2 INSANE AI TOOLS: X-Portrait 2 + Detail Daemon

[@MikevomMars]

OMG! It animates long hairs! It also animates shoulders! You can turn the head! All those things LivePortrait could not do 😱

[@OlivioSarikas]

Yes, treally stunning. This actually feels like a real animation in the best cases

[@darrenlewis8273]

@@OlivioSarikas How do we access it please>

[@hybridroid]

Man, 1 year from now, imagine what will be possible... this already goes beyond my expectations

[@pedxing]

looks like Bytedance has their hands on X-Portrait. coming soon to a paid subscription near you. ;-)

[@ielohim2423]

Its supposed to be open source.

[@-kuler882]

@@ielohim2423 Why do you think it will be open source?

[@zacharyshort384]

@@ielohim2423 OpenAI was supposed to be open source too :P

[@hotlineoperator]

This is great, but the limitation is that only one character can be controlled at a time. Making two different characters have dialogue in the same frame is still a challenge, not to mention other interactions.

[@darrenlewis8273]

Not at all just use your common sense. You crate one image with both characters in them as a still image. Take that image into photoshop and split it into two image vertically into two equal size images. Animate them separately then sandwich the video back together by placing one next to the other in any video editing software of yoru choice. its not hard lol

[@MarcoCholo-iz9js]

Hollywood's days are numbered when this tech takes off. Imagine the level playing field toward the end of this decade. Not just by indie groups with small budgets but even smaller nations with extremely talented people

[@ammokha]

How on earth will "extremely talented people" manifest their talent when there will be nothing for them to do other than pressing some buttons and writing some prompts?

[@niroknox]

When will this be available for ComfyUI? Is there anything on this level for Comfy?

[@amerikyuss]

Need xportrait 2 now :D

[@AdvancExplorer]

Is that the "denoise output" or the "output" to connect to the VAE ? Thx

[@theartificialrealms]

I don't understand why a tutorial would not explain the node connections. I have never seen "sampler" and have no idea what node I need to add to connect this.

[@blasteekmag]

Hi how do we access something similar available now?

[@wildone106]

Nodes nodes nodes! God I loveeeeeeeee nodes /S

[@joechip4822]

In the design of electronic circuits there exist so called 'auto routers' which place every part at an optimal place and avoids the chaos of spaghetti routing, like it is permanently present in most Comfy workflows. The Comfy guys need to implement something like this urgently. I won't use Comfy as long as they avoid this.

[@murphylanga]

Cool, now you've checked it out 👌

[@SchinTeth]

I have a problem, when i open the example workflow it shows an empty canvas

[@jamesbadman2279]

Me Too, I'm on ComfyUI v0.2.7-10-g8b90e50

[@archael18]

How much time does the detail daemon add to the generation? And does it work with schnell?

[@TerminatorSAW2k]

Question: Could i use X-Portrait 2 4 Live Realtime Avatars in Obs Studio als Overlay or Plugin? When yes with which other Tools and Plugins?

[@TerminatorSAW2k]

= virtual Vtuber

[@TerminatorSAW2k]

could u make please a V Touber tutorial please 4 Obs Studio? :)

[@onlineispections]

Hi. Where can I find the link to install x-portrait 2 on local with workflow?

[@Airbender131090]

Its not opensource yet

Woaw great !!! I really need to test Flux. I have a Geforce RTX 4070 12 Gb, which Flux model I need to use with Comfy UI ? The original one or an other based on the original ? Which Flux model do you use the most ? Thx

[@torosmedia007]

Hi Buddy, thanks for the excellent videos, i just got a new hardware and i will love to know what is the best ai photo generator at the moment, photos for websites, can you point me the right direction please. Thanks in advanced!!!!

[@OlivioSarikas]

The Flux Model is the most advanced model right now and there are a lot of loras for it. But SDXL is still also very good if you look at the community trained models

[@CoconutPete]

too many tools to keep up with 😂

[@OlivioSarikas]

Enjoy it until everything is owned by Adobe 😅

[@FusionDeveloper]

Cool.

