I see. Well, I might agree with your creative choice if warmer is what you wanted. However what I see is not warmer, bur rather it is a very unpleasant magenta cast on the faces of actors, that was either not there in the original or it was no where near as strong. If warmer is what you want, you could add warmer tint like more orange or yellow, definitely not magenta. I don’t know how much experience you have with color grading, but I think you made it worse. When it comes to skin tones and color grading, magenta is a big no, no.
Perhaps you can suggest an alternative workflow on the color grading. How do you color grade? I usually use DaVinci and the color tabs.
1 Like
DaVinci is an excellent choice, yes. But I would suggest, not so much a workflow change, its more a matter of experience with color grading, something like feature films or TV shows etc.
Probably the best thing to do would be to leave the colors more or less as the original since it was done by people who knew what they were doing and what they wanted, creatively.
But if you do plan on making your own tweaks, because you see something you don’t like or feel that they didn’t have the technology or budget to fix some smaller issues, some experience with color grading would be required.
In a normal, non sci-fy show skin tones are usually standardized, with proper white balance for the overall scene and if there is discoloration you can use vectorscope and skin tone line.
Its been a while since I’ve seen Star Trek Enterprise TV show, but if my memory serves me, one of the potential problems might be that many humanoid alien spices wear make up or mask so color of skin is all over the place and also because there are lot of colorful lighting set pieces there can be all kinds of color cast coming from the environment, and altering that might change the mood of the scene from what it was indented. That is why I think it might be best to leave it as original.
If you do see technical problems, like color spills or too saturated colors or something like that, again depending on the level of your experience, you could try to select that using qualifier tools and windows in DaVinci and tweak it to be more consistent. But usually this requires someone with experience with color grading to be able to judge the scene and make good decisions. If you are not at that level, I would simply suggest to leave it as original for now, and focus on up scaling the resolution and bringing out the details and all that.
P.S.
In the scene with the sun set and silhouettes on the beach that you show, before and after I see that upscaled version introduced a lot of halos around the tree branches. I would try to minimize that. Otherwise the impression is that it looks worse than original, because the little halo problem in the original , got really amplified in the upscale version. I think Topaz , not sure which model offers de-halo option, so maybe for some scenes it might be worth it.
Update: 4K Video Sample Link blow from the Enterprise-Pilot (about 3 minutes, 520 MB, no password, free to download) - Scene: Enterprise crew assembled and NX-01 launching from space dock.
Link: Dropbox - ENT 1x01;02 - 4K Sample.mkv - Simplify your life
Have fun!
1 Like
Hi all - New to this forum - been using most of the software for a little bit and Joel H has been very kind to supply me some guidance on email. Been going OK but really with my PAL sources being a pain in the butt - looking to compare workflows - mostly StaxRip from DVD Decrypt via DGIndex but setup AVSPmod and AVSynth+ too. I then have full Resolve which I am learning as well. My youtube has some recent encodes I did from the DS9 documentary - MyYTube
Where are people at in terms of StaxRip scripts? I can share mine here if people want to compare, I learn from everywhere 
I’ve found first 3 seasons of DS9 and VOY are a dumpster fire - emissary is terrible in places. Jumped over to test Voyager and ran into a wall with caretaker lol - all fun and games. I definately need to learn about slowing down the audio from PAL next - I can see from CodeGrey’s past posts I am missing steps from my process. That being one of them.
Also how many have spotted errors on the VFX in places - 6 odd mins into DS9 S04E01 you can see this, I’ll need to isolate those out and fix in Photoshop - Who missed this I wonder I know DS9 footage in TNG HD Birthright has a similar issue
Could someone tell me if this staxrip script will be fine for de-interlacing a DVD with QTGMC, I have a AMD Ryzen™ 9 5950X with 12 cores and 32 threads.
SetFilterMTMode("DEFAULT_MT_MODE", MT_MULTI_INSTANCE)
QTGMC(preset="Slower", InputType=0, sourceMatch=3, sharpness=0.2, tr2=2, ediThreads=32)
Prefetch(32)
Thanks
Having seen this thread
https://community.topazlabs.com/t/video-enhance-v2-6-4/27636/575?page=28
I think these settings would be ideal,i am unsure on the number of cores threads i should use for epithreads and prefetch.
