FaceFusion: The Definitive Deep Dive & Walkthrough - Everything You Always Wanted to Know About...

Summary

This tutorial video guides users through updating and using Face Fusion 2.1.1 for face swapping in images and videos. It covers installation, selecting execution providers, and adjusting settings for optimal performance. The script explains how to choose face swapper models, enhancers, and use face debugger tools for precise masking, ensuring a natural look in the final output. Viewers are encouraged to experiment with different settings for the best results.

Takeaways

• 📝 Ensure you have the latest version of Face Fusion (2.1.1 at the time of the video) for the best experience and features.
• 🔄 If you have version 2.0.0, follow the upgrade instructions in the video to update to the latest version.
• 💻 Use the terminal to navigate to the Face Fusion directory and execute commands to pull the latest version from GitHub.
• 🛠️ Create a virtual environment for Mac/Linux or use specific commands for Windows to prepare for the installation.
• 🖥️ Select the appropriate execution provider based on your system (Core ML for Mac, CUDA for Nvidia, etc.).
• 🔄 Be prepared for additional downloads and installations when you first run the updated Face Fusion.
• 🔧 Adjust the execution thread count based on your GPU's VRAM for optimal performance.
• 📁 Set a custom output path for your creations, but be aware of potential issues with spaces in the path and the path not being saved.
• 🎥 Use the face swapper tool to replace faces in images or videos by selecting a source and target.
• 🔍 Utilize face analyzers to select specific faces in the video based on order, age, and gender when needed.
• 🛡️ Experiment with different face mask types and settings in the face debugger for precise control over the face swap effect.

Q & A

What is the latest version of Face Fusion mentioned in the video?

The latest version of Face Fusion mentioned in the video is 2.1.1.

If someone has never installed Face Fusion, what should they do according to the video?

If someone has never installed Face Fusion, they should watch the clean install video linked in the top right of the video or in the information below.

What is the first step to upgrade Face Fusion if you already have version 2.0.0 installed?

The first step to upgrade Face Fusion is to open the terminal application and ensure you are in the Face Fusion directory.

What command should be typed in the terminal to pull the newest version of Face Fusion from GitHub?

The command to pull the newest version of Face Fusion from GitHub is 'git pull'.

For Mac and Linux users, what commands should be used to create a virtual environment after pulling the latest version?

For Mac and Linux users, the specific commands to create a virtual environment are not provided in the transcript, but it is mentioned that they should type the commands for Mac and Linux.

What is the purpose of the 'face selector mode' in Face Fusion?

The 'face selector mode' in Face Fusion is used to choose the face that you want to swap in the target image or video.

How many face swapper models are available in Face Fusion by default?

By default, there are five different face swapper models available in Face Fusion.

What does the 'keep FPS' option do in Face Fusion?

The 'keep FPS' option in Face Fusion keeps the output file to its original target's frame rate instead of converting it to the default 25 FPS.

What does the 'face debugger' tool in Face Fusion do?

The 'face debugger' tool in Face Fusion is used for face masking and helps to visualize which face will be swapped and what area will be affected.

What are the three different face mask types available in Face Fusion's face debugger?

The three different face mask types available in Face Fusion's face debugger are 'Box', 'Occlusion', and 'Region'.

What is the recommended face enhancer model for a more realistic film look according to the video?

The recommended face enhancer models for a more realistic film look, according to the video, are 'Code Former' and 'GPen etc'.

Outlines

00:00

📲 Face Fusion Software Update and Installation Guide

This paragraph provides a step-by-step guide for ensuring the viewer has the latest version of Face Fusion software, version 2.1.1, and how to upgrade it if they already have an older version installed. It covers the process of opening the terminal, navigating to the Face Fusion directory, pulling the latest version from GitHub, and setting up a virtual environment for different operating systems. The instructions also include choosing the correct system for the installation, such as default for M1 Macs or Cuda for Nvidia GPUs, and the initial steps to run the software after installation.

