Color conversion

[salvo00786]

Hi... I have a new question for you.

First of all, After deinterlacing a file, is better to convert it from bt601 to bt709 before or after using premiere? Premiere can only read from Bt709 and if I import a bt601 file, I can see a color change during the work with Premiere. Premiere can't have an output of bt601.

What is the best tool for convert from bt601 to b709. I can choose between Clever FFmpeg gui, vidcoder or Staxrip using HDRMatrix pluginm with the code: Matrix(From=601, To=709, Bitdepth=8).

Do you have other suggestions? Thanks

[poisondeathray]

Usually as early as possible, and especially if you're doing work in premiere - because if you add graphics , titles, color work etc.. those additions will be shifted to something incorrect if you do it afterwards

Another option is to use a LUT to change 601 to 709 (or vice versa) within premiere
https://forum.videohelp.com/threads/406379-Merging-original-PRORES-and-converted-to-h2...em#post2662557

[salvo00786]

Ok, thanks. My file is a MOV PRORES DEINTERLACED DVD with STAXRIP and QTGMC. In Mediainfo I don't have information about the colours. In Premiere I can import these Luts with color correction, lumetri correction and basic correction, right, or is better if I use Magic Bullet Colorista?

[salvo00786]

I have seen that i can apply these lut with Ripbot, create an mp4 file and after that I can import it into Premiere.

[salvo00786]

The Ripbot conversion is good but is tagged as bt601 and if I import these files in Premiere, there is another colour change. How can I tag bt601 as bt709 for using these Lut with Ribpot, or is better to use directly these luts in Premiere?

[poisondeathray]

If staxrip does not tag the prores correctly, you can use ffmpeg to write the prores metadata color information

https://ffmpeg.org/ffmpeg-bitstream-filters.html#prores_005fmetadata

ffmpeg -i input.mov -c copy -bsf:v prores_metadata=matrix_coefficients=smpte170m output.mov

From a DVD source, there should be no colormatrix filter change or LUT applied (those alter the actual colors ). It should be actually 601 , and tagged 601 (smpte170m has the same values as 601) . If premiere mishandles it, then you can use a LUT in premiere

For the tagging, you can do it with the x264 and x265 encoders directly, not sure about ripbot or staxrip , but they should have those options, or you might be able to add it if they have additional commandline options

[salvo00786]

Poisondeathray, Premiere open the bt601 files, but in the output you can only export to a bt709 file. For this reason, the new file will take the bt709 tags, with the bt601 altered (from Premiere) colours. For this reason, with these LUTS, i can change the colours before exporting to bt709. The result will be a BT709 file, with BT709 colours. After that I want to upscale this file.

[salvo00786]

If my source is limited range, is better to convert to full range or is better to not touch that setting?

[poisondeathray]

The result will be a BT709 file, with BT709 colours. After that I want to upscale this file.

If my source is limited range, is better to convert to full range or is better to not touch that setting?

601 to 709 color shift is ok for upscaled SD to HD... this is actually desired

Usually leave it as limited range if your final export is YUV...but there maybe some other details or issues you might be leaving out...

For example some "upscalers" work in RGB eg. VEIA , and if you use a limited range YUV to full range RGB conversion in VEIA, you will lose some data on some sources. Limited range sources (like DVD) can have details that get clipped in the RGB conversion - it depends on the specific source and what you're doing afterwards

[salvo00786]

Ok, thanks and how can I convert from yuv limited range to rgb full range? It's possible in Premiere?

[poisondeathray]

Ok, thanks and how can I convert from yuv limited range to rgb full range? It's possible in Premiere?

Not directly in premiere ; unless the file is flagged as full range. But you can use levels filter to get similar results. The levels method is not necessarily as good, because you're "squishing the Y range" from Y 0-255 to 16-235 - this can introduce banding artifacts and loss of data as well. Whereas a full range YUV to full range RGB "maps" Y 0-255 to RGB 0-255. The later "proper" method is not "squished", it's "1 to 1" so each "slot" is represented fully. In contrast, a standard YUV range conversion to full range RGB "maps" Y 16-235 to RGB 0-255 . So you lose values <16, >235. Normally that is not necessarly that bad, because most sources shouldn't have that much usable data in those ranges. But some do, and those are clipped in a standard RGB conversion. So this workaround is only necessary if you are certain you have usable details in those ranges (Y <16 , Y>235 for 8bit formats)

For a standard range source with usable "overbrights/overdarks" - this full range conversion will temporarily look washed out (low contrast) but the goal is to preserve the shadow and highlight details that would otherwise be clipped by the standard RGB conversion. So "best practices" for those types of sources is to use full range conversion to something like an image sequence => whatever the RGB process is => full range conversion back to YUV for your end delivery format , so the contrast "looks" essentially the same as the original (because "black" is back at Y=16, "white" at Y=235, ie. it's still limited range, but you preserve the Y<16 and Y>235 data)

[salvo00786]

Ok, In fact after Premiere I want to use Topaz Video Enhance AI. Is good to upscale with Topaz after a Premiere output of a MOV PRORES 4444 XQ. Or I have to convert from limited range YUV to limited range RGB. Because of the problem you mentioned, It's possible to convert limited range YUV to limited range RGB without going in full range, for the file to better work with Topaz and if is possible, how can i do this conversion?