This is for a delinterlace and upscale of 80’s dvd concert footage in NTSC
SetFilterMTMode("DEFAULT_MT_MODE", MT_MULTI_INSTANCE)
# can be configured at: Tools > Settings > Source Filters
AssumeTFF()
QTGMC(preset="Very Slow", InputType=0, EZKeepGrain=0.5, sourceMatch=2, MatchPreset="Very Slow", Lossless=2, sharpness=0.1, tr2=0, Sbb=0, ediThreads=8)
Prefetch(12)
should be ok!
I’m very Interested in knowing your work with Stargate SG1…
Its been a while for this topic, but I did finally manage to get a settings setup in the new V3 Proteus that was both faster and reasonably good in comparison to the older V2 so thought I would post a clip as a reference:
Total episode processing time was almost under 10x hours, though it probably could be if you didn’t have a system like mine that is around 6x years old at this point.
1 Like
Interesting, care to share specifics?
I extracted the image sequence (NTSC Episode) using the Svisynth script:
Note that I am using a method to avoid any issues with variable frame rates, but it produces a 59.94 sequence. That part is mostly optional and given most of Voyager seems to not suffer much with this issue, can probably straight TDecimate it.
I also left in some of the scripts I had been experimenting with but commented out (not used above).
SetFilterMTMode("DEFAULT_MT_MODE", 3)
D2VSource("C:\Users\Vaengence\Downloads\Episode 1\VTS_03_1.d2v")ConvertToYV12()
checkmate(tthr2=0)
Bifrost(interlaced=true)
A=Tfm(field=1,mode=0,slow=2,pp=2,mchroma=false,cthresh=-1,micmatching=0).converttorgb().generalconvolution(matrix = "0 -1 0 0 4 0 0 -1 0",divisor=2,auto=false).converttoyv12()
B=Tfm(field=0,mode=0,slow=2,pp=2,mchroma=false,cthresh=-1,micmatching=0).converttorgb().generalconvolution(matrix = "0 -1 0 0 4 0 0 -1 0",divisor=2,auto=false).converttoyv12()
C=Tfm(field=1,mode=0,slow=2,mchroma=false,cthresh=-1,clip2=A,d2v="C:\Users\Vaengence\Downloads\Episode 1\VTS_03_1.d2v",flags=1,micmatching=0)
D=Tfm(field=0,mode=0,slow=2,mchroma=false,cthresh=-1,clip2=B,d2v="C:\Users\Vaengence\Downloads\Episode 1\VTS_03_1.d2v",flags=1,micmatching=0)
interleave(C,D)
#SMDegrain(tr=12,thSAD=500,contrasharp=0,str=1.2,refinemotion=true,lsb=true,interlaced=false,Globals=2, chroma = false, plane = 0)
#TemporalDegrain2()
#####Deblock(quant=25, aOffset=0, bOffset=0, planes="yuv")
GradFun3()
EdgeCleaner(strength=10, rep=true, rmode=17, smode=0, hot=false)
LimitedSharpenFaster(ss_x=1.00, ss_y=1.00, strength=20, overshoot=0, undershoot=0, soft=0, edgemode=0)
Crop(6, 6, 708, 468)
xaa(ow = 720, oh = 540, mask = 1, chroma = 1, nns = 4)
Prefetch(6)
That image sequence then fed into the V3 Proteus on Manual settings: 25, 55, 35, 15, 30, 14 and upscaled 2x.
Then I did some post processing to recombine it in Premiere, including running the Neat video noise removal, and color correction (not that this would be obvious in the B&W sections).
Its not perfect - I am still not entirely convinced with the new Proteus yet - but its the best I have worked out so far using the latest versions of Topaz.
2 Likes
So a few more notes.
So far, everything I have touched with QTGMC results in worse upscales - I know for some of the content, particularly the PAL DVD’s it may be unavoidable, but I have rarely achieved good results with it. Its why I have tried to avoid it if I can.
Given most of Voyager can be essentially straight IVTC with no real variable framerates (except in specific episdoes) I have experimenting with dropping back to a more standard IVTC wihout frame doubling - if for no other reason than to shrink the image sequences, save space and more than half the processing time.
