airduster_9000

I assume its a controlled roll-out so they can evaluate any misuse and issues with a smaller sample size first.

Ok but the real question is: is this mix (or set) available somewhere? I need it

Here you go :

Breaking Bad.

AMC/Sony created Breaking Bad - Netflix just license it

I would assume just as many - if not more are using Microsoft Azure OpenAI services. Especially in EU due to GDPR.

You're on the dot. Both have their positives and negatives. Alternatively I would say "do you like messing around with modules and connections to get something you combined?" Go comfy. "Do you want to quickly make something, with no experience at all?" Automatic1111.

Would you agree with this I made for a work related project:

Pretty good swap - but it looks a bit like a mix of the two. These new faster ways to do deepfake video often seem to combine the two people as a lot of generalization is going on to be able to faceswap with as little as 1 source image.

Så længe vi ønsker at cigaretter og alkohol er lovligt - er der ikke et intelligent argument der gør vi skulle behandle hash anderledes. Cigaretter og alkohol er begge mere skadelige og dyrere for samfundet.

This game looks insane on a PS5. I bet it looks ridiculous on a 4090

Its hard not to stop and do "photos" all the time. The cut-scenes were also really well done with natural acting, camera angles changing often and movement.

The sweet spot I found for me is 0.75 weight on SD1.5 and 0.65 weight on SDXL. One or two specific sources will require me to bump that up or down, and I tend to start much lower for art styles.

0.7 - 0.8 for being able to do art/style transfer - and atleast for InstantID you then go 0.9 - 1.0 to get precise photos with the person/face.

So... Are these efficiency gains or lay offs to reduce labor expenditures?

I would love to do the South African gangster shrimps situation before going all in on space fascism.

Not sure what UI you are using, but for face swaps/character consistency, I generally begin with a weak IPAdapter plus face to get the general head and face shape, then pass that through face ID, and will do a final step using reactor (depending on the facial expression needed in the output vs the reference image-- reactor isn't great for much other than happy or neutral expressions in my experience, your results may vary).

I have a similar setup - where I use either IP-Adapter or InstantID to create a person 90-95% identical - and then if needed use ReActor to do a faceswap as well.

can you make a 4k version for the 16/9 guys like myself that would be soo cool if u can

Upscaled

I would watch this 😅

Well it’s a real show made in the UK

If you want a model to produce images of a specific product I think its easier to train a model for each product using Dreambooth - than trying to get one model to do all the products. But its a common question on this sub-reddit, and there is also a growing field of websites/apps who does it all for you.

I've seen this before but it's always cool to see someone's dreams come true in real time. Like that video of Pharrell listening to the Maggie Rogers track.

The video Maggie

Looks like 80s inspired steel frame design. Search for that and you get hits for similar

…. How??? SÅ MANGE SANGE JEG ØNSKER AT PUTTE I DERES KORREKTE GENRE AAAAAARRRGHH!

De laves med

Hvem synger?

En AI laver både musik og sang - baseret på teksten jeg har givet den.

Er sangen også AI-genereret? Den lyder som nordmænd eller svenskere der synger på dansk. Egentlig meget fedt.

Ja det er 100% AI udover sang-teksten. Ofte ser du at de modeller der er trænet blander især svensk ind i både dansk og norsk - da vores sprogområde ikke har fokus endnu.

Reminds me of the music video to Star Guitar by The Chemical Brothers: