Personally, I export in PNG. The advantage is that it is lossless and less heavy than TIFF. Then I recompile with Adobe Premiere pro and Adobe media encoder.
by importing into Adobe premiere pro as an image sequence, it is quite easy afterwards. You replace the sound track and that’s it. For me, an episode of a series of 42 minute 25, with VEAI 1920 x 1080 to 1920 x 1080 with the “Proteus 6-Parameter v1” model, I take +/- 4h00 to improve the sharpness. Then I recompile with Adobe premiere pro and media encoder for export. It takes me between 1h10 and 1h20 +/- for the final render. So it’s still ok. In one day I can do 2 enhancements with VEAI and 3 final renders with media encoder.
Very good, Proteus v2 has completely fixed the burn in artifacts when upscaling to UHD resolutions I had on select sequences with v1.
For best results, I would advise doing it in several passes. The first one is with one of the dione models to de-interlace. Or with the Handbrake software which also allows to deinterlace well with bob filter. Then, for the second pass (if the video has noise and compression artifacts), I would advise using Proteus 6 parameter v1 or Proteus 6 parameter v2 with the settings for noise and artifacts. Then, for the 3rd pass, try Gaia CG to do the 4K upcalle. This model will not degrade the image with built-in parameters such as noise reduction or over-sharpening. And finally, for the 4th pass, use Proteus again with only the sharpening parameters like deblur and sharpen.
All passes must be made in image sequence to preserve image quality. I recommend PNG which is a good compromise between quality and size of allocation on the disk. You will also need a fairly large disk because a 42 minute 25 episode is a 126GB PNG image sequence.
2 Likes
Excellent news, because the improvements of version 2.4.0 for updates is very practical, to avoid being eliminated and reinstalling with each version, so if you can continue to have the preferred models of each one with each version it will be great.
Thank you very much Taylor for your work and disposition
Not quite that easy because I use 8TB SSD for VEAI only. And the max read is around 550mb/s. My machine is too fast, but my SSD is too slow, so it can’t keep up. Then my encoding speed in Media encoder reduced by 50%. That’s for TIFF image sequence.
I don’t know how fast mine is but since I have a 465 GB M2 SSD, I installed VEAI and the templates on my 2nd internal 2TB HDD. That’s enough since my image sequence folder for a 42 minute 25 episode is 126 GB.
So, for you it should go normally
I just looked at my SSD, on the screenshot, these are the data collected from Samsung Magician because my SSD is a Samsung.
And here, those of my 2Tb internal drive too
si pas deja fait je te conseillerais d’acheter un radiateur pour ton SSD ! les M2 et Nvme chauffe beaucoup plus que les ssd en Sata 3, et avec la carte mere qui chauffe pas mal aussi, sans parler de l’été suivant ou tu habites…
(don’t know how to tell this in english, sorry guys,)
Im having a serious issue with colour, brightness and contrast here.
Gaia seems to be increasing the brightness and changing the colours and raising the black levels, and ive had a serious trouble trying to fix it after in editing, so i cant use this release version.
1 Like
4K image sequence has 24mb/pic. Media encoder will be slowed down.
And a massive 1TB image sequence of a 25m video. Yikes.
It happens on all models. Not just Gaia.
How do you add Chronos v2 and Proteus v2 models in v2.3?
FFMPEG windows command prompt command to convert the currently selected folder to a h265 video in a matryoshka container (you need to set the framerate, and the parser seems to have issues with ‘%06’ if you run it from a batch file):
ffmpeg -f image2 -framerate 24 -start_number 0 -i %06d.png -vcodec libx265 -crf 15 -pix_fmt yuv420p test.mkv
Then you use mkvtoolsnix to assemble the final container however you wish (adding audio, subtitles, original video, or whatever).
Less painful than losing an entire render, and much more in your control too.
1 Like
Have you tried using ffmpeg to extract the frames first? I’ve had far less issues with demuxing/decompression weirdness since I started using a .png source.
As I think I’ve mentioned before, if you use lossless png your filesizes will be smaller, so less storage and bandwidth is needed, at the cost of a tiny amount of extra CPU/cache overhead.
1 Like
I don’t use compressed format if I want image sequence. I want as much detail as possible.
I use Media Encoder for encoding. It’s faster.
Try to compare BEFORE-AFTER using a TIFF frame sequence, and I believe you will have no problems, at least with Proteus.
