You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When I tried InfiniteZoom together with upscaling, I found that it upscales every each frame of the video, which is unnecessary in this case. It would be enough to upscale only the main generated frames and do the zooming on the upscaled ones. That would save a lot of time.
The implementation seemed simple to me, so I wanted to take it on and code this feature directly. However, I found out that there is already the optimize_upscale branch by @GeorgLegato that does exactly this.
I want to ask, what does the future look like with this feature? If that branch isn't in the development, I'd be happy to take it on, I just find the modifications unnecessarily complicated. My idea is more simple:
after generating the images, upscale all of them.
skip/disable upscaling everywhere else
The text was updated successfully, but these errors were encountered:
check please the branch overmask_...
beside to smarter upscaling it needs to write upscale/interpolated frames as a stream to video.
older approach is to keep al in ram which is insane for 2 mins/60fps/4k
in the overmask_ branch use the "corner" outpainting strategy, the Center one is not frame corrected but could still work
Corners will extend the target size by your given outpainting width, then zoom in.
When I tried InfiniteZoom together with upscaling, I found that it upscales every each frame of the video, which is unnecessary in this case. It would be enough to upscale only the main generated frames and do the zooming on the upscaled ones. That would save a lot of time.
The implementation seemed simple to me, so I wanted to take it on and code this feature directly. However, I found out that there is already the
optimize_upscale
branch by @GeorgLegato that does exactly this.I want to ask, what does the future look like with this feature? If that branch isn't in the development, I'd be happy to take it on, I just find the modifications unnecessarily complicated. My idea is more simple:
The text was updated successfully, but these errors were encountered: