highest quality mpeg2?

[glm1051]

Hello,
I recently bought the advc 100 to capture vhs tapes. It does a perfect job,
no droped frames,...
I then use TMPgenc to convert the avi file into a dvd compliant mpg using settings from one of the guide on this site "high setting for mpg dvd compliant" I author with Ulead dvd workshop. When I play the cd, I am not
pleased with the quality. I know it's compression and loses quality. However, I was led to believe that you could achieve results that are near
perfect. I don't expect to get something better than my original VHS tape, but I do expect to get identical quality. Can anyone direct me towards a solution?
Thanks
Gui

[Bravoxena]

if the AVI file doesn't show the quality you're expecting, it could just be your capture card.

[ZippyP.]

What resolution is your avi and what resolution for the DVD?
What codec is your avi, DV?
The quality of your avi is a big factor! The more details you give the more people may suggest solutions.

[Fishingeek]

Your AVI should be fine. I would consider using CCE to convert to MPEG2. I bought the basic version and have been very impressed the the quality that I get. I also use the ADVC-100 so I assume your AVI quality is as good as mine. The speed of CCE is very nice. I am hard pressed to tell any difference from the original VHS.

Hope this helps,
Fishingeek

[Triffid]

I've found that guide to be quite a good guide, actually.

However, the result you get is really going to depend on the bit rate you can throw at the thing. So some really important parameters are: How long is the movie? How many disks do you want to use? And, Do you really care about getting the BEST audio?

To put a 2 hour movie onto one disk with PCM audio will SEVERELY restrict the bitrate you can use.

I would recommend either: 2 disks OR MPEG layer 2 audio (even then your still restricted a bit, if you keep full resolution).

What I have found works well (in TMPGEnc) for single disk movies is:

Half D1 resolution
MPEG layer 2 audio (Toolame)
Noise reduction IN (in TMPGEnc)
Soften Block noise ON
Constant Quality (85%, 0% P/B frame spoilage Max 8000, Min 1000) (But there is a risk the file could get too big...and this does take a long time)

With the noise redution ON, you can get very low quantisation even at bitrates of about 3000 (sometimes as low as 2000), so there is plenty of spare rate for peaks up to 6-8 k when needed. (and if you do starve for bits, the soften block noise will cut in and help out)

[glm1051]

When you say: "Your AVI should be fine. I would consider using CCE to convert to MPEG2. I bought the basic version and have been very impressed the the quality that I get. I also use the ADVC-100 so I assume your AVI quality is as good as mine. The speed of CCE is very nice. I am hard pressed to tell any difference from the original VHS."
You are right. My AVI are fine; they look like the tape. I capture in full DVD resolution. When I convert in Tmpgenc, I convert at 8,000 bitrate. I use almost the highest settings as reccomended in the guide mentioned above. I do not care about audio for now so I just encode at MP2, 128K. All I am interested in achieving so far is as good as VHS quality when I compress in Mpeg. I am getting better results when I do 2 VBR in Tmpgenc, Apply a couple effects, but it then take 5 hours. I have T-bird 900Mhz, with 512 of ram, 7,200Rpm drive...
How much is the CCE? that you are refering about? I have tried some trial version it does look faster, however, when I encode it says that parameters are invalid, going to default. Any Idea on that?
Anyway, the quality degradation that I am talking about is some blocks as a result of the compression.; it's just not as good as the original VHS.
Thanks for the help,
Gui

[rob1]

I'm trying to do the exact same thing- convert VHS to DVD using the ADVC-100 and get results that look like the original.

I'm using CCE Basic as well (around $50). Taking the DV straight into CCE doesn't really produce quite the result I would like- the resulting mpeg2 is fairly 'noisy'. I'm pretty sure I've read CCE is fast but output tends to be a little noisy, while TMPGENC tends to be a little blocky.

To get good results, I've been filting the DV first (via VirtualDub), then sending it to CCE. This seems to result in good quality, but it's slow as hell. I figure it's time to learn Avisynth to balance speed and quality.

But yea- I'm liking CCE as long as it's fed a pre-filtered DV. It might be worth a try. That, or play with the TMPGEC filters some more. Be curious to see what others are doing to get a high quality DVD out of VHS material.

