Mr DeepFakes Forums
  • New and improved dark forum theme!
  • Guests can now comment on videos on the tube.
   
Thread Rating:
  • 1 Vote(s) - 5 Average
  • 1
  • 2
  • 3
  • 4
  • 5
HaDenGLiv Tyler Had It Rough
#1
Hey all,

For this fake, first I trained Liv's face-set to itself for about 100k. I believe it converged faster and learned more details this way. After that I continued regular training.
For convert options, I used overlay, erode 10, blur 10, rct as color transfer and enabled super resolution. You can find model details below.
Enjoy.

You are not allowed to view links. Register or Login to view.

Mega:
You are not allowed to view links. Register or Login to view.

===== Model summary =====

== Model name: SAE

==

== Current iteration: 154872

==

== Model options:

== |== batch_size : 8

== |== sort_by_yaw : False

== |== random_flip : False

== |== resolution : 144

== |== face_type : f

== |== learn_mask : True

== |== optimizer_mode : 2

== |== archi : df

== |== ae_dims : 512

== |== e_ch_dims : 42

== |== d_ch_dims : 21

== |== remove_gray_border : False

== |== multiscale_decoder : False

== |== pixel_loss : False

== |== face_style_power : 10.0

== |== bg_style_power : 10.0
#2
Wow I should try that technique. How long did you train after swapping the facesets?
#3
Around 50k more. I trained it for background and face style
#4
Thank you HaDenG.
#5
Wow.  I've had the hots for her since I was a teenager and first saw Lord of The Rings. ?
#6
very nice
#7
A very good one! Very well fed beaver))

Forum Jump:

Users browsing this thread: 1 Guest(s)