05:01

🛠️ Customizing Face Fusion Settings and Interface Overview

The second paragraph delves into the customization options available within Face Fusion, including execution providers, thread count, and memory settings. It discusses the importance of selecting the right execution provider based on the user's hardware and the impact of the thread count on rendering speed. The paragraph also touches on the output path settings, which dictate where the final videos and images are saved, and the limitations encountered with path customization. Additionally, it outlines various options like keeping the original frame rate, retaining temporary files, and skipping audio or downloads for certain models.

10:04

🔍 Face Selection and Swapping Process in Face Fusion

This section explains the process of selecting and swapping faces within Face Fusion. It introduces the face selector mode, which allows users to choose a reference face or select multiple faces for the swap. The paragraph discusses the use of face analyzers to refine the selection based on order, age, and gender. It also covers the 'many' mode, which applies the source image to as many discernible faces as possible, and the importance of the face detector model and face detector score in accurately identifying faces for the swap.

15:04

🎨 Face Swapper Models and Enhancer Settings in Face Fusion

The fourth paragraph focuses on the different face swapper models available in Face Fusion and their respective performance in terms of rendering time and output quality. It presents a comparative analysis of models like INS-Swapper 128, Blend Swap, and Sim Swap 256, highlighting the best options based on the test results. The paragraph also explores the face enhancer settings, detailing various models like GFP-GAN 1.4, Code Former, and Gpen bfr 512, and their impact on the final output's realism and plasticky appearance. It concludes with a discussion on the blend setting in the face enhancer and the use of frame enhancer for video upscaling.

20:05

🖼️ Advanced Masking Techniques and Face Debugger in Face Fusion

This paragraph introduces advanced masking techniques in Face Fusion, such as the face debugger, which allows for precise control over which parts of the face are affected by the swap. It explains the use of face mask types like the box, which can be adjusted for padding to fine-tune the area affected, and the blur setting, which helps in making the swapped face look more natural. The paragraph also covers the occlusion mask, which intelligently handles objects partially covering faces, and the region mask, which can be used to preserve specific facial features. Additionally, it discusses the use of the face detector score to fine-tune face selection and avoid switching between similar-looking individuals.

25:05

📝 Conclusion and User Engagement Invitation

The final paragraph wraps up the video script by summarizing the key points covered in the tutorial and inviting viewers to engage with the content. It encourages viewers to ask questions in the comments section if they have any, and to like, subscribe, and leave a comment if they found the video helpful. The paragraph ends on a positive note, wishing viewers a wonderful day and reminding them to take care.

Keywords

💡Face Fusion

Face Fusion is the software being discussed in the video, which allows users to perform face swaps on images or videos. It is the central theme of the video, with the script focusing on its installation, upgrade, and usage. The video script mentions the latest version, 2.1.1, and how to upgrade to it, indicating its importance in the video's content.

💡Virtual Environment

A virtual environment in the context of the video refers to an isolated space where the Face Fusion software operates, separate from the main system to manage dependencies and versions. The script instructs viewers on how to create this environment for Mac and Linux, and provides alternative commands for Windows users, emphasizing its role in the setup process.

💡Terminal Application

The terminal application is a command-line interface where users can execute various commands. In the video script, it is used to navigate to the Face Fusion directory and to pull the latest version from GitHub, showcasing its importance in the software's installation and upgrade process.

💡GPU (Graphics Processing Unit)

A GPU is a specialized electronic circuit designed to manipulate and alter memory to accelerate the creation of images in a frame buffer intended for output to a display device. The script discusses the selection of Cuda or ROCm based on whether the user has an Nvidia or AMD GPU, which is crucial for optimizing the Face Fusion software's performance.

💡Face Swapper

Face Swapper is a feature within the Face Fusion software that enables users to replace the face in a target image or video with a source face. The script describes different face swapper models like 'in_swapper 128' and their performance, highlighting the core functionality of the software.

