Not Pokimane volunteers her face for some model tests - 256 DF-D vs DF-UD

I like this video

156 likes

TMBDF

Donate

Duration: 19:47 Views: 163 523 Submitted: 1 year ago Submitted by:

Celebrities: Pokimane

Original Pornstar: Unknown

Description: Some more model tests, this time I've done same scene as df_7777 //deep.whitecatchel.ru/literotica/user/302949 used in his video: //deep.whitecatchel.ru/literotica/video/11644/fake-pokimane-casting-couch using the shared DST from Groggy4 //deep.whitecatchel.ru/literotica/user/3163 and SRC from df_7777.

His model is 448 DF-UD, I've used my two DF-D and DF-UD models at 256 (Full Face) and with slightly higher dims on both.

Merged both twice, first with same merger settings as df_7777 and no post processing or merger upscale and 2nd time with slightly different merging settings, merger upscaler and post processing, 5th run was done on my bigger SRC set for half the time the others were trained for and on the same base DF-UD model and merged with slightly different merger settings too so in total there are 5 different videos here, at the end there are still frames without any extra post processing, straight from the merger of all 5 attempts.

Obviously I didn't include df_7777 attempt here so for that you will have to watch his video. In either case SRC was quite blurry as Pokimane doesn't have a lot of good 1080p, let alone 4K out there so GAN didn't improve results that much (all mine and his models were trained with GAN set to 0.1).

Let me know in the comments which out of the 5 of my attempts look best and if I managed in your opinion to match the 448 model quality. I'm myself not seeing that much difference between non upscaled and upscaled + post -D model, with -UD the difference is much bigger and somehow the -UD upscaled model ended up being more blurry than either -D or non upscaled -UD which is really weird, especially since at the end of training -UD had lower loss and for some time -UD model was trained on higher batch size (in the later stage all models were trained at batch 6 and then on LRD and LRD + GAN batch size of 4 was used). -UD definitely looks better when the head is facing down, profiles seem to also look a bit better and overall lighting seems to be a bit better but -D somehow manages to look sharper. Also -UD model had slightly higher dims (336/84) vs the -D model (288/72) and was pretrained with smaller pretrain dataset that had only white and asian faces and more high res faces, -D model had bigger pretrain dataset that had all races and had more blurry faces in it.

If you enjoyed it make sure to press the like button, add it to your favorites and subscribe to my profile to be always up to date with my newest releases!

If you liked this video make sure to hit the like button, add this video to your favorites and subscribe to my profile! (you can then find my videos in your subscription page, upper right corner, next to messages and log out).

I put a lot of work and time into making those edits, please consider supporting me with a donation or video download using tokens (which supports me and the website). To donate to me click on the donate button next to my profile name under any of my videos and select cryptocurrency to find my BTC address or PM me for address.
Also the more given celebrity lookalike gets downloaded with tokens the more content of her I make!

Check my forum thread too where you can vote in the poll and decide what I make next! You can also suggest pornstar matches to celebs I make or plan on making, suggest scenes to edit or just simply request celebs and see what I'm working on right now, talk with me and so on:
//deep.whitecatchel.ru/literotica/forums/thread-request-poll-voting-vote-for-which-deepfake-you-want-to-see

I also make custom fakes of all celebs, if you want to order a public or private video please contact me here on the website or on our forum for details.