to make sure it wasn't something on my end, and it ends up in the photos/videos album as soon as it's done downloading.  BTW, it does seem like these v2.0 vids are downloading.  It takes around 45 seconds or so and finishes, but it's just not there.  So weird! Maybe iPhones don't see the files v2.0 is producing as a video for some reason? Is that possible?

UPDATE.... Also, thanks for this solution. It works! (for those of you that might run into this same issue)

If you're on PC you can try editing VideoEd file in F:\DF\DFL2.0\_internal\DeepFaceLab\mainscripts and change line from
output_kwargs.update ({"c:v": "libx265",
to
output_kwargs.update ({"c:v": "libx264",
  •  Previous
  • 1
  • 2
  • 3
  • 4(current)
  • 5
  • 6
  • ...
  • 14
  • Next 

Forum Jump:

Users browsing this thread: 1 Guest(s)