However, I have been encountering anomalies part way through season 5 - the IVTC process is leaving interlaces frames which TFM simply doesn’t seem to fix - eitehr by finding the right frames or post processing deinterlacing. Given its on the CGI sections so far I see it in, I am guessing its a byproduct of the conversion they used for the CGI in Voyager to not be variable.
Either way, the simplest way of dealing with it I found was using AnimeIVTC in Avisynth. This seems to pick up on the interlaces sections and deinterlace it fairly well whilst leaving the rest IVTC properly.
So current script is this:
SetFilterMTMode("DEFAULT_MT_MODE", 3)
D2VSource("D:\Video\NTSC Season 5\Disk 3\Episode 11\VTS_03_1.d2v")ConvertToYV12()
checkmate(tthr2=0)
Bifrost(interlaced=true)
AnimeIVTC(mode = 1)
GradFun3()
EdgeCleaner(strength=10, rep=true, rmode=17, smode=0, hot=false)
LimitedSharpenFaster(ss_x=1.00, ss_y=1.00, strength=20, overshoot=0, undershoot=0, soft=0, edgemode=0)
Crop(6, 6, 708, 468)
xaa(ow = 720, oh = 540, mask = 1, chroma = 1, nns = 4)
Prefetch(4)
Lastly, with tweaking on settings, using Proteus V1 with 20, 55, 20, 25, 10, 50 - no grain.
Proteus V1 requires using V2 of Topaz and this is intentional - there is a bug in the noise reduction in the current versions and going back for a few now. Unsure if it is V3 itself of the Proteus model as went to V3 from memory before they removed the numbers. Either way, the noise reduction is completely different, so the settings above won’t get the same result, and you will periodically have shimmering that doesn’t exist in V1 Proteus.
I am waiting till thats resolved so I can make used of the faster speeds for testing, but until then, the above script output takes 25 min for image sequence, upscale is 3.5 hours, and then assembly, some additional noise reduction in Premiere and colour balancing exports is about 2-2.5 hours, making the whole process about 7x hours ish. Using my 5-6x year old PC with a 1080.
The link below is an extract from my last episode completed. Its certainly not perfect - still trying to reach that mythical goal, but for a 7x hour process that can be sped up with the newer version (once issues resolved) it at least puts the upscale of the whole series in a more realistic goal of achievement for a single person.
Don’t watch this clip via the google drive, it tends to re-encode it. Need to download it.
Also, the sharpening and edge cleaning was left over from me using Artemis - its possible I can remove them and experiment further given Proteus can manually do a lot of this now, but haven’t tested yet.
1 Like
Hi, what is the original resolution? is it possible to see the same passage of the source video? for proteus, you use v1 of version 2, like 2.6.4? the problem is that there is no relative to auto, so it’s not processed frame by frame differently but all the frames with the same parameters, so I imagine that you make proteus settings according to the scenes? for the quality of your video, the upscale is well done I find but still this plastic effect when you go from low resolution to fhd or 4k, I like to add grain to hide the few imperfections it can stay or even hide the plastic effect, adding grain, it also gives a little sharpness depending on the size of the grain.
I am not sure if you know how to read the Avisynth scripts, but it just so happened the first question you asked is stated in that script.
But to be specific, I start with the original DVD resolution (NTSC) which 720x480, then crop it to remove the edges that look terrible, then I minor upscale it to 540 within Avisynth before loading it into Topaz. This is because I found historically that doing a straight 2x in Topaz (input starting at 540 and ending 1080) seemed to yield better results than leaving it native and then doing a 2.x upscale to get it to “1080”.
I can upload the same section as source as reference, but I extract in images and I don’t usually reassemble the source before it goes into Topaz so I will have to create that first. I should be able to do that in the next day. Image comparisons however are straightforward as they are both images, so this is “an” example to give a reference for now:
Note however, that since my last post, I was experimenting with removing some of the additional -preprocessing (edge cleaner and sharpening) to see if Proteus can take up the slack and speed up even more, so this comparison is from that run. That run had Proteus using 20, 55, 20, 45, 10, 50.
