Using TMPGEnc Master Works 7 to encode DV-AVI to mpeg2......Best settings.

[ConsumerDV]

I just tried saving a frame picture from the software itself, but the size is only 655x480, way too small to see any detail.

Not surprising, considering that your original video is DV, and your tool saves frame grabs with square pixels.

[davexnet]

This whole thread is veering into the absurd

[jagabo]

Lossless h.264 encoding is only available with the x264 encoder. I'm not aware any other h.264 encoder that has a lossless mode. Using a quantizer of 0 (ie, qp=0) would not normally result in lossless encoding. x264 takes it as a flag to switch to a special lossless encoding mode, not just to use a quantizer of 0.

And lossless x264 is lossless relative to the uncompressed YUV video it's given. It's only lossless relative to the source video if you avoid any other changes before encoding. If you're NTSC DV AVI is converted from YUV 4:1:1 to YUV 4:2:0 (as far as I know, no other encoder support 4:1:1, so you will be changing the color format) there will be losses before the encoding. And since DV is interlaced you need to convert to 4:2:0 with an interlaced algorithm and encode interlaced, bottom field first. If you are seeing differences between the source and a lossless x264 encoded video it's because of those issues.

For A/B switching I use AviSynth with a script like:

A = LWlibavVideoSource("source.mp4").Subtitle("source")
B = LWlibavVideoSource("encoded.mp4").Subtitle("encoded")
Interleave(A, B)
ConvertToRGB(matrix="rec601", interlaced=true)

Then open the script in Virtualdub where you can flip back and forth between subsequent frames with the left and right arrow. I also use Windows' built in Magnifier to zoom into the details.

[Purple2112]

I just tried saving a frame picture from the software itself, but the size is only 655x480, way too small to see any detail.

Not surprising, considering that your original video is DV, and your tool saves frame grabs with square pixels.

Yeah, don't even get me started on the whole rectangle/square pixel thing. When viewing the original DV videos they look grainy and I can actually see the pixels moving around. I'm fine with that because it's still a sharper image/better detail than after encoding to mpeg2/H.264. When I encode to mpeg2/H.264 the picture appears less grainy, but the pixels also seem to be much smaller and with that comes a loss in picture sharpness. When I encode to (in my case) VP9 lossless (because I don't have H.264 lossless with my software) it gives me that original grainy look with the floating pixels. Is this because lossless is keeping the original DV rectangle pixels VS mpeg2/H.264 changing them to smaller square pixels as part of the compression process?

[Purple2112]

This whole thread is veering into the absurd

Well, I'm learning something new with every new post and isn't that the whole point of using this forum. Unfortunately, sometimes the answers can lead me in another direction ( I try to stay on track) or create even more questions.

[davexnet]

I thought you might have had, by now, some results to share with us.
Your goal, as far as I can see, is to find the best format/quality that will play on your playback device.
Seems relatively straightforward. Looks of of hints and direction in this thread

[Purple2112]

Lossless h.264 encoding is only available with the x264 encoder. I'm not aware any other h.264 encoder that has a lossless mode. Using a quantizer of 0 (ie, qp=0) would not normally result in lossless encoding. x264 takes it as a flag to switch to a special lossless encoding mode, not just to use a quantizer of 0.

Like I said in a prior post, unfortunately, I don't have H.264 lossless, I have VP9 lossless & the VP9 quantizer with the QP setting. The whole writing script thing scares me, I've used command prompt before, but I was just copying the command directions from the internet, I have no idea how to come up with my own script.

And lossless x264 is lossless relative to the uncompressed YUV video it's given. It's only lossless relative to the source video if you avoid any other changes before encoding. If you're NTSC DV AVI is converted from YUV 4:1:1 to YUV 4:2:0 (as far as I know, no other encoder support 4:1:1, so you will be changing the color format) there will be losses before the encoding. And since DV is interlaced you need to convert to 4:2:0 with an interlaced algorithm and encode interlaced, bottom field first. If you are seeing differences between the source and a lossless x264 encoded video it's because of those issues.

