deep.whitecatchel.ru Forums

Some content may not be available to Guests. Consider registering an account to enjoy unrestricted access to guides, support and tools

  • Forums successfully migrated to new platform. Please report any bug HERE

[GUIDE] MVE - Editing Landmarks

SPT

Moderator
Staff member
Moderator
Verified Creator
If the pictures don't show, download them here :




Hello,

here I will explain how to fix the landmarks Manual re-extract couldn't fix using MVE.
It will basically be the same thing Angeloshredder explained, but I believe repetition helps as well as saying the same thing differently.

Important note : In case you use it for a dst and you have a huge number of landmarks to correct, like 500 to 2000, something you might not want to do in one day :
The way to do it is to fix 50-100 images for example, but selecting those corrected images in MVE (click above or under the image and then you'll have a large green zone under and above), then export these corrected images in the folder you chose.
But next time when you fix 100 additional images, if you export in the same folder, your previous images will be deleted, so you just have to create a new export folder for each new batch of images you will create to avoid this.
Then you just have to assemble all the fixed faces from export 1, export 2, export 3 in your aligned DFL folder to use them.

So you need MVE, and you need to be done with manual re-extracting of your project in DFL (I will not re-explain this).

Once you're done, create a folder for your project and inside put 3 folders (it's better not to directly use the folders inside of DFL Workspace but to copy them in a separate folder) :

1 folder with all your aligned faces
1 folder with all your frames (the image version of your data_dst video which is located in the root of your data_dst folder)
1 empty folder for corrected faces

Naming is not important (just don't use space or special characters), but for this example I will name folders respectively :
Faces
Frames
Corrected

Note : In my example I will show you how to use it with src pictures, but it works exactly the same with dst pics.

Open MVE.

Open your main folder (the one containing the 3 others).

9FX5Ufmh.jpg


Open images from your face folder.
(that's where I put the green dot)
hgjehvwh.jpg


Click on the icon on top of the right panel.

Change Video frame to face
eaI1aH4h.jpg


Then new options appear, set them as in the image, end with import face data, and let it do its thing :
s8y8Xrjh.jpg


Once done, save the panel, as in this image :
iyUikRuh.jpg


check if DFL landmarks loaded correctly :
oW4G61eh.jpg


then open images from your your frames folder :
20iulYkh.jpg


l4HddLNh.jpg



Filter the image as "no face"
QvZdScHh.jpg


Then you see the frames where DFL didn't detect frames :
7dllug5h.jpg


Then read the text on the next pictures to know how to add a face :
W96Nku7h.jpg


4C0XCaKh.jpg


mSOQ12sh.jpg


VD4dO8eh.jpg


Srqyy6vh.jpg


i9fR1Oph.jpg


EeVZNdBh.jpg


tpcrb35h.jpg


End.

- Made with Adobe's Shitty Green Arrows plugin for Photoshop ($115.95 but it's worth it).
 

TMBDF

Moderator | Deepfake Creator | Guide maintainer
Staff member
Moderator
Verified Creator
To be honest that example isn't that hard to extract with manual extractor and would take approximately 1 second to align it :D

Sticked.
 

SPT

Moderator
Staff member
Moderator
Verified Creator
TMBDF said:
To be honest that example isn't that hard to extract with manual extractor and would take approximately 1 second to align it :D

Sticked.

True. But it would have worked in MVE if she had a large object in front of her face, I just wanted to make a family friendly tutorial :D
 

TMBDF

Moderator | Deepfake Creator | Guide maintainer
Staff member
Moderator
Verified Creator
SPT said:
TMBDF said:
To be honest that example isn't that hard to extract with manual extractor and would take approximately 1 second to align it :D

Sticked.

True. But it would have worked in MVE if she had a large object in front of her face, I just wanted to make a family friendly tutorial :D

I see, after all we're all about that family friendly content over here ;)
 

etsnotes

New member
I'm always using MVE for my fakes. You can check alignments very quickly, and if there are faces without alignment between 2 aligned faces, you just have top copy/paste to retrieve them.
An advise: enable Auto Apply Changes to avoid forgetting to apply !
 

