Not Gal Gadot invites you for a blowjob and fucking. POV. Model comparison.
Duration: 14:28
Views: 28 565
Submitted: 1 year ago
Submitted by:
Celebrities:
Gal Gadot
Original Pornstar: Nora Doll
Description:
More boring tests... wish I could just pick one and start making stuff more frequently but at this point I'm pulling my hair out to figure some training issues some models seem to be having. Anyway as you can see it's the old DF192 vs 224 with modded face type that covers slightly more than full face (but it's not that noticeable, especially since both full face DF 192 and full face "extra" 224 were done with WF Xseg so the coverage is kinda similar between them). Hence the "extra" face type didn't really improve quality of seams and even made color matching more difficult (see my Miley fake, pay attention to forehead and it's shifting colors on some shots) I've redone that 224 and then trained it longer and also applied GAN but seems like I "overganned" it. Also I've tried my 256 but with less GAN. I feel like 192 is still decent compared to 256 but 224+ models seem to handle teeth better and some other tiny detail. Not sure about -U flag, doesn't seem to make face that much more SRC like, I feel like default or -D blends with DST better than -U does but you tell me that. All models trained and merged with the same settings with exception of merger up-scaling settings:
192 at 50.
224 "extra" No GAN at 45.
224 GAN and 256 GAN at 40.
192 and 224 "extra" trained with xseg mask applied, 224GAN and 256GAN with no xseg applied, merger masking mode, learned dst + src + xseg dst + src.
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.
192 at 50.
224 "extra" No GAN at 45.
224 GAN and 256 GAN at 40.
192 and 224 "extra" trained with xseg mask applied, 224GAN and 256GAN with no xseg applied, merger masking mode, learned dst + src + xseg dst + src.
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.
Tags:
Gal Gadot
gal gadot porn
gal gadot fake porn
gal gadot nude
gal gadot fake nude
gal gadot deepfake
gal gadot deepfake porn
gal gadot sex
gal gadot sex scene
gal gadot blowjob
gal gadot cumshot
tmb
tmb-df
model test
dfl test
model comparison
df
df-u
df-d
df-ud
dfuhd
saehd
dfl 2.0
model
gan
lrd
learning rate dropout
true face
style power
machine learning
Test
vs
tmbdf
truly mindblowing deepfakes
Download now to support TMBDF
Not Gal Gadot invites you for a blowjob and fucking. POV. Model comparison.
You currently have 0 tokens available
Insufficient tokens - 200 Required
Buy more tokensDonate to support TMBDF
Confirm tokens to TMBDF
1
2
3
4
5
6
7
8
9
10
11
12
«Preparing another model right now, finished previous polls, will be starting making fakes from them right now, unfortunately I won't be making the GITS VR one due to difficulties of the scene, will do some image fakes and regular cosplay instead.
Regular videos poll: />VR cosplay videos (regular celebs) poll: />VR cosplay videos (video game and animated characters) poll: />»
Reply
«more wwe please»
Reply
Reply to: markos.calderon
1 year ago«I've added Alexa Bliss to the poll
»
Ответить
Reply to: TMBDF
1 year ago«Thank you!
»
Ответить
«I dunno they all have extremely minor strengths and weaknesses comparitively. I thought the 256D held up a bit better in matching the fake face to the model's surrounding cheeks and forehead but the 256D also had some odd, brief color mismatches.»
Reply
«14:04 pretty much is the perfect frame to see which one matches the best on best conditions:
LowerLeft (so 224 longer training)
Overall, I found the 256 less convincing
224++ > 192 > 224 > 256
(at least that's my personnal take)»
Reply
Reply to: fradecroix
1 year ago«I agree with fradecroix.»
Ответить
Reply to: fradecroix
1 year ago«Technically that 256 is closer to 192 one since I've only used -D architecture type to get higher performance/resolution, dims are close to default so both models should and when you check it in other places it shares the look of base DF architecture. -U in the -UD ones (224) changes how lighting appears on the face (seems like shadows are a bit softer) and also for some reasone -U produces sligthly better, whiter teeth.
Maybe that 224 can be improved or I'll try to get other 256 ones working but this time also as -UD as opposed to just -D.
The -U in -UD is supposed to make faces resemble source better but it has disadvantage of making some angles for which ideal face in src is not present look worse, kinda makes it even less like LIAE, can't "cheat" that much and approximate so either face looks good or looks bad whereas -D and base compromises on resemblance a bit and thus makes the faces look more equal across all angles/expressions at the cost of aforementioned source resemblance.
256 doesn't benefit (at least not at default dims) as much from upscaling algorithm, 192 does so it makes face look sharper. At 14:06 when viewed in windowed mode 192 actually appears sharpest, followed by 224 with GAN, then 224 Ganless and 256 with GAN but thats because as I've mentioned I only did really low value for 256 as I wanted to avoid severe noise/grain issue I had on 224 (GAN at 0.1, 256 was done at 0.04).
Maybe the best one would be 224 with lower GAN value or maybe I just need to train 256 DF-UD and let it run longer to produce sharper results on it's own. Also 224 has higher dims, maybe it has something to do with it all too.
Will do more tests shortly, I'm now working on 2 commissions, after they're done I'll make that VR fake and after that do more commissions and comparison of different 256 resolution models (both full face and whole face).»
Ответить
«Please do a WWE Alexa bliss deepfake in the future!
»
Reply
Reply to: Polpol1
1 year ago«She is in the poll, go vote for her
»
Ответить
«I've uploaded this video in 1440p, hopefully it get's processes as such, but even if it's just 1080p watch at full screen so you can notice all the differences and let me know which one you prefer, note that none of them were "really well" trained, by that I mean long, had some issues with GAN at 224 but decided to still include it because at least you can check the difference between 224 that was trained with WF XSeg masks applied and one that was trained on default masks from extractor. But still the 192 trained for about 8 hours, 224 for 10, 224 with GAN I gave it about 12 hours and the 256 one was about 24 hours, all models were pretrained but 192 was already trained before to 280k, 224 a bit less at 240k and 256 started at about 50k iters.»
Reply