[glm1051]

I wonder what other people do. So far it seems that the two best choices is to either use Tmpgenc or CCE. There does not seem to be an easy answer either. Time is a big factor. I do have about 30 tapes worth of "tv stuff" that I want to save into dvd, and to encode an avi file of 4 GB in 8 hours or more seems almost impossible and discouraging. Also sounds like I need a CPU upgrade. nevertheless, my concern rght now is to achieve near perfect quality.

[av-freak]

-----
Time is a big factor. I do have about 30 tapes worth of "tv stuff" that I want to save into dvd, and to encode an avi file of 4 GB in 8 hours or more seems almost impossible and discouraging
-----

I know what you mean. I have about 50 tapes myself I want to save, but the problem is definately time!

I use a Pinnacle DV-500 capture card which has a harware encoder on it. If I use DV footage from my camera then it is perfect, since there is little noise. If I use VHS footage, then I use Tsunami, because of its noise reduction. However, as we all know, it is really time consuming and sometimes I end up with a none compatible stream, which is very odd since I always use the same dvd-compliant settings.

It takes me about a week to build one dvd with 2 hours footage and menus and stuff. So, with my 50 tapes I'll be finished somewhere next year, without any vacation

I have tapes from the 80's and every week I spend making crappy streams these tapes lose more quality, so what I do now is letting VirtualDub smooth the picture a little bit (setting 2 or 5) and after that let my hardware encoder encodes the picture. The output is nearly as good as Tsunami and definately worth the time profit.

Happy encoding!

[glm1051]

[quote="rob1"]
To get good results, I've been filting the DV first (via VirtualDub

What kind of filtering do you apply?
GUI

[nickerous]

I use the ADVC-100 and TMPGenc to encode my svcds. Time is not an issue for me. I absolutely adore the stuff I'm encoding and won't sacrifice quality to get the job done faster.

My settings:

Noise reduction, highest-quality motion search, inverse-telecine (my stuff is not 30 fps although I capture it at 30 fps), and I use the audio edit to get the audio at 100 percent. The resulting mpegs are, basically, better than the source video. I say basically, because, while the conversion takes out all the VHS noise, the picture can show some pixels when there is heavy action. In whole, I am very satisfied with the outcome.

It takes around 7-9 hours to encode 20 minute episodes. I set up one before I go to bed and another before I go to work. This way, I'm doing other things while my computer is doing the encoding. I burn the discs when I get home in the evening and then capture 2 more episodes that night.

[DJRumpy]

If you have any number of VHS tapes to convert, you definately need to learn AVISynth. VirtualDub is nice for beginners, but it's very slow. It works in the RGB colorspace, which is way to slow. AVISynth can work in native YUY2, which is much faster. You should also look at the CCE basic package. It runs about the same price as TMPgenc, but the speed difference is well worth it. Combined with AVISynth, you can encode at well over 1X plaback speed (realtime).

For VHS captures, make sure you capture at full resolution (720x480 NTSC, or 720x576 PAL, and with a lossless codec if possible (like Huffman). If you don't have the drive space for a lossless codec (captures could go well into the 30-50 Gigabyte range), opt for a codec like MJPeg, or DivX/XviD. Note, that these are all lossy, causing a slight decrease in quality.

To pre-filter your video, you need to first clip off any noisy edges, or simply black them out (you can do this with either VirtualDub, or AVISynth and the LetterBox command). Leaving these noisy edges will eat up bitrate, and reduce your overall quality. You normally don't need to blacken more than 10 pixels, but this varies, depending on your VHS, and the condition of the heads. Examine your capture in media player to determine how much edge noise there is.

Consider performing inverse telecine (IVTC) on your video, if it's telecined. The reduced framerate, and elimination of those interlaced frames will improve your output. Of course, if your material is true interlaced, then this isn't an option. You can perform IVTC in both VirtualDub, and AVISynth.

You also need to use a good Temporal Smoother to remove analog noise. I use AVISynth, and the TemporalSmoother that is included with MPEG2DEC (you can find both in the TOOLS section). A typical script would look like this:

LoadPlugin("C:\My Plugins\MPEG2DEC.DLL")
AVISource