DeliveryBoy

Member
Hey guys is it possible to make this lesson into a video? I didn't go to school and my reading really sucks :( i am a slow learner.

i was born in a war torn country we dont have school :( i had to look after my family at really early age because most adults are killed by soldiers or taken away.

Just to prove my point, i cant even find MVE download :( <-- never mind i found MVE download on another thread.
 

SPT

Moderator
Staff member
Moderator
Verified Creator
thisdudethe7th said:
Any idea how to re-extract only the missing faces that I manually added in MVE?
Say I used DFL to extract thousands of faces, but did not use the built in manual extract tool for the missing faces.

If I used MVE to get the extracted faces metadata copied onto the original frames and then manually added
the missing faces like you described here, Can I re-extract just the missing faces I added?

Or do I have to re-extract them all?
No need to re-extract them all.

But first, because I'm not sure if that's clear for you : whatever you modify in MVE, you need to re-export the faces. If you just modify pics in MVE and close it without exporting, your faces are not modified.

Also, you always need to export in a new folder, otherwise MVE deletes everything in a non-empty folder before exporting.

So once you opened and parented to frames your aligned folder, you then modify whatever pics you want, and once done, just select "modified" in the upper selector (the one with "non", "1face in frame", "no faces"...etc).
Then you will only see the pics you have modified. You can select them all, and export to a new folder.
Once they are exported in this folder, check it so that the faces are numbered the same way as in your current aligned folder : if they have a "_0" at the end, and your aligned folder doesn't, then you have to remove the "_0" at the end of your MVE extracted images. Or leave it if your aligned frames also have the "_0". So that MVE new frames can replace some of the previous ones in your aligned folder in case it's needed (most of the time it is). Also check for the presence of files ending in "_1" in the MVE Folder and rename appropriately, as you should have done already for your aligned folder before all that.

Hope it helps, feel free to ask in case it's unclear.

Also, there's a bug in latest MVE, your frames must have "_0" at the end of their filenames or else MVE exported faces won't be recognized by DFL when you train and merge.
So don't use the latest MVE yet, use an older version, as long as you can see the "modified" tab, it means it's fresh enough.
 

SPT

Moderator
Staff member
Moderator
Verified Creator
thisdudethe7th said:
SPT said:
thisdudethe7th said:
Any idea how to re-extract only the missing faces that I manually added in MVE?
Say I used DFL to extract thousands of faces, but did not use the built in manual extract tool for the missing faces.

If I used MVE to get the extracted faces metadata copied onto the original frames and then manually added
the missing faces like you described here, Can I re-extract just the missing faces I added?

Or do I have to re-extract them all?
No need to re-extract them all.

But first, because I'm not sure if that's clear for you : whatever you modify in MVE, you need to re-export the faces. If you just modify pics in MVE and close it without exporting, your faces are not modified.

Also, you always need to export in a new folder, otherwise MVE deletes everything in a non-empty folder before exporting.

So once you opened and parented to frames your aligned folder, you then modify whatever pics you want, and once done, just select "modified" in the upper selector (the one with "non", "1face in frame", "no faces"...etc).
Then you will only see the pics you have modified. You can select them all, and export to a new folder.
Once they are exported in this folder, check it so that the faces are numbered the same way as in your current aligned folder : if they have a "_0" at the end, and your aligned folder doesn't, then you have to remove the "_0" at the end of your MVE extracted images. Or leave it if your aligned frames also have the "_0". So that MVE new frames can replace some of the previous ones in your aligned folder in case it's needed (most of the time it is). Also check for the presence of files ending in "_1" in the MVE Folder and rename appropriately, as you should have done already for your aligned folder before all that.

Hope it helps, feel free to ask in case it's unclear.

