Read the FAQs and search the forum before posting a new topic.
This forum is for discussing tips and understanding the process involved with Training a Faceswap model.
If you have found a bug are having issues with the Training process not working, then you should post in the Training Support forum.
Please mark any answers that fixed your problems so others can find the solutions.
Hi guys,
Just started playing around with faceswap. I've done some nice ones before using other 'tools' out there, but this one seems really nice. Great work!
I was just curious, do you think that in the (near) furure, we'd be able to work with higher resolutions? (say 256x256 or even 512x512) .. What's currently the limiting factor (patience?)
Thanks!
Ton
The limiting factor is GPU RAM. For every doubling in output size, the model needs to be scaled exponentially to handle the increase in data (as an example, a 64px image is 4,096 pixels, a 128px image is 16,384 pixels, 512px is 262,144 pixels). The model has to be scaled to handle this.
Resolution isn't the be all and end all, detail is the most important thing, with a decently scaled output. Strides are being made in that direction, but it's a balancing act not to have a higher resolution output that looks like crap.
Higher resolution models also take exponentially longer to train.
My word is final
