ok, i change settings to try.
i did not find the 0.20 keepgrain setting : do you speak about EZKeepGrain ? (but it is set to 8.00 on your capture)
or do you talk about the grain you set in tvai ?
kr
fred
my issue related to color (or should i say the black), just after deinterlacing, and before TVAI treatment
Left is the deinterlaced result in FFV1
Right is the original interlaced video in mpeg2
trying to find out what s going on
ok , fixed the part of color/brightness.
now i ll switch back on movies instead of toon, as it was the original purpose
weird, maybe depending on the source the details are the same with the arrangement of colors, because in your example there is just more noise but it doesn’t bring out more details like for my sd videos
Slightly off topic but I found an interesting site where you can compare the results of various Deinterlacers using a visual comparison. They have an image from the Tokyo market. I chose the sign which shows how each method combine frames. For instance the built in Premiere method produces combing. There seem to be some new methods such as MFDIN L which produces a great result. There’s a science behind this, as below. I’ve found recently with QMTGC that I get better results when the filters are Placebo. Those Vaporsynth filters can soften and enhance artefacts which can be exaggerated in VEAI.
2 Likes
Could you share your workflow please? Thanks in advance!
Sure. I start by prepping the source files. If they are super blocky (crappy compression) or are very low contrast, I process them first in hybrid by running a deblock filter and/or tweaking color by increasing contrast a little and reducing brightness a little. The extra contrast helps Iris pick up extra details.
If the file is medium to low quality I use Iris V1 and if it is a good quality source I use the latest Proteus and upscale x4. I use manual settings, get the estimate for settings and then usually increase improve detail to around 70 - 80. I test to make sure the teeth on people dont look crazy and tweak by increasing or reducing until it looks good. I will also increase sharpness quite a bit and increase dehalo a little to compensate for the extra sharpness. Recover original detail set to 100. Lots of trial and error tweaking to get the look you want.
I then use StaxRip to reduce the file to a 1440p or 1080p file using DPID algorithm. If it looks good enough at 1440, cool. If not, I go down to 1080. Then, I bring the file back into Topaz and use Iris V1 or Proteus (whichever looks better) to do another pass without changing resolution. I get the estimated setting again and then set Preserve Details to around 50 - 70 (again checking teeth, etc) and Deblur set to around 70-80 (sometimes lower, depending on how sharp the image is). I leave dehalo, denoise and revert compression at the estimate settings at first, Preserve Original Details at 100 and sharpness set to 0. I then begin to tweak Deblur, Sharpness and Dehalo to get the look I want. I also try a preview with Revert Compression set to 0 and see if it looks better than the recommended. Sometimes it does look better, sometimes not. Depends on the source.
After that, I color grade/correct using Adobe Premiere Pro. Then finally, I do any final tweaks that may be needed (usually some dehalo stuff since I try to get the image as sharp as possible) including MCDegrainSharp or QTGMC to “deshimmer” the picture so it looks cohesive.
If the source is extra low resolution (like a 240p file) I will use SuperResXBR in Staxrip to double the resolution to 480p before first processing it in Topaz.
2 Likes
Thank you very much. I’ll try these on my old videos. All the best
1 Like
To Fredo:
The manual settings you mentioned are for Iris V1 or V2?
From my experience with Iris (both V1 and V2), Recover original details less than 85 results in a horrible image quality. I usually set Recover original to 100 (source 720x480 DVD remux).
From my experience with progressive/telecined DVD/SD upscaling Topaz rarely gives a good result, since many DVDs are way too much oversharpened and denoised either too much or too little. So SD to HD upscale works only when you have an exceptionally good quality SD source (probably 1 out 10 DVDs).
When a DVD is not oversharpened, upscaling with Dione Robust (v4 or v1) to 720p is usually the best (always keep the original pixel rate for DVDs when upscaling in Topaz).
With interlaced DVDs you need to use bwdif (without double framerate) to deinterlace them first and save the deinterlaced video as same as source mkv with FFV1 lossless codec (Shutter Encoder does this in a couple of minutes), since Dione models are trained with bwdif deinterlaced sources.
sorry i missed this post.
unfortunately i did not yet find my best settings, still restarted from scratch, and also gave a shot with vapoursynth and VSGAN…
but still working on… with my sources movies sleeping on my drives…
