Stable diffusion thread

[doogyhatts]

Hunyuan Custom is a multi-modal video model, focused on subject references.
https://hunyuancustom.github.io/

It is launching today!

 

[doogyhatts]

 

[doogyhatts]

Music video.

 

[doogyhatts]

 

[focus1974]

Music video.

this one very well done!

 

[doogyhatts]

@AikiBoy
Got other channels like melodyisheremusic and sherryishere4yt.
Both channels are from the same creator.
His "melody" channel is monetised, but I suspect his "sherry" one could not.

 

[AikiBoy]

@AikiBoy
Got other channels like melodyisheremusic and sherryishere4yt.
Both channels are from the same creator.
His "melody" channel is monetised, but I suspect his "sherry" one could not.

Lol... what is advance AI. Those video all same girl n randomly sitch together... no story telling..

 

[AikiBoy]

Made another one.

 

[doogyhatts]

Lol... what is advance AI. Those video all same girl n randomly sitch together... no story telling..

You see his "melody" channel.
He release one video every day, with a duration between 2-3 mins, with a higher 1440p resolution.
He did not reuse video clips within the same video.

The characters have specific poses and their fingers remained mushy sometimes.
The clothing is not fully consistent throughout the video.
But still he can monetise the channel within 2 weeks and the retention of his videos is likely to be very high.

He might be paying Hailuo AI for the unlimited plan and all his images were prepared beforehand, focusing on faces.
Then everyday just generate 200-300 video clips using their Minimax's Subject Reference model.
This would be the easiest solution, but it also deters a lot of other creators from doing so.

 

[AikiBoy]

You see his "melody" channel.
He release one video every day, with a duration between 2-3 mins, with a higher 1440p resolution.
He did not reuse video clips within the same video.

The characters have specific poses and their fingers remained mushy sometimes.
The clothing is not fully consistent throughout the video.
But still he can monetise the channel within 2 weeks and the retention of his videos is likely to be very high.

He might be paying Hailuo AI for the unlimited plan and all his images were prepared beforehand, focusing on faces.
Then everyday just generate 200-300 video clips using their Minimax's Subject Reference model.
This would be the easiest solution, but it also deters a lot of other creators from doing so.

lol some face also not the same leh. warp after the ai video.

Plus i think ppl go there for visual comfy lol

 

[AikiBoy]

I think unless you go for hot chicks. I think hard to grow

 

[doogyhatts]

I think unless you go for hot chicks. I think hard to grow

Even if it is hot chicks, we still cannot use certain lewd clothing or do provocative poses, or even use camera focus on certain areas.

That's why the Sherry channel never got monetised, he had to create the Melody channel, and then shift the audience over to it.
But eventually, the new channel boosting will end, and his Melody channel will have lower impressions and lesser views.

 

[sick_and_tired]

What hocus pocus is this...

 

[AikiBoy]

Even if it is hot chicks, we still cannot use certain lewd clothing or do provocative poses, or even use camera focus on certain areas.

That's why the Sherry channel never got monetised, he had to create the Melody channel, and then shift the audience over to it.
But eventually, the new channel boosting will end, and his Melody channel will have lower impressions and lesser views.

So far i only got 1 subscriber for my Ai channel..

 

[doogyhatts]

So far i only got 1 subscriber for my Ai channel..

I think it is easier to do animals.

 

[doogyhatts]

ByteDance has a new image model that has very realistic lighting, but the background is blurry.

 

[doogyhatts]

@AikiBoy
You can use Dreamina to generate images too.
Comfy can be used for add detail, outpainting and upscaling instead.

 

[AikiBoy]

@AikiBoy
You can use Dreamina to generate images too.
Comfy can be used for add detail, outpainting and upscaling instead.

Too many... too little time to test

 

[doogyhatts]

 

[AikiBoy]

Japanese still like anime style