For A/B switching I use AviSynth with a script like:

A = LWlibavVideoSource("source.mp4").Subtitle("source")
B = LWlibavVideoSource("encoded.mp4").Subtitle("encoded")
Interleave(A, B)
ConvertToRGB(matrix="rec601", interlaced=true)

Then open the script in Virtualdub where you can flip back and forth between subsequent frames with the left and right arrow. I also use Windows' built in Magnifier to zoom into the details.

I understand that the encode will never be an exact match to the source, but what I'm seeing is not just some fuzziness or artifacts. I ended up just taking two screen captures from the exact same point in a video and used the Windows Magnifier that you mentioned to zoom in on it. I know it's not the best way to check for differences, but hopefully you can still get an idea of what I'm talking about. All my videos are of nude female friends and girlfriends so I had to blackout 98% of the captures. What you will be looking at is a crease under a breast. The 1st one was taking from a standard H.264 (constant bitrate) encode. You should notice that the crease is a solid, dark, brown line.

[Attachment 64782 - Click to enlarge]
The 2nd capture was taking from a VP9 VBR(Constant Quantization) encode with QP set at 10. (The same thing happens when encoding to VP9 lossless) You should now notice that the line is no longer solid, it looks more like a white and brown striped snake. This is very noticeable when watching the video, it almost looks like a brown and white strobe light.

[Attachment 64784 - Click to enlarge]
I don't expect you to know why it's happening, just wanted you to see what I was talking about. The other two problems with the lossless that you might have answers for is that the picture looks stretched. I tried changing it to 640x480 and 720 x540, but when I do that it still doesn't look right and I'm getting a small black bar on top and bottom, instead of just on the sides where they belong. When using the standard H.264 encoding I don't have this problem it just somehow comes out correct, not sure if BD player or TV is auto correcting it, but it plays correctly even though it's 720x480 resolution.
Lastly, the outline of everyone is covered in little Z's which as far as I know is an interlaced issue. I can notice this in the source file a little if I really look for it, but after encoding it looks really bad. I've been encoding all of them as bottom field first as you mentioned, could the interlaced algorithm with my software be that bad?

[ConsumerDV]

I did not know this, but apparently VP9 and AV1 do not support interlacing. This is why you have your problem #3 ("little Z's") and problem #1 ("striped snake").

Problem #2 ("the picture looks stretched") is not because of the codec, but because you have failed to specify correct pixel aspect ratio for your video. The way you specify it depends on the tools you use to render the video, and on the container.

I thought you wanted to watch it on your TV? Render into H.264 either as interlaced, or deinterlace it into 60p and render as progressive. Then either burn a BD or put it on a thumb drive. Does your BD player support VP9? If you want to use VP9 you must deinterlace first.

If you use standard authoring tools for DVD or BD, they will take your DV video and package it in the format your BD player can understand. Do you watch Hollywood movies off BD? They are not uncompressed, but they look good enough, do they?

[DB83]

Maybe I missed something along the way but I fail to see any actual video before/after samples.

Given that I have too much free time on my hands I conducted a wee experiment using the 30-day trial of this s/w.


I do not have a NTSC sample to use but I do have a PAL one which I have used in other topics. The first clip is the DV.avi and the second just throwing some s/w settings at it. Is the resultant mpeg-2 any worse - visual wise - than the original ?

[DB83]

And another experiment pumping up the bitrate. Does this look any better than the earlier mpeg-2 sample ?

[jagabo]

And another experiment pumping up the bitrate. Does this look any better than the earlier mpeg-2 sample ?

Yes. There are far less blocking artifacts in the higher bitrate video.

[jagabo]

The 2nd capture was taking from a VP9 VBR(Constant Quantization) encode with QP set at 10. (The same thing happens when encoding to VP9 lossless) You should now notice that the line is no longer solid, it looks more like a white and brown striped snake. This is very noticeable when watching the video, it almost looks like a brown and white strobe light.

[Attachment 64784 - Click to enlarge]
I don't expect you to know why it's happening, just wanted you to see what I was talking about.

The video was poorly deinterlaced. Either before encoding or during playback.

The other two problems with the lossless that you might have answers for is that the picture looks stretched. I tried changing it to 640x480 and 720 x540, but when I do that it still doesn't look right and I'm getting a small black bar on top and bottom, instead of just on the sides where they belong.

You should not be resizing the video at all if you want lossless encoding. Use aspect ratio signalling (that is what your DV source does) and the player will do the appropriate aspect ratio on playback (although some players may ignore the signal depending on the codec/container).

Lastly, the outline of everyone is covered in little Z's which as far as I know is an interlaced issue. I can notice this in the source file a little if I really look for it, but after encoding it looks really bad. I've been encoding all of them as bottom field first as you mentioned, could the interlaced algorithm with my software be that bad?

This is definitely an interlacing issue. Without before/after samples I can't tell you exactly what.

But all those problems either occurred before encoding, by incorrect encoding, or are playback problems.

[jagabo]

Here's DV83's DV AVI file encoded losslessly with x264 (no audio). PAL DV actually has a sampling aspect ratio of 12:11 but I've encoded it here as 16:15 because most players will play the PAL DV that way.

[ConsumerDV]

Here's DV83's DV AVI file encoded losslessly with x264 (no audio). PAL DV actually has a sampling aspect ratio of 12:11 but I've encoded it here as 16:15 because most players will play the PAL DV that way.

VLC plays OK, but MPC-HC and Splash do not. I realize that these are old versions, there is new better and free Spash, but I don't feel like upgrading every time a non-compatible video comes along. I like my Windows 7 install with all the old stuff (including Photoshop 6.0). This is not even H.265 or VP9 or whatever, so I expect Splash to play AVC in MKV.

Also, I don't see what is the point of re-encoding 36 MB DV as 55 MV uncompressed AVC. DVD-legal MPEG-2 with less than 10 Mbps is more than enough for enjoying naked bodies on a TV from a sensible viewing distance. After all, the original DV has been captured and is not going anywhere.

[DB83]

Here's DV83's DV AVI file encoded losslessly with x264 (no audio). PAL DV actually has a sampling aspect ratio of 12:11 but I've encoded it here as 16:15 because most players will play the PAL DV that way.

Well I could not play that even in VLC - total picture breakup. Granted that my system - given specs are accurate - is slow, memory-starved and ancient but it is all I need.


Of course all that matters is what the OP sees. 'Best' settings may not work for him.

[jagabo]

VLC plays OK, but MPC-HC and Splash do not. I realize that these are old versions, there is new better and free Spash, but I don't feel like upgrading every time a non-compatible video comes along.

I warned you in post #55 that not all players support x264 lossless.

Also, I don't see what is the point of re-encoding 36 MB DV as 55 MV uncompressed AVC. DVD-legal MPEG-2 with less than 10 Mbps is more than enough for enjoying naked bodies on a TV from a sensible viewing distance. After all, the original DV has been captured and is not going anywhere.

You where the one asking for lossless encoding. I even asked why you needed it in post #59. You replied:

Why wouldn't I want lossless, I started this thread asking for help getting the best possible quality regardless of bitrate and finished file size.

All lossless codecs will deliver larger files than DV and MPEG.

[ConsumerDV]

Interesting quoting, @jagabo.

[Purple2112]

I did not know this, but apparently VP9 and AV1 do not support interlacing. This is why you have your problem #3 ("little Z's") and problem #1 ("striped snake").

Problem #2 ("the picture looks stretched") is not because of the codec, but because you have failed to specify correct pixel aspect ratio for your video. The way you specify it depends on the tools you use to render the video, and on the container.

I thought you wanted to watch it on your TV? Render into H.264 either as interlaced, or deinterlace it into 60p and render as progressive. Then either burn a BD or put it on a thumb drive. Does your BD player support VP9? If you want to use VP9 you must deinterlace first.

If you use standard authoring tools for DVD or BD, they will take your DV video and package it in the format your BD player can understand. Do you watch Hollywood movies off BD? They are not uncompressed, but they look good enough, do they?

I wanted to try using a lossless format, but unfortunately my software https://tmpgenc.pegasys-inc.com/en/product/tvmw7.html doesn't support H.264 lossless, only VP8 & VP9. I did not know that VP9 doesn't support interlaced video. In light of the VP9 interlace issue and aspect ratio issue it's probably best that I just stick with encoding to Mpeg2 or H.264 for now. I'll always have the original DV-AVI captures to go back and play with in the future. I bring up Mpeg2 because for some reason it seems like Mpeg2 seems to give me slightly, sharper image quality. I don't know if it's because Mpeg2 uses less compression than the H.264 or what. With Mpeg2 I can can see the tiny, light reflection twinkle in people's eyes. The H.264 seems to mute that detail down (dulls the sharpness). I would like to learn about the best settings for both though.
As I said in the start of this thread I will not be burning the encoded files to disc. They will be stored on a 4TB external HDD that's connected to my Oppp BD player. That is why I keep stressing the point that the finished size is not important and would like to use the least amount of compression as possible and the highest bitrate needed, hopefully giving me better quality.
I need to jump back to one of the first issues that I'm still not positive about, Field Order. As far as I know DV should always be BFF. However, when I used MediaInfo to check the DVD's that I made from all these DV-AVI files back in 2006 it shows them as being TFF. At the time I used Pinnacle software to capture and then Nero software render to DVD using all the default settings. This time I just used https://www.videohelp.com/software/WinDV set to (type-2 AVI) to capture all the source videos. I know this is a long shot, but is it possible that WinDV captured the DV-AVI's as BFF, even though they were actually TFF? I tried using some of the methods mentioned on this forum to test them, but I really can't see any difference. The TFF dvd's aren't jumpy when being played.

[DB83]

Perhaps you should save your energy and actually comment on the samples recently posted rather than theories.

BTW


1. Interlaced video straight to mpeg2 = TFF
2. Interlaced video straight to DV = BFF
3. Interlaced DV (BFF) back to mpeg2 = BFF

[Purple2112]

Maybe I missed something along the way but I fail to see any actual video before/after samples.

Like I said before all these videos are totally nude female friends and past girlfriends and can't post them over the internet.

Given that I have too much free time on my hands I conducted a wee experiment using the 30-day trial of this s/w.


I do not have a NTSC sample to use but I do have a PAL one which I have used in other topics. The first clip is the DV.avi and the second just throwing some s/w settings at it. Is the resultant mpeg-2 any worse - visual wise - than the original ?

Yes, I can definitely see the degradation in the mpeg-2 encoded video.. Then again the DV-AVI source file itself doesn't look very good to begin with. Thank God the DV-AVI source files that I'm starting out with look much cleaner and the encoded Mpeg-2 came out better. Then again I changed the Profile & Level to MP@HL, used CBR(Constant Bitrate) at 40Mbps, set encoding speed to Slow, went into GOP settings and changed B frame count to zero, and lastly, changed the DC Component Precision to 10Bit.

[Purple2112]

And another experiment pumping up the bitrate. Does this look any better than the earlier mpeg-2 sample ?

Yes. I can see the difference, just upping the bitrate seemed to help a lot.

Have you figured out the whole VBR(Constant Quality) setting yet? When I asked the software developers this is how they explained it.

[Attachment 64990 - Click to enlarge]
Although, I can't imagine setting the bitrate to 60Mbps as they are stating.

[DB83]

Ok. I never stated that my sample was straight-to-DV. I was merely attempting to use the s/w to create a mpeg2 from a DV source (with mine from VHS(tv recording) hence the less that perfect result) but with a selection of random bitrates just to show what the s/w can do. And all this with just a few minutes of using it. Others with more knowledge of mpeg2 and encoder settings could probably get more out that even that.

Yes. You have already stated that you can not upload actual examples. But I still think you are chasing your own tail here. Bitrate is not everything. DV is a legacy format which had its champions 20 years ago. Mpeg2 is only still 'suffered' due to dvd. Even mpeg2 at the same bitrate (or higher) as DV will not duplicate what you see on the DV (whatever the source)


Modern codecs such as AVC and HEVC will give you better results. And at a fraction of the bitrate. Yet there are other factors such as noise reduction which we can not even comment on simply because we can not see any original footage.

[Purple2112]

The 2nd capture was taking from a VP9 VBR(Constant Quantization) encode with QP set at 10. (The same thing happens when encoding to VP9 lossless) You should now notice that the line is no longer solid, it looks more like a white and brown striped snake. This is very noticeable when watching the video, it almost looks like a brown and white strobe light.

[Attachment 64784 - Click to enlarge]
I don't expect you to know why it's happening, just wanted you to see what I was talking about.

The video was poorly deinterlaced. Either before encoding or during playback.

The video looked the same when played as interlaced (no Deinterlacing) on TV. ComsumerDV stated that VP9 does not support interlace video.

I did not know this, but apparently VP9 and AV1 do not support interlacing. This is why you have your problem #3 ("little Z's") and problem #1 ("striped snake").

I think it's time to just let the whole Lossless thing go at this point (stop wasting everyone's time) and go back to learning about the best quality setting for standard Mpeg2 and H.264. I've been trying and comparing different GOP's like you mentioned. (Mpeg2) - One with default GOP, one with no B Frames, and one with no B or P frames (only I frame), but I can't see any difference between the three. I really need to find a sample DV that I can make changes to and then post the results.

Here are the Advanced Setting for Mpeg2, pretty simple compared to H.264.

[Attachment 65007 - Click to enlarge]
Can I uncheck (Encode Keyframes into I Frames) since I have no chapters and aren't doing any cutting? I should be able to set the (DC Component Precision) to 10 since I am using a high bitrate and not concerned about encoding time.......correct? Found some info here:http://https://forum.videohelp.com/threads/225262-TMPGEnc-s-DC-component-precision-setting

[Purple2112]

Here's DV83's DV AVI file encoded losslessly with x264 (no audio). PAL DV actually has a sampling aspect ratio of 12:11 but I've encoded it here as 16:15 because most players will play the PAL DV that way.

I had no problem playing this with Zoomplayer https://www.inmatrix.com/files/zoomplayer_download.shtml

[Attachment 65008 - Click to enlarge]

[ConsumerDV]

• native DV is BFF
• Video digitized from analog into DV is often TFF, which throws NLEs off because they expect BFF, but this is not your case.
• MPEG-2 is TFF, and good software makes sure to convert DV/BFF to MPEG-2/TFF behind the scenes.

I apologize for forgetting that you don't want to burn a DVD or BD, your mentioning of an Oppo player threw me off. I get it now that you want to use the Oppo as a network player, and it cannot play DV-AVI. This sucks, few players can.

I have no idea why H.264 looks less sharp. Can be related to bitrate, to levels, to interlacing... With everything being equal, MPEG-2 is no worse than DV and H.264 is no worse than MPEG-2, and usually they are better at the same bitrate or as good as DV at lower bitrate, this was the whole point of developing new codecs.

The original is DV 25 Mbit/s intraframe, so MPEG-2 10 Mbit/s interframe or H.264 5 Mbit/s will be visually about as good (I don't think there is a lot of difference between frames in your content, neither I think there is a lot of high-frequency detail). I pulled these numbers from thin air, if you really care you can output logs from an encoder and figure out how close the encoded video is to the original by analyzing encoder statistics, although I am not sure how this can be done if you change scanning type from interlaced to progressive... OTOH, they do compare deinterlacers somehow, so I guess this can be done. I have never done this myself and have only a vague understanding what numbers to look at. People who upload to torrent sites use such analysis religiously. I either set bitrate that I feel is sufficient, or use a CRF value.

I don't see the point of increasing bit depth to 10 if the original is only 8.

There is a reason to use 4:2:2 to preserve as much chroma as possible, but in real life you won't notice the difference. If it looks faded to you, you can boost color a little bit. You can also verify that the levels are correct.

Since you are not going to upload it on YouTube, you can keep it interlaced, the Oppo has one of the best deinterlacers.

[Purple2112]

VLC plays OK, but MPC-HC and Splash do not. I realize that these are old versions, there is new better and free Spash, but I don't feel like upgrading every time a non-compatible video comes along.

I warned you in post #55 that not all players support x264 lossless.

Also, I don't see what is the point of re-encoding 36 MB DV as 55 MV uncompressed AVC. DVD-legal MPEG-2 with less than 10 Mbps is more than enough for enjoying naked bodies on a TV from a sensible viewing distance. After all, the original DV has been captured and is not going anywhere.

You where the one asking for lossless encoding. I even asked why you needed it in post #59. You replied:

Why wouldn't I want lossless, I started this thread asking for help getting the best possible quality regardless of bitrate and finished file size.

All lossless codecs will deliver larger files than DV and MPEG.

Also, I don't see what is the point of re-encoding 36 MB DV as 55 MV uncompressed AVC. DVD-legal MPEG-2 with less than 10 Mbps is more than enough for enjoying naked bodies on a TV from a sensible viewing distance. After all, the original DV has been captured and is not going anywhere.

You where the one asking for lossless encoding. I even asked why you needed it in post #59. You replied:

I think you're confusing (ConsumerDV) response with me. I wasn't complaining about any of the info that you've given me.

[Purple2112]

Ok. I never stated that my sample was straight-to-DV. I was merely attempting to use the s/w to create a mpeg2 from a DV source (with mine from VHS(tv recording) hence the less that perfect result) but with a selection of random bitrates just to show what the s/w can do. And all this with just a few minutes of using it. Others with more knowledge of mpeg2 and encoder settings could probably get more out that even that.

My bad, I though that was footage that you shot with a DV video camera.

Yes. You have already stated that you can not upload actual examples. But I still think you are chasing your own tail here. Bitrate is not everything. DV is a legacy format which had its champions 20 years ago. Mpeg2 is only still 'suffered' due to dvd. Even mpeg2 at the same bitrate (or higher) as DV will not duplicate what you see on the DV (whatever the source)

Modern codecs such as AVC and HEVC will give you better results. And at a fraction of the bitrate. Yet there are other factors such as noise reduction which we can not even comment on simply because we can not see any original footage.

Yes, I know, that's why I'm willing to try encoding to H.264 as well. Yes, I've been experimenting with different settings on my own in both Mpeg2 & H.264 encodings. Unfortunately, there are some settings that I can't find any info for on the internet. Actually, H.264 has a sh_tload of advanced settings that I still need to find out exactly what they do and the s/w developers for this s/w are not much help. Yes, I know bitrate is not everything, that's why I've been trying other settings that members on this forum have suggested. For instance, (Jagabo) suggested doing away with B and P frames and using only I frames. I also found that in the Mpeg2 advanced settings that I can change the (DC Component Precision) from 8bit to 10 which helps to improve the quality. (changing the above two settings requires using a much higher bitrate) I know I'm limited with what I can do with DV from back in 2004, but it's tweaks like these that I'm trying to learn about.
As for samples, I've searched the internet for sample DV, 720x480, 29.9 frame rate and at least 24Mbps, but can't find any. I would love to have a DV sample that I can try different settings with and then post the results for all to see. Here are my actual DV-AVI specs, if you happen to find a matching sample let me know.

[Attachment 65010 - Click to enlarge]
So setting everything aside that we've been discussing, what do you actually think of the software. I think it cost about $120, but because I already own a rendering version of their software I got it for much less. It's suppose to be really good software, especially for timeline editing, unfortunately, it requires a really good, powerful PC.

[DB83]

For the record, there is nothing 'special' about the specs as posted above. These are just 'bog-standard' NTSC DV. (compare a mediainfo report of my DV.avi sample as above)

As for samples, I am sure that our friend ConsumerDV has uploaded some in his recent topics. Some of which I contributed to.


As for the software, I understand it has a good reputation but I only ran the trial version twice for the examples so created and not really interested in persuing it. And my PC is probably not 'powerful' enough.

[jagabo]

I think you're confusing (ConsumerDV) response with me. I wasn't complaining about any of the info that you've given me.

Yes. Sorry, I lost track of who had posted what.

[Skiller]

• native DV is BFF
• Video digitized from analog into DV is often TFF, which throws NLEs off because they expect BFF, but this is not your case.
• MPEG-2 is TFF, and good software makes sure to convert DV/BFF to MPEG-2/TFF behind the scenes.

So much wrong here, sorry.

DV is always BFF no matter the source (native or from analog). Show me one device that does not comply with this. It is correct to automatically assume BFF with DV.
MPEG2 by itself as well as DVD-compliant MPEG2 can be both TFF and BFF. It's just a flag. There is no need to convert BFF to TFF for MPEG2. All that matters is that it does not get mixed up and that the flag is indicating the correct field order.