[@vecterror]

Non of that Workflows works for me ? Cannot open any workflow in Comfy ... hmmm

[@OlivioSarikas]

Go to manager and click update all. If that doesn't solve it, please ask in my Discord

[@sternkrieger1950]

Detail Daemon is promising, but it changes the composition of the image too much even at 0.1, unlike Add Details for SD1.5/XL.

[@OlivioSarikas]

I would say in my test images it stayed pretty much the same. But yes, the times of easily adding details are over in Flux. It's a new kind of beast. But it can also do a lot more

[@huschangetemadi-zanganeh7247]

Maybe you can use a Depth Map Controlnet to keep the original composition and add details on top?

[@ritpop]

Quick idea, maybe won't work, load the image and encode to latent then use it with a lower denoting value.

[@dreamzdziner8484]

Facial mocap tools are going to be obsolete in a few months. This is freaking insane.

[@ARNABOSS]

16 gb vram is now the minimum requirement for desktop computer 😅😢

[@amykpop1]

no it's no lol

[@ANAN-iy2jn]

how can I use X-Portrait 2 free?

[@MarkBowenPiano]

The X-Portrait videos look okay albeit they're moving way too fast and look like they've all taken some pills or something but their eyes all look emotionless almost like the eyes of a dead person plopped into the video. I don't think I've seen one video yet where it's not given away by the eyes. The most difficult thing to get right on a person.

[@Darkwing8707]

I assume that some are fast in order to emphasize how well it can keep up with motion.

[@joechip4822]

​@@Darkwing8707 And it reflects the attention span of the average KZbin watcher 😅

[@TeddyLeppard]

With stuff like this coming out now, it's no wonder people like Ridley Scott and James Cameron are suddenly very interested in using generative video like this.

[@OlivioSarikas]

Yep, the big studios are very interested in all of this. They can't talk public about it, because they fear backlash, but they are all doing research :)

[@pong4289pan]

The workflow is best for 1024x1024 pixels; others result in poor image quality.

[@AnAntidisestablishmentarianist]

The people on the X-Portrait 2 page look like they're having a meltdown after the election 🤣

[@CoconutPete]

lol I gotta check that out

[@NLPprompter]

no code yet... you can also add fight fight fight meltdown LOL

[@peacetoall1858]

dishonesty factor - 🤣🤣🤣

[@OlivioSarikas]

what do you mean by that? is this aimed at me?

[@peacetoall1858]

@@OlivioSarikas It's one of the options in the Lying Sigma Sampler @9:05 in your video. I found the name funny.

[@xcom9648]

Looks like the detail daemon is changing to much if your trying to make consistent character.

[@acerzwolf4780]

Just use the PAG node for detail, same effect just 100x easier

[@PixelsVerwisselaar]

You oke?😅 btw?😅😅

[@Zerod-rn3ye]

I normally like your videos but you go over X-Portrait 2 while never once mentioning Live Portrait and how one might compare to the other. Live Portrait already does this and based on your presentation X-Portrait 2 is kind of a potentially irrelevant technology for something that already exists. In addition, this should have been covered at the very start of your video as that is going to be the #1 question of most of your target audience: How does X-Portrait 2 improve, or not, over Live Portrait?

[@ritpop]

I think is very clear how x portrait is doing better than live portrait , but tbh is very unlikely that it will be open source

[@Darkwing8707]

​@@ritpopThey released the code for the original X-Portrait. Why do you assume they wouldn't for the 2nd version?

[@Darkwing8707]

They can't really be compared yet because the code for X-Portrait 2 hasn't been released yet. But from the examples, it's looking really good.

[@-The-Darkside]

@@Darkwing8707 he's not assuming they aren't, you're assuming they will. Assumptions...

[@gnarlymode]

X portrait looks better than live portrait imo

[@respectvetman]

Ok

Would you have a $10 free invitation to replicate, please? Tks!!!

[@eiermann1952]

so shocked, imagine what this means for tiktok. US will regret to have banned it : )