Open cyrilzakka opened 4 years ago
The method is trained to operate at a resolution of 1024 * x
or x * 1024
where x
< 1024 (your input image automatically gets rescaled). There may also be some added blur due to resampling, which should be difficult to notice though. So yes, the fidelity of the input image is currently not truthfully preserved.
Are there any plans to add support for higher resolution?
Loving the paper and its implementation. The results really do look stunning! I'm just having an issue with the output resolution being severely degraded in comparison to the input image. Everything seems so blurry. Is this just an artifact of the animation process itself?