MrDeepFakes Forums
Some content may not be available to Guests. Consider registering an account to enjoy unrestricted access to guides, support and tools
-
We are looking for community members who are intested in helping out. See our HELP WANTED post.
Gillian Anderson - Candyman -
- Thread starter WilhelmVacker
- Start date
Wow---I thought I was the only person in the world who did SFW ones with her. I've done a 2-person fake before, but that's longer and more complex than any of mine. Did you do each scene individually then compile together, or did you "sequentially build on each previous version"? (I do the latter----basically faking a fake, doing one face at a time then making that new fake the destination)
Any info on the model settings? And did you do a single source or a huge compilation? (I've tried using both only single-interview vids as well as just "every pic on the net" for making a source model for her)
Any info on the model settings? And did you do a single source or a huge compilation? (I've tried using both only single-interview vids as well as just "every pic on the net" for making a source model for her)
It is complex...
I am using DeepFaceLab_CUDA_10.1_AVX
== Model options:
== |== batch_size : 8
== |== sort_by_yaw : False
== |== random_flip : True
== |== resolution : 128
== |== face_type : f
== |== learn_mask : True
== |== optimizer_mode : 3
== |== archi : df
== |== ae_dims : 512
== |== d_ch_dims : 21
== |== batch_size : 8
== Running on:
== |== [0 : GeForce RTX 2070]
train SAEHD to 150.000 (one week an an RTX 2070) data_scr 4120 hand sorted images
convert SAEHD
Use interactive converter - n
Choose mode: 1 overlay
Choose mask mode: (6) learned*FAN-prd*FAN-dst - takes long to convert.
Choose erode mask modifier [-200..200] (skip:0) : 0
Choose blur mask modifier [-200..200] (skip:100) : 300
Choose output face scale modifier [-50..50] (skip:0) : 0
Apply color transfer to predicted face? Choose mode ( rct/lct skip:None ) : rct
The main problems with multiple face is the time you have to invest to sort out the thousands of aligned images and match the wanted and the training time.
I am using DeepFaceLab_CUDA_10.1_AVX
== Model options:
== |== batch_size : 8
== |== sort_by_yaw : False
== |== random_flip : True
== |== resolution : 128
== |== face_type : f
== |== learn_mask : True
== |== optimizer_mode : 3
== |== archi : df
== |== ae_dims : 512
== |== d_ch_dims : 21
== |== batch_size : 8
== Running on:
== |== [0 : GeForce RTX 2070]
train SAEHD to 150.000 (one week an an RTX 2070) data_scr 4120 hand sorted images
convert SAEHD
Use interactive converter - n
Choose mode: 1 overlay
Choose mask mode: (6) learned*FAN-prd*FAN-dst - takes long to convert.
Choose erode mask modifier [-200..200] (skip:0) : 0
Choose blur mask modifier [-200..200] (skip:100) : 300
Choose output face scale modifier [-50..50] (skip:0) : 0
Apply color transfer to predicted face? Choose mode ( rct/lct skip:None ) : rct
The main problems with multiple face is the time you have to invest to sort out the thousands of aligned images and match the wanted and the training time.