💡Source Image

The source image is the image from which the face is taken to be swapped onto another image or video. The script explains how to select and use a source image within Face Fusion, such as using Robert Downey Jr.'s image for the example, which is essential for understanding the face swap process.

💡Target Image/Video

The target image or video is the media onto which the source face is to be swapped. The script details how to select the target and how the face swap process involves changing the person's face in the target media, as demonstrated with a clip from a Tyler Hubbard video.

💡Preview Window

The preview window in Face Fusion is a tool that allows users to see the changes before rendering the final output. The script mentions its importance and how it can be used to adjust settings and select specific frames, which is vital for the video's tutorial on how to use the software effectively.

💡Face Debugger

Face Debugger is a feature in Face Fusion that provides visual cues for the face swapping process, such as a red square around the face to be swapped and key points for tracking facial features. The script discusses its use for fine-tuning the face swap and ensuring accuracy.

💡Occlusion

Occlusion in the context of the video refers to a smart masking tool within Face Fusion that automatically handles obstructions in the target video, such as a beer mug covering part of a face. The script demonstrates how occlusion can improve the face swap by correctly masking out such obstructions.

💡Face Mask

A face mask in Face Fusion is a tool that defines the area of the target image or video that will be affected by the face swap. The script explains different face mask types, such as 'Box' and 'Occlusion', and how they can be adjusted to control the extent of the face swap effect.

Highlights

Ensure having the latest version of Face Fusion (2.1.1) for new features and improvements.

Watch a clean install video for initial Face Fusion setup if not installed.

For version 2.0.0 users, a quick upgrade process is outlined.

Using terminal commands to pull the latest Face Fusion version from GitHub.

Creating a virtual environment for Mac/Linux and specific commands for Windows users.

Installing the new version with system type selection for optimal performance.

Potential automatic downloads and installs when running Face Fusion for the first time.

Basics of using the app, including selecting face swapper and execution providers.

How to choose the correct execution provider based on the platform and GPU type.

Using the source and target images or videos for face swapping.

Explaining the execution thread count based on VRAM and optimal speed.

Adjusting the output path for saving created videos and images.

Options for keeping the original frame rate or skipping audio in the output.

Using face selector mode to choose which face to swap in a video.

Utilizing face analyzers to select the desired face based on order, age, or gender.

Different face swapper models and their impact on render time and quality.

Choosing the best face enhancer model based on the desired output look.

Adjusting the face detector score for better face selection accuracy.

Using the face debugger for fine-tuning the face mask and ensuring accurate swaps.

Exploring different face mask types for various scenarios in face swapping.

The importance of fine-tuning face mask padding and blur for natural results.

How to use occlusion for handling face swaps with objects partially covering the face.

The use of region masking to preserve specific facial features during a swap.

Invitation for viewers to ask questions and engage in the comments section.