In relation to setting Proteus for scenes, my goal from beginning was to simplify the process to the minimum amount of time spent to get effective results. There are 172 Episodes of Voyager alone, not to mention DS9 or anything else to be done, and you have to do them all alone due to copywright, so self processing 172 episodes with half a day each is already 2-3x months and thats assuming you can run it constantly (which I cannot) so its more like 3-4 episodes a week maximum. More realistic timeframe is over 12x months.
So the quicker and simpler the better. Trying to adjust the settings to match scenes is the antithesis of that idea - sure I can spend more on specific scenes, but then do I finish in 2-3 years instead? I have in the past treated the CGI separately, but this current method does enough with it to be acceptable and saves time working on it. So I generally am trying to find settings that will be acceptable for the whole episode.
I have also added grain the past - actually for probably close to 2x years. Then I realised that depending on the settings and output, it looked better to me without the grain, so I have generally avoided it since, but it is personal opinion in that regard so up to you.
The slightly plastic effect comes from the noise reductions - and a lot of the footage is very noisy, so removing it has side effects. Its very hard for any algorithm to make proper distinctions between noise and finer details when the object is moving, so you tend to get less detail on movement and much better detail when still - IE for example the comparison shot above.
I will say, though that avoiding that plastic look is one of my goals and will tend to find it is a lot less plastic than many other examples of the upscales I have seen. I am not aiming for perfection, I just want additional clarity, with better quality footage with a goal to make it wherever possible that someone watching it without realising what you have done, doesn’t pick up that the video “looks weird” or “what did you do to their faces?” etc. Its not perfect, may never be, but I still keep trying just in case a miracle happens ^^
ok, i didn’t pay attention to the resolution, i can’t read scripts i understand your method, for the noise did you try the keepgrain option with hybrid? I don’t really know why but the higher you go, the more detail there is in the video, but the more the grain reappears. to gauge according to the source, at the moment i’m processing two old clips, and i noticed that using gaia hq or gaia cg before proteus, it’s not bad gaia cg slightly reduces artifacts and pushes some details, gaia hq reduces artifacts a bit more and gives a bit more detail, but some fine details are overemphasized. again, it depends on the source. I stopped working on my sd videos two weeks ago, I calm down because I spend too much time on them, I aim for perfection so much that I’ve been working on quality for 7 years, my goal is more ambitious , I want no more artifacts and no loss of detail, or so minimal! haha it’s because it’s family videos from 1996 to 2002 so it’s more important than the two clips I put myself on. all that to say that when you care about the videos you work on, you have to devote time to them.
You may need to clarify what you mean by “keepgrain” as none of the processes I use have a function named that - on the contrary the specific pre-processes I use try to avoid removing much grain/noise intentionally.
Gaia CG and HQ can be good, but I don’t use them often. The main issues are that Gaia CG broke on version 1.6 of Topaz - if you want to use it effectively, you have to go all the way back to 1.5.3. On current versions, its virtually pointless to use.
And yes, I know what you mean by trying to make them perfect - especially with personal videos which you can probably spend significantly more time on - just with Voyager for example, far too many episodes and a lot less personal to make it perfect ^^ Around 3x years now playing with the software.
the keepgrain option in qtcgmc of hybrid it gives a lot of details so it blurs a lot without adding grain if you don’t push it too high. for me it’s the opposite, gaia cg and hq are quite fast with 2.6.4 as long as I don’t upscale (720p it works) beyond that it becomes slow, but with the latest alpha, gaia is much faster and he does a good job
Ahh ok - I don’t currently use QTGMC. None of these episodes are interlaced, which is what that filter is used for. But thats specific to the NTSC DVD’s - the PAL ones I am pretty sure are, so you would have to handle them differently.
There are some small sections, usually only around 10 frames max, where the CG sections have been handled badly and/or encoding to DVD makes so IVTCing that section misses some interlaced frames - but they are rare and usually make up less than 0.01% of the file images meaning using QTGMC to deinterlace them is overkill.
I did experiment with handselecting just those frames to QTGMC - and it works and if I was a stickler for perfection this is probably the best option, but it takes a lot more effort to find and remove those frames than having Avisynth do it for you - that was why the AnimeIVTC was a good option. It finds those automatically and deinterlaces them, saving me having to manually do those leftover frames.
ok, but you can use qtgmc without deinterlacing! I sometimes use it for the keepgrain option precisely to gain sharpness