[poisondeathray]

Ok, In fact after Premiere I want to use Topaz Video Enhance AI. Is good to upscale with Topaz after a Premiere output of a MOV PRORES 4444 XQ. Or I have to convert from limited range YUV to limited range RGB. Because of the problem you mentioned, It's possible to convert limited range YUV to limited range RGB without going in full range, for the file to better work with Topaz and if is possible, how can i do this conversion?

Prores XQ4444 is usually ok , unless...

**It's only a "problem" if your source actually contains usable data in that range, otherwise you are worrying about nothing.

It's fairly common for consumer shot video to have usable data in the 16-255 range. That 236-255 is usually visible on a regular HD/TV display , because they typically use limited range RGB levels (RGB 16-235 black to white, so the overbright region is visible) . That 236-255 range will be clipped by VEIA

If it's a retail theatrical DVD it's less common to have usable data in the overbrights/darks, but they can have still have overshoots which might be useful. I mentioned this because you wrote "deinterlacing" - so it's less likely to be a commercial theatrical DVD, but possibly some documentary, or some home video. (And if it is some theatrical release, then you most likely should not be deinterlacing)

You can "legalize" levels in premiere (bring everything to 16-235), but again that is "squishing", a 2nd best method - but at least you'd be using some of the data in that overbright/dark region, instead of discarding it

Otherwise you can use ffmpeg or avisynth or vapoursynth for the full range conversion, and then out of VEIA, unless you want to keep full range (and tagged full range), you have to reverse the conversion but using full range RGB to limited range YUV again

[salvo00786]

Thanks Poisondeathray. All my works are music concerts for now and, yes these are commercial dvds.

First of all:

1) How can i work on these files without deinterlacing?

2) My files are limited range, not full range. For this reason, is really necessary to convert to full range or I can convert to limited range yuv to limited rgb? The problem for VEAI, is in the YUV/RGB or in the limited/full range?

3) If I want to try an upscale with Staxrip, what is better NNEDI3_rpow2 with spline36/64 or SuperRes? Please can you give me a good line of code for the best of the two plugins or is better to use the combined two plugins ? What are the differences between avisynth superres and super resolution used by this software http://www.infognition.com/articles/video_resize_shootout.html?

4) Probably the best sharpener is LFSMOD,right? Or do you have something better? Is better to sharpen before or after upscale?

5) It's true that is better to upscale to a resolution over 1080p and then resize with spline64 to 1080p?

[poisondeathray]

All my works are music concerts for now and, yes these are commercial dvds.

1) How can i work on these files without deinterlacing?

If they are music concerts, then likely the content is interlaced. You will have to deinterlace at some point before upscaling anyways

2) My files are limited range, not full range. For this reason, is really necessary to convert to full range or I can convert to limited range yuv to limited rgb? The problem for VEAI, is in the YUV/RGB or in the limited/full range?

Very unlikely that you have to worry about this for a commercial DVD concert.

Just use everything normally, but you can check to see if you're missing anything

3) If I want to try an upscale with Staxrip, what is better NNEDI3_rpow2 with spline36/64 or SuperRes? Please can you give me a good line of code for the best of the two plugins or is better to use the combined two plugins ? What are the differences between avisynth superres and super resolution used by this software http://www.infognition.com/articles/video_resize_shootout.html?

4) Probably the best sharpener is LFSMOD,right? Or do you have something better? Is better to sharpen before or after upscale?

5) It's true that is better to upscale to a resolution over 1080p and then resize with spline64 to 1080p?

There are no "rules" .

It depends on the source characteristics. You have to try some things out for that particular source. What works well for one DVD concert might not work as well for another...

Even when upscaling with VEIA (or anything), it depends on the source and model. Sometimes sharpening is built into the model.

eg. Sometimes LSFMod is good for one source but it doesn't do as good job with high frequency sharpening. You might want to combine some sharpeners are low strength. There are dozens of types/categories of sharpeners, and some are good at some types of sharpening, but not good at others

Super Resolution for AviSynth is just the plugin by infognition to use in a script directly. Same results as the standalone version. It's not very good by today's standards

[salvo00786]

Ok, Poisondeathray, thanks for all your replies.

Now, please, someone can tell me:

1) in Staxrip, I have seen in resize section Hardware Encoder, but there are no explanation on how to use it and I want to know if the results are good.

2) I have an RTX3070, but with PRORES, i can't use the hardware encoding in Staxrip. Please can you tell me the best output container and the best settings for for h264 output?

What are the differences between in the final files between the software encoded and the hardware encoded file? The quality with the hardware encoded file is bad?

For your last response, in general is better to use only nnedi or combine nnedi and superres?



In the meantime, I worked on nvenc settings. This is a file with my settings. The screenshot are only of the pages with settings changed. The rest that isn't in the screenshot are on default. Let me know if I can optimize some of these settings or other settings.


[Attachment 66701 - Click to enlarge]

[salvo00786]

Hi... I want to upscale x2, like 1280x720 to 1920x1080. When I use NNEDI, is different the quality if I use a x4 factor instead x2 factor (For example x4 is more precise), or is only a waste of time?