Transcripts

hey everyone before we begin looking at

the interface and figuring out what

everything does make sure you have the

newest version of face Fusion installed

which as of this video is release is

version

2.1.1 that added a few new features and

improvements if you do not have face

Fusion installed at all make sure to

watch my clean install video that I will

have linked in the top right of the

video right now or in the information

below the video however if you already

have version 2.0.0 installed let's do a

quick upgrade on it you can skip to the

next chapter if you've already done this

first open up your terminal application

and make sure you are in the face Fusion

directory you should be able to just

type

CD face

Fusion once you're in the proper

directory type the following command get

poll and hit

enter what this will do is it'll it'll

pull or basically download the newest

version of face Fusion from GitHub as

I've had problems in the past I'm going

to next create the virtual environment

so type the commands for mac and Linux

if you're on either of

those

and and there now you're in a virtual

environment for Windows users these are

the commands that you should be using

instead of those last

two then to install the new version

that's been hold type in this

command at this point as per before you

need to pick out which type of system

you're using so for me I'll just be

picking default as I am on M1 Mac or any

Mac you'll be picking default if you're

using an Nvidia based GPU choose Cuda

otherwise I believe rocm is for anyone

using an AMD GPU CPU I'm not sure who

would actually pick just

that then for another part of the

install you choose again kind of the

same things but instead for Mac you will

choose between Legacy which is if you're

on an Intel Mac otherwise you will

choose silicon if you are on an M1 M2 or

M3

Mac Cuda again once again if you are on

Nvidia and I'm not sure which of these

two are what you're supposed to be using

if you're on AMD I believe it's one of

those

two watch it do its

installation and that's

it at this point it should be installed

but as per usual I believe when you run

it for the first time you will get some

more downloads and installs

automatically let's find out and as

always to run it type the command

python

run.py looks like there aren't any more

installs that was it so you're ready to

go at that point pick up the URL here

and copy that and put it into your

browser and here we are starting off I'm

going to show you the absolute Basics so

that you can easily and quickly create

your first swap if you already know the

app pretty well or just need to know

about specific parts of the app feel

free to skip around the chapters I've

set

up first up in the top left you should

already have face swapper

selected next go down to the execution

providers and depending on which

platform you are working on make sure

you have the correct selections um well

selected on a Mac if you can make sure

core ml is

selected

on an Nvidia GPU machine make sure you

have Cuda

selected on an AMD GPU machine make sure

you have well whatever option you're

supposed to have not sure to be honest

weirdos again for quick and simple skip

everything else on the left now in the

middle you'll drag an image to

Source or you can click on and find your

file this is the image you want the

person in the new image to look like

below that is the target drag or click

and find your image or video that you

want to change the person's face from

into the source

images at this point if your image or

video only has one obvious person in it

that you want to swap you should be able

to click the start button Below in the

middle and then wait for the output and

done that at times can be literally all

there is to it now what the heck is

everything else El let's look at the

execution providers I already mentioned

it but just make sure you have corl Cuda

or whatever it is for AMD selected

otherwise renders will take much much

longer the execution thread count is

based on the amount of vram you have so

this depends on the GPU that you have I

have a 16 GB M1 Mac Mini so I set it to

16 and that seems to work closest to the

optimal speed but because it's shared

memory on Apple silicon anywhere from

about 10 to 16 gets roughly the same

results apparently with dedicated

graphics cards you can go with a 2:1

ratio so if it's a 16 GB card you can go

with 32 threads the execution Q count I

don't touch most people will probably

keep that at one I would only try

increasing it if you are already getting

really good speeds but try it out with a

small file and test it for

yourself max memory is there more as a

limiter leave it at zero and it will use

as much memory as it can otherwise if

you need a certain amount of memory

available for other tasks on your

computer you can change it to suit your

needs now output path is something that

I wish worked a little bit better you

can see that by default it has a very

weird path this is where any videos and

images you create will automatically

reside you can change it to whatever

directory you like however I ran to a

problem where it doesn't seem to like

any spaces in the path and if it did

have any spaces it would fail if I tried

to Output anything the other problem

that I have with it is that it doesn't

save the path that you change it to so

the next time you come into phas Fusion

it will be set back to its default I

just don't change it and I make sure I

have easy access to the finder for that

directory on a Mac in order to see that

directory you need to show hidden files

to do this type command shift period and

again to hide them once visible you can

easily find the V or VAR directory in

your user directory and go from there on

Windows and Linux I'm not sure what the

deal is sorry but it's actually probably

just easier to get to the options are

pretty simple Concepts keep FPS will

keep your output file to its original

targets frame rate instead of converting

it to the default 25

FPS keep temp will make it so that all

of the images created for each output

before it merges them will be kept in

the default folder therefore no longer

temporary skip audio will strip the

audio from your target well technically

it just won't add it back in when it's

creating the output file and Skip

download is kind of a weird one that has

little use in my opinion when you switch

to a particular face swapper model or a

face enhancer model it has to download

it the first time if you have this

selected it won't download it at least

that's how I've come to understand it

it's something I've never never had to

use for but maybe useful in some case

that I don't understand maybe if it's

corrupt or something for the rest of the

features we're going to need to actually

show what is going on so I'm going to

put in a Target I'll use a short clip

from a Tyler hubard video with that in

now you can see a ton of stuff lighting

up and change so let's start from the

top right with the preview

window this is probably the most

important window as it's where you can

see any and all changes before or

rendering out your final picture or

video just below that is the preview

frame slider it's currently on frame

zero but you can slide it around to a

different number to see a different

specific frame in the preview window

switching between which frame it is on

is very useful when you have multiple

Cuts or scenes within a single video for

reasons I'll explain with the options

below first with trim frame start and

end by default they will be set to the

first and and last frame of your entire

clip and if you don't change it it will

render the entire video when you hit

start but if you only want one portion

of the video you use the preview frame

above to see what frames you actually

want to render out and then set the

start and end to those say I only wanted

to show one quick cut of Terry Crew

dancing I can see that at frame

24 it starts this cut and then it ends

at frame 40

so I can just easily change the trim

frame start and the trim frame end to 24

and 40 and that way it'll only render

out that one little cut of 16 or so

frames under face selector

mode you can see there are three options

reference one and many with reference

selected you can see the reference face

just below that shows a bunch of

faces to show how this Works we're going

to need to bring in our source

image it seemed appropriate to use

Robert Dy Junior for this example I'm a

dude playing a dude disguises another

dude you might not see anything happen

at first and that could be because it's

trying to put the face on someone that

was in the beginning of the clip but not

in this shot so what you want to do now

with reference turned on is select the

face that you want to swap Robert Dan Jr

on to

you can just as easily pick someone else

from the references because the preview

is such a low resolution it can be

difficult to see how much is actually

changing until you render it out while

still using reference let's look at the

reference face

distance I'm not really sure why it's

called distance but what it does is find

faces that are either less or more like

the image you chose if you turn it up

too high it will start to think everyone

is your reference and and change

multiple

faces but if you turn it down too

low it will think it's a different

person just between frames or scenes

with different lighting when it's still

the same person so you'll get flashing

in and out if it's close or just shots

that aren't changed at all usually the

default works great but you'll find

reasons to play with it a bit depending

on your video after reference is one

this will always only pick one face from

the entire video the trick is getting it

to pick the one you want now if you only

have one person in the entire video then

it's going to be pretty simple if you

have two three maybe still pretty easy

but when you have more faces in the

video especially when you get pretty

high up this can become a huge

paa but if you skip down to the face

analyzers which they spell incorrectly

here you have a few tools to try and

pick the the one you want and these

analyzers work for reference and mini

also first with order given the first

two options you'd either get the person

on the far right or the far left of the

picture and it's for the first frame of

your video if it's someone not in the

first frame good luck if we wanted to

pick Terry Cruz top to bottom would work

and with large to

small I have no idea how best or worst

are decided but in this situation if you

wanted one of the guys just to his left

or right I don't see how this would work

the age analyzer it's pretty obvious but

it will allow you to narrow down if you

have a range of obvious age differences

within the video if I quickly go back to

reference you can see that accurately

portrays them all as adults no children

teens or seniors the last one is

gender I shouldn't have to explain this

one as it's the the same as age but well

gender male or female however this one

doesn't seem to work as well just to

