The character's facial proportions change in the animation, which is a complete no-go. Programmers wont care, sure, probably marketing neither, but this is really what separates artists from people who get duped by these tools. It doesn't look good. To a trained eye it looks like shit.
You can argue it will get better, sure, but you're showing what it currently is, and that's all it can really be judged on.
This isn't moving the goalpost - It's impressive it can do this. Its just not good enough for me to use it, and that goalpost has been there a while.
In case anyone was wondering what mild hallucinations are like when taking LSD or mushrooms, that last video with the landscape is a realistic simulation.
Comments
The character's facial proportions change in the animation, which is a complete no-go. Programmers wont care, sure, probably marketing neither, but this is really what separates artists from people who get duped by these tools. It doesn't look good. To a trained eye it looks like shit.
You can argue it will get better, sure, but you're showing what it currently is, and that's all it can really be judged on.
This isn't moving the goalpost - It's impressive it can do this. Its just not good enough for me to use it, and that goalpost has been there a while.
The context is previous versions of stable diffusion animations, which were significantly worse.
In case anyone was wondering what mild hallucinations are like when taking LSD or mushrooms, that last video with the landscape is a realistic simulation.
Is anyone running these models on consumer grade hardware?
How do these compare technically or philosophically to NERFs?
https://github.com/continue-revolution/sd-webui-animatediff