This forum uses cookies
This forum makes use of cookies to store your login information if you are registered, and your last visit if you are not. Cookies are small text documents stored on your computer; the cookies set by this forum can only be used on this website and pose no security risk. Cookies on this forum also track the specific topics you have read and when you last read them. Please confirm whether you accept or reject these cookies being set.

A cookie will be stored in your browser regardless of choice to prevent you being asked this question again. You will be able to change your cookie settings at any time using the link in the footer.

[BUG] Deep Remaster and Deep Reference modes both crash; ColorMnet reference mode is OK
#9
(27.12.2025, 17:41)Ricks Film Restoration Wrote: Thank you very much. Having to delete the Vapoursynth folder before installing the colorizer stuff is not described in the HAVC User Guide PDF.
Instructions how and what to do need to be improved. I am not an amateur but ICT professional.
Scrolling through the comments it seems not many people are using the video colorizer modes. There is hardly any useful debate about it.
I hope somebody will develop better video tools on the basis of AI diffusion technology (alike Topaz Starlight/ Astra).

Below are two files that show my reference frame versus what the Vapoursynth colorizer produces (one frame difference): disappointing.

Your reference frames seems to be produced using a DiT model. But keep in mind that during the colorization process will be kept the resolution and grain of the original B&W frame, not the one of the reference frame. I recently started to do some experiment using the DiT models as described in my recent post: #5

I know that ColorMNet does not apply perfectly the colors, to obtain this result you have to check the box Vivid and provide very frequent reference frames (at least a key frame every 20 frames) as reported in this post: #1.268. ColorMNet does not store the full reference frames, but in order to reduce the memory usage it keep only the most relevant pixels, this is by design and I cannot change it. But providing a key reference frame every 20 for me is not practical. Recently I colorized using a DiT model a film with 138687 frames, providing a colored reference frame every 20, imply to colorize about 6934 frames. Given that to colorize a frame with a DiT model in my case are necessary 12sec, the colorization of 6934 will require about 23h, too much for me. Moreover even the frames colored with DiT models are affected by color instability, using ColorMNet with Vivid unchecked (set to False) will allow to mitigate this problem, since ColorMNet will try to keep the color consistency across the frames and the time. So if you have a lot of reference frames with consistent and stable colors, you can use Deep-Exemplar or Deep-Remaster, otherwise the only available choice is ColorMNet (try to color a clip with and without Vivid to understand better the difference).

Dan
Reply


Messages In This Thread
RE: Deep Remaster and Deep Reference modes both crash; ColorMnet reference mode is OK - by Dan64 - 27.12.2025, 22:06

Forum Jump:


Users browsing this thread: 1 Guest(s)