show what I

mean if I choose

male there's only one image and that

looks like it's a female and for some

reason that's it for reference if I put

the female for some reason that thinks

all of them including Terry Cruz is a

female so back to the face selector

mode let's check many what this one is

going to do is use your Source image on

as many faces as it can make out well

enough and you can use the analyzers for

exclusions 99% of the time I'm using

reference mode and almost never use the

analyzers there are rare occasions where

I find them useful but not too often the

last thing on the right side that isn't

part of the masking is the face detector

model section I've always left this on

retina face however with Retina face

selected it makes the next part the face

detector size absolutely worthless if

you try anything other than the than the

default 640x

640 it will come up as an

error if you switch it to

unet and then switch it

around it will allow you to change a

size I haven't played around with this

much but what it seems to be is

something that helps you select faces

when they're too close to the edge or

cut off normally I wouldn't use it but

if you can't seem to get a face selected

try this

out as for the face detector

score though I don't use this too often

it can be useful this is usually good

set to the default Center but this is

how tight or Loosely determines what a

face is within your image or video

turn it down a bit to try to recognize a

face that wasn't all that

obvious but go too low and it will think

all random things are

faces turn it

up and you'll get fewer faces which may

be a good thing as the ones that doesn't

pick up still probably wouldn't have had

much of a good output anyway but go too

high and it won't pick up

anything now let's switch back over to

the very first thing we looked at for a

half second earlier frame processors

face swapper is what you're going to

want to use almost all of the time for

obvious reasons this gives you the

option of five different face swapper

models by default it's set to ins

swapper 128 and it's a pretty good one

in fact maybe the best option I was

going to let you figure it out what is

best on your own but instead I ran a

test of each one to show you what they

all look like with the basic settings

here you can see that blend swap looks

pretty bad and it took 56

seconds in swapper looks pretty good and

took 15

seconds in swapper fp16 looks marginally

better but also took 2 minutes and 53

seconds Sim swap 256 looks pretty bad

but only took 15 seconds

and the 512 unofficial took longer at 26

seconds but looks even worse so as I

said in swapper 128 is probably the best

option unless time really is of no

concern and you want the marginally

better output of the

fp16 and just to show you how it turns

out if you put on the default face

enhancer on the two good ones and the

maybe okay one here are how they turned

out both of the in swappers look pretty

darn good but I don't think the face

enhancer was able to save the Sim swap

so definitely no point in trying the

other

two now that you've seen what the

default face enhancer can do let's take

a look at its other models and

settings face Fusion gives you seven

different model options I've not tried

all of them but I'm pretty happy with

the ones I've used like face swapper

let's look at the outputs of the four

that I have used first here's what the

clip looks like without turning on the

enhancer for not being enhanced I'd say

it turned out pretty decent and now here

are the four results gfp Gan 1.4 looks

great and took 2 minutes and 3 seconds

to

render however it can look a bit

plasticky depending on the

scene code forer took a bit longer at 2

2 minutes and 58 seconds looks a bit

better by not looking so

plasticky gpen bfr 512 looks almost

identical to code former and took the

exact same amount of time at 2 minutes

and 58

seconds restore former took 4 minutes

and 2 seconds and seems to have some

problems you can see how it even

distorted a large portion of the clock

so to me it's a tossup between code

former and gpen etc for the more

realistic film look while gfp

gan4 is faster and also works very well

for cartoonish and current AI videos

being made with Pika Moon Valley and Gen

2 Etc there's still one more setting for

face enhancer and that's the

blend it's default is set to 80 and

tends to be good enough but you can put

it down a bit to get rid of some of the

plasticky look I never go below 50 or up

for some reason if you want more more

now frame

enhancer I never use basically it's a

video upscaler if you need your video to

be larger in scale then maybe this will

be useful for you you have three options

of

models the first one will double the

resolution while the other two will

quadruple the resolution so if you start

with a 1080p it'll get 4k or 8K

respectively I use real ESR Gan so

either one of these two all the time for

still images and would recommend going

with that the main reason I never use it

though is the additional time is kind of

insane if you're ever wondering why a

render is taking so long double check

that you didn't accidentally choose

frame enhancer instead of face

enhancer now the last one is face

debugger which I think should be renamed

personally as you aren't really doing

any debugging but this is where all the

face masking comes into play Let's turn

that on and see what we get so let me

bring in a new

Target and a new

source just to make sure that she is the

one that's being selected let's choose

