How to reuse a models training on another project with same destination actor?

[wolfcreative.com.auParticipant]

I’ve followed the 3h long in depth tutorial and had great success with creating some deep fakes.

I was hoping to train a model that would replace my face onto an actors that I would reuse for new scenes over time?
Also helping with my lower VRAM GPU each time I produce a successful deepfake I was hoping it would speed up the overall time spent on the next one if the model could retain some or all of the previous work?

I know this is possible but cant seem to find clear guides for the process, I did a quick search in the forums but the reply was a little complicated,

Do I need to remove all my destination aligned files from JOB1 and replace them with ” new Aligned files from JOB2 ” and then keep training the model?

[defalafaParticipant]

depends what model you use

when df models are used:

simple delete the part you want to change … DST oder SRC or both and put inthe new files incl. aligned folder with faceset

then XSeg mask the models , train the XSeg mask, appy XSeg setting

start train SAEHD again

for LIAE models

the 3 scenarios below, what would be the recommended workflows?

1) Keep Source the same, replaced Destination actor
(since its the same source actor, i usually never delete any files)

2) Replaced the Source, kept the Destination actor
(deleted inter_B, but do i delete inter_AB?)

3) Replaced the Source, and replaced the Destination actor
(deleted inter_B, but do i delete inter_AB?)

[Author]

Posts