FaceFusion 3.0 Epic Trailer - Completely AI Generated and Showcasing Amazing New Features

Рет қаралды 8,676

Күн бұрын

Пікірлер: 48

@YAHURDME 4 ай бұрын

Thanks for covering this Awesome Ai software! As far as I know, you are the most in depth KZbin channel covering it's progress!

@AlgorithmInstituteofBR 4 ай бұрын

trailer was epic but still not epic enough for the product. this is amazing!

@laszlonemeth8032 3 ай бұрын

Fantastic program i follow it almost from the start . What more incredible is what a developement !!! Absolutely incredible guys !!! i love it

@shadyendeavor 3 ай бұрын

Though I was asked to create this trailer, I am not affiliated with the developers of FaceFusion. Just fyi.

@KILLEGAH 4 ай бұрын

Great Trailer ! 🔥🔥🔥

@PoldaranOfDalaran 4 ай бұрын

That was super neat

@amekomidarake Ай бұрын

I've managed a few videos with Face Fusion but the majority of them just hang with "ERROR" in the output window either during the initial processing phase or later in the process. I can't see what is different between the ones that work and the ones that don't.

@hashanperera5012 Ай бұрын

wow! do u have a discord community?

@shadyendeavor Ай бұрын

I do not, though I've thought of creating one. I'm just not sure how many people would actually use it.

@gamingdayshuriken4192 4 ай бұрын

loving it !

@johnunity3451 4 ай бұрын

👍Love it.

@YossiJimmy 3 ай бұрын

Does anybody knows what the expression restorer option does?

@shadyendeavor 3 ай бұрын

@@YossiJimmy it's supposed to restore whatever expression your target had after a face swap as sometimes it loses some of the, well, expressions when doing so. Mouth might not open or eyes might not close or look around. Honestly, it didn't work as well as I'd hoped, so I left it out of the trailer. It works a bit, buts it's usually too subtle to notice unless you're really looking.

@YossiJimmy 3 ай бұрын

@@shadyendeavor Thanks! I'm asking because I been working on a big project lately (which we actually finished on the same day FF3 was anounced...) and one of the biggest problems we were facing is that all of our source images were ones with a smile while the actress we used for the target wasn't smiling (as it should be). the result was a dramatic acting by the actress, but with a slight smile that came from the source. I wonder if that feature could have helped in this situation.

@shadyendeavor 3 ай бұрын

@@YossiJimmy It is supposed to, but results will vary. I'd say it's worth trying on a quick clip.

@YossiJimmy 3 ай бұрын

@@shadyendeavor Project is already delivered but yeah, it's worth a shot just out of pure curiosity. anyway, thanks for all the great videos, there isn't much information about FF online, so you are totally making a difference here!

@douglaspriester 3 ай бұрын

it uses liveportait

@gumvue.studio 3 ай бұрын

thanks, what do i need to have to use this? GPU 24GB?

@shadyendeavor 3 ай бұрын

Pretty much anything above a potato. But seriously, it's just as long as your system is new enough to actually be able to install the different components. As long as they can install, the app should run. It's just a question of how slowly.

@SuryaMusic-90 3 ай бұрын

You won't have any problem.. I'm running this is 8GB CPU.. and it's pretty fast..

@kevinehsani3358 4 ай бұрын

Any reasons skin color transfer and hair color models can not be added like 'live portrait' model which is added.

@shadyendeavor 4 ай бұрын

I'm not sure. Maybe you should ask the developers of FaceFusion. I am not one of them. You should ask on their Discord channel - which is linked in the description.

@chariots8x230 4 ай бұрын

It’s impressive, but I hope the facial expressions & lip-syncing will improve to look more natural. If we’re using a script that we want our character to follow, then it would be cool if we could control the lip-syncing & facial expressions separately. For example, for lip-syncing not to sound monotone, we may want to emphasize certain words, add pauses, use a certain tone of voice, a certain style of speaking, maybe apply a specific accent, and also show certain emotions in the voice. But, the facial expression is something that would need to be manipulated separately, because the tone of voice & manner of speaking, does not always accurately reflect the facial expression that we’re looking for. Also, I really like some of the AI tools that I’ve seen recently, where you can input a video of yourself talking/acting, and the AI would apply those same facial expressions & body movements to your character. I think that could be really useful because it would allow for much more expressive faces. Also, I’m looking for an AI tool that does well with scenes that have multiple consistent characters in them. I need options to manipulate each character individually, even though these characters are together in the same scene.

@shadyendeavor 3 ай бұрын

@@chariots8x230 I had hoped that the LivePortrait features that were being put into 3.0 were going to work along these ideas, but the developer definitely has his own ideas, to put it nicely. Currently, I think the best option would require some kit-bashing with FaceFusion and LivePortrait and maybe ElevenLabs.

@laszlonemeth8032 3 ай бұрын

Its all what you just listed is in the 3.0 version ....and f...in working beautifully

@berndvonherrnburg3799 4 ай бұрын

swap multiple faces (target ) in a single job!? How ?

@shadyendeavor 4 ай бұрын

Watch the video that I purposely linked at the end of the trailer and also linked in the description that shows you how to use the Queued Jobs feature.

@clintjoselito1567 4 ай бұрын

👌

@slow-moe3518 4 ай бұрын

Does it still suck for faces that are at a steep angle?

@NinjaKittyBonks 4 ай бұрын

Hmmmmm..... 1 thread...... 5 replies...... shadow-banned.... wonder who would write such a dissident comment? 🤔

@chariots8x230 4 ай бұрын

How is it being trained on the faces though? Do you just provide one image, or can you provide multiple images of your character for the training? For example, I’d want the AI to be able to accurately depict my character with different expressions, not only with their mouth closed, but also with their teeth showing. If we just provide one expressionless image for the training, then the AI probably won’t be able to depict the character’s teeth accurately & consistently when the character is showing expressions like laughter & anger, during which their teeth would be showing.

@TheLoy71 4 ай бұрын

I am not convinced if you look closely without bias. Very uncanny in some cases.

@laszlonemeth8032 3 ай бұрын

nothing in this world wich is nearly as perfect as you !! 😥

@rameshdevasi6720 4 ай бұрын

make for anime or stylized character

@A_Little_O_Everything 4 ай бұрын

Man, I tried to watch your video, but KZbin hit me with a 4 minute ad!

@algore3462 4 ай бұрын

Excellent AI software for sure but it is not true that there are no limits. There are actually limits to the video size e.g. 500MB and greater, if you are doing faceswap via FF.

@shadyendeavor 4 ай бұрын

Well, that's just absolutely not true. Before 2.6.0, Gradio (not FF) would forcefully convert files over 512MB. FaceFusion fixed this back in 2.6.0 so that it wouldn't do that anymore. I use files multiple GBs all of the time. But thank you for watching!