I upscaled and interlaced my Video8 clips at the time with TVAI v3.07 all in TVAI (Diode TV/DV) and the quality of the output file was very good. I tried doing the same with the same exact Video8 Clip on v3.1 and the video quality now is very bad… it’s like the interlace function is not working well. or it could be because in v3.1x Topaz increased performance it cutting corners on quality.
please look into it and fix it.
In the mean time I am reverting back to v3.07…
UPDATE: with v3.07 I noticed I encoded the file using H.265 Auto. while with v3.1.2 I used ProRes HQ.
When I encoded the file again on v3.1.2 , this time using H.265 the quality of the output video was much better and identical to the previous v3.0.7 H.265.
So my conclusion there is a serious issue with ProRes HQ, it outputs degraded quality (at least when de-interlacing is needed).
Even though it doesn’t makes sense it would, as ProRes should be nearly lossless, but in practice on the field it yield worse quality output file then the H.265 Auto.
With ProRes the de-interlacing is not working well… that is my suspicion…
UPDATE 2: just Encoded the same Clip again using H.264 Auto, on v3.1.2 and the quality came out the same good quality as the H.265 on both the v3.0.7 & v3.1.2.
So it is a clear cut that the issue is with ProRes HQ. it yields crappy quality when needs to de-interlace.
Then don’t let the source stay interlaced. Use QTGMC or so in an Avisynth/Vapoursynth script, or let TVAI deinterlace. Interlacing has always been a dirty hack to deal with NTSC, and should never have existed to begin with.
Your conclusion is wrong. ProRess 422 HQ is, for all purposes and intent, de facto lossless. Your H265 encoding may already be doing the deinterlacing for you (unbeknownst to you), and that would account for the better result. ProRes 422 HQ preserves the scanning method used in the source material. So, unless you tell TVAI to deinterlace, the output will stay interlaced.
for your point 1. QTGMC is not the greatest option as I noticed Progressive upscaling in TVAI is not as good as interlaced. read more here.
for your second point, I did not say ProRes is bad, I said it is bad/worse in TVAI.
i used both ProRes and H.265 in TVAI only… and H.265 provided a better quality on interlaced material. facts don’t lie, even if it doesn’t make sense… the field practice prevails…
Wait, wut? We must be at cross-purposes here. Progressive always trumps Interlaced, always. And QTGMC deinterlacing is superior to that of TVAI. Unfamiliar with Bob, I must admit, but QTGMC does doubled-frames too (in fact, it’s default). Just make sure you set InputType=0 for an Interlaced source.
What do you mean? ProRes 422 HQ = ProRes 422 HQ. There’s no ‘doing it worse.’
That is likely because, like I said, TVAI’s H265 deinterlaced for you, whereas ProRes left the output Interlaced.
there is no point discussing this further, obviously you did not read my post. you are locked in your opinion.
this post is for Topaz to read as they will know what i mean after they will repeat the testing and review their algorithm… and read my post with all the steps and facts…
@Akila It always helps to show actual screenshots or video exports to help others see exactly what you’re seeing. For example, I used TVAI 3.1.2 to deinterlace and export samples using h265, ProRes HQ, and h264, all with identical settings except for the encoder used during export. The encoder used during export is generally (at a high level) only going to impact what artifacts are created during output (as in each encoder will have different artifacts because they compress differently), how colours are represented, and file size. GENERALLY. How a video is deinterlaced is done identically regardless of output encoder because it is necessary to render each frame fully before compressing it for output. I acknowledge that I’m not addressing your concerns about differenced between versions of TVAI, so my experiment is different than yours, but personally I like to test changes on one common version completely and then re-run the entire set of tests in a different version if needed. I just don’t have time to download and rerun the tests on the older version of TVAI right now.
Okay, here’s what I see:
1080i source video, deinterlaced, using Dione DV 2x FPS, encoded using H265 Main at 24Mb/s
Everyone has different criteria for determining what video looks best to them, given multiple samples. For me, the ProRes 422 HQ encoded video shows richer colours more natural colours and smoother shadows where I see large compression artifacts on the front speaker for example in H265 and H264. However, that’s just my opinion.
The bottom line though is that whether it is H265, H264 or ProRes for the export encoder, it has no impact on how deinterlacing occurs, that is done earlier in the pipeline long before it ever gets near an export encoder. **If this isn’t the case for you, it would be great to get a sample of the video you are working with so myself and others can try to help you replicate and diagnose the issue you are seeing to put more weight behind a request to Topaz if indeed there is a reproducible error. **
And that doesn’t come as a surprise to me. Unless you fiddle with the command line, H265 output in TVAI is 8-bit default, whereas ProRes 422 is, well, 4:2:2 at 10-bit: better color, better gradients, better shadows, etc.
That’s really interesting, I’ve never done anything with B&W. Would love to see the output from the video you’re working with if only just to educate myself on the potential impacts of the various encoders on your type of source material.
Output to Tiff and then compare Tiff with ProRes HQ… What gives?
Or simply check the bitrate of the ProRes HQ output Vs. H.265.
The think is… H.265 will encode to a much lower bitrate and for that it needs to remove all the crap it finds, on the contrary ProRes HQ encodes everything
Use QTGMC or so in an Avisynth/Vapoursynth script, or let TVAI deinterlace. Interlacing has always been a dirty hack to deal with NTSC, and should never have existed to begin with.
