Dub with Lip Sync Any Video using ElevenLabs and FaceFusion (Kit Bash)

[@StormlightStudiosOffical]

Hi i need help which is the best enhance face model

[@shadyendeavor]

I generally use the default gfpgan_1.4. If it's looking too plasticky, try turning down the Face Enhancer Blend to 50-60. Otherwise, try out codeformer, which is better for still images, but on video takes a lot longer time than gfpgan.

[@StormlightStudiosOffical]

@@shadyendeavor okay thanks, I have managed to do a really impressive brad pitt deepfake already and because it's kinda a closeup I wanted to know which is best as my un edited outputs were pretty mild I appreciate that

[@StormlightStudiosOffical]

@@shadyendeavor also where it say output speed I think its under output quality should I set it to medium or low for best quality output or leave it at very high

[@shadyendeavor]

@@StormlightStudiosOffical to be honest, the best option would be to use the new Pixel Booster, but currently it isn't free. This allows a much better initial swap that instead of the inswapper_128, you can do 256, 384, 512, 768 or even 1024. I use the 512 most of the time and then I end up not even needing the face enhancer. I plan on doing a video on these soon. And if you don't want to spend the money, they should be available for free when 2.7.0 comes out. I'll put a link to the place to purchase all of the extras for FaceFusion in the description now.

[@shadyendeavor]

@@StormlightStudiosOffical nothing says "output speed", so I assume you're talking about the "Output Video Preset" where you can choose between "ultrafast" and "veryslow". This is telling ffmpeg how fast you want it to encode your video once it's done rendering all of the frames. This shouldn't change the size or quality all that much. I tend to always keep it on the default "veryfast", but if I was worried about the output, I would maybe put it up to "medium", but never higher. And making it any faster than "veryfast" doesn't really help all that much if you're worried about time. So short answer, just keep it between "veryfast" and "medium". Outside of those are unnecessary.

[@katherinesera]

Interesting how it worked, and yet it didn't. The final output worked great though.

[@shadyendeavor]

Anna did all of the hard work.

[@juncando]

Sounds just like Chipp Walters

[@shadyendeavor]

You're saying I sound like him? Either way, I'll be honest, I have no idea who that is.

[@NinjaKittyBonks]

"... and well, some not so great" 😅🤣😂 As we see more and more developers throw their AI generated hat into the market, those prices shown, will drop like a rock. A purrfect example of just how amazing the free market can drive up quality, options and ease of implementation. This has led to the point at which I can push plastic buttons on my desk to create lines of code onto a server on the opposite side of the planet, only to then have some fool who has never met me, decide these lines of code are a CG / ToS violation and erase them. We are talking billions of $ in development costs, many thousands of hours of millions of people, so that this do this all at a microscopic fraction of that cost, only to complain when I return here in 5 min. to find it was all a waste of my time. *see what The Kitty did there? 😸

[@shadyendeavor]

Funnily enough, after all of this time (maybe 1 1/2 years) that I've really been following AI, I have yet to see a price drop on a service. Instead, it has tended to just push the competition's tech advancement and more features. And the rate of advancement has been insane. So did this comment work right away? Or was this your fifth attempt?

[@NinjaKittyBonks]

@@shadyendeavor ... Was my first attempt, as was both of the "Kitty's comment system" ones. My guess is that there is one person who has an axe to grind, so puts a "time-out" on my abilities. This sounds insane to consider I am only one of, what... 2B users here? Seems this "time-out" is not renewed when the night-shift guy comes in, so on your account and maybe 1 or 2 others, first time is the charm. On others... no quite so charmed... lol . RE: Prices dropping: True. There are a number of those leading this AI development and they they are competing with each other and not yet been able to relax and just tinker on the user-friendly side of things with tweaks, when the costs are far lower. It is all engine pushing (graphically this is HUGE), as it then takes the hardware developers to engineer and create faster more powerful CPU's and such. . This is where my Win XP is exceedingly short of all this... the processing of graphics. Worked out fine, as I am an audio guy and the editing I was doing, though extremely good, for use of a basic WAV editor within Nero 6 (ya... THAT long ago 😸), does not have any such demands on processing as what you are doing. Graphics are always gonna push those boundaries, so while the chip manufacturers are gonna always have their hands full, generating faster, smaller, more efficient hardware, eventually, the developers of the programs you are using, will hit a bit of a wall and slow down, to work on those interface / user tweaks. That is when the prices will begin to fall, because remaining on an engineer's dinner table napkin, is the next gen 128 core processor that will be well beyond Germany's Enigma machine, but fit on the head of a pin 🙀 . Speaking of... there are some fascinating movies, documentaries and some videos right here on the Enigma machine. That was an absolutely incredible piece of engineering and is born of the same minds as that guy who will be distracted from dinner with his family, to make drawings on a napkin.

[@agnesslovehealz]

DID easier way

[@shadyendeavor]

D-ID is great, except it's a different product. It doesn't generate audio from an original or specific voice. It doesn't translate audio either. So that already has no comparison to this video. And also, it's super expensive. $29/month in order to get a measly 15 minutes if you don't want the watermark and/or if you need it for commercial use.