Also, there's a bug in latest MVE, your frames must have "_0" at the end of their filenames or else MVE exported faces won't be recognized by DFL when you train and merge.
So don't use the latest MVE yet, use an older version, as long as you can see the "modified" tab, it means it's fresh enough.

Thanks. I don't have a modified option, MVE too old?
Probably. Latest version is 0.8
 

SPT

Moderator
Staff member
Moderator
Verified Creator
thisdudethe7th said:
I've ran into multiple problems with MVE, not being able to extract faces from frames extracted using DFL.
It seems like either I'm missing something, or the WHOLE process should be done either in MVE or DFL.

(For face set extraction I get 'invalid directory name' error. Dependencies are fine, I tried linking to my local DFL and using Github, neither worked)

Using your local DFL should work, try to link to base folder for it, and if it doesn't work try to link to "internal" folder.

In any case, that's not what I do, I didn't link my DFL to MVe.
I extract in DFL, sort in DFL, then I open data_dst or dat_src folder in MVE and go on from there. First importing landmark info and parenting, then saving tabs, refresh, and then you can open your frames and start to fix landmark data.
 
Top


/video/12832/bibibeautypalace-new-song-wap-wap-bla-blaemma roberts leaked nudesÖzge Törer deepfake"alanah pearce nudes"emma langevin fakes"marisha ray porn"Vivi alto nude"anna camp nude"/photo/15101/mrs-morello-everybody-hates-chris-todo-mundo-odeia-o-chrisnoraxxxxseyma subasi deepfake pornAnmol Larke Dabal 3x Photo/photo/4792/anna-reshma-rajan-lichi/video/24870/a-pov-experience-with-tiffany-ann-hsu-hsu-wei-ning-1-full-version"selena gomez deepfake""dodger nude"Kajal2023xxx/video/21631/adept-gets-the-juice-bavfakes-preview"swimsuit succubus""jennifer carpenter nude"Sri divya failed attempt deepfakeAudrey Fleurot leakedscarlett johansson nudekajal is fuckAbbaxxx"gal gadot topless""elyse willems fakes"/video/28753/tanya-asmr-vertany-pornmadison beer deepfake.emilia bunnysusu.jpg nude leakporno meghan markleItzy yunapussydasha taran pornMillie bobby brown dilldo deepfake"ariana grande sex tape""elyse willems fakes"luxurica deepfake/photos/celebrities/mandy-takhar迪丽热巴 受不了的绝色旗袍美女"tessa thompson naked"Peyton List Titty-Fuck Deepfake"frivolousfox nude"/video/26906/not-burcu-ozberek-realistic-part-1-trailer-dm-for-full-10-00-minsCarice van Houten nude fake"deep fake emma watson"Jodie Comer naked deepfakedeepfakesxxx.com jennifer lopezPia wurtzbach deepfake pornDanIca Patrick deepfake/video/6974/alexa-rivera-2xvidsall live tayker xxx deeofake xxx/photos/celebrities/gouri-g-kishanNa eun pissing deepfacke/tags/gima-ashixxnxxkariyoona deepfake picture"kristen bell deepfake"danielle panabaker showing pearls /photo/1265/zendaya-facialtom holland fakes nudes"nicki minaj deepfake"AH02 : Amber Heard Vs Johnny /video/13146/luxurica-disha-patani-gives-you-a-sensational-handjob-full-version-1080p"leah remini nude"kat dennings pornfakesValkyrae deepfakewww kiss xxx chetraCatherine Middleton deepfake pornkyedae deepfake"chloe bennet porn"Lee Sung-kyung deepfakemr.babe porn/video/18887/twice-nayeon-trans-photoshoot/tags/oilSophia Diamond boob"alyson hannigan porn"Emma Watson Squirting DeepfakeAlexa bliss Mondomonger videosxxxAmber LaUra Heednadia streamer nude/tags/melissa-fumero-deepfakekatrina kaif 2021 xxxdina tlou sucking dick"kim basinger nude"