her in the reference face and it's

actually doing the wrong thing right now

it should be showing a box and not doing

what occlusion actually does let me try

switching those around see what

happens

there we go weird so now with just her

selected you can now see a red square

around her face that's the face mask box

to show you which face will be swapped

and what area will be affected as you

can see face mask is selected under the

face debugger

items turn that off and the square goes

away KPS is also turned on but it

doesn't seem to be doing anything with

this Frame as I believe her face is too

small in this to show up so let's go to

a later point in the

clip and because I don't want this to

get too weird let's pick a different

Source picture for

this you can see when I pick his

reference face that he now has a square

and five dots that signify where his

eyes nose and lips are for tracking I

think KPS stands for key points or

something like that now let's make him

Steve Corell

now B

box I'm assuming bounding box as far as

I can tell doesn't really do anything

other than highlight which face will be

swapped turn that

on okay now it's not getting any so

maybe this is a bug and this is

hopefully something they'll fix really

soon but it seems whenever I actually

put the source image in now during face

debugger it will then completely get rid

of all the information in the preview

frame to see what is actually shown for

masking so like right now if I put Steve

car back

in it changes the image however the

masking information for the debugging is

gone changing frames or whatever does

not change that if I get rid of Steve

krell's

image sure enough there the information

is back again if I choose somebody else

let's go with Michael J

fox it changes the image but again lose

all the

information so for now I'm just going to

keep the source out of there so you can

see what's supposed to be happening

without the image changing and the last

of the face debugger items is score you

turn that on it puts up a number in the

box and based on that number that's how

high you can put up the face detector

score slider and still have it select

that face so seeing that it's 0.9 91

right now I should be able to put it all

the way up to

0.9 sure enough it still shows his face

selected if I put that to

0.95 no longer

selected the real reason for using this

is maybe if you have so many people that

are in it that looks somewhat similar

and you want to get as close to possible

to only selecting that person maybe you

can fine-tune it so that it won't

switch I don't know so now back to face

mask let's first turn off the score and

let's turn off the bounding box now take

a look over on the right where all the

masking tools are available first we'll

take a look at the box that's titled

face mask type as it set up now you can

see where that square is around his head

everything within the square can be be

affected by the face swap including his

jacket and shirt most likely nothing

will happen to them with a simple shot

like this but if you want to make sure

not as much is affected that's where you

can choose the different padding sliders

to make the square smaller or larger in

this case maybe you just want to slide

it up a little bit from the bottom so

you don't get so much of a shirt so you

just do the face mask

padding bottom and slide that over a

little bit see how much that changes

oh that changed the left side for some

reason let's try that

again okay I don't know where that came

from but all of a sudden the left side

is all messed up so let's do the

left bring that in a

bit that seems to be

working you just want to get close you

don't want to go over the actual head in

any way whatsoever unless there's

something on the face you do not want to

be mased

the other option is the blur this is the

amount or I believe more accurately the

width of the outside of the mask that is

blurred to make the new face look more

natural on the target video or image

it's very difficult to see anything

change on such a small preview that

you'll have to test it out for yourself

on your own

projects now instead of the Box face

mask type there is

occlusion

this seems to be a bit smarter masking

tool and I hope it is as there are no

settings for it with this current shot

though there isn't much of a reason to

use a cusion so let's take a look at

another

frame you can see here that the man in

the middle has a beer mug partially

covering his face usually this would

have been a big issue and caused weird

artifacting now with

occlusion let me first select his

face you can see how it automatically

mask out the mug with zero work go back

a few

frames and you can see how the mask

opens up at the bottom as more of his

face is

revealed and for the final option which

is

region let's bring in one last

clip

this is what you use if you don't want

something to change say you want to keep

someone's crazy eyes or keep their

specific eye color and the way they look

around you can turn off the

eyes and they won't be affected by the

swap looking at the other options I'm

sure you get the idea you can of course

also use the different face mask types

at the same time so if you like those

eyes but he's drinking a beer at the

same time use both occlusion and

Regen

anyway I think that's more than enough

for now if you have any questions please

feel free to ask in the comments I

appreciate you watching the video please

like subscribe and leave a comment below

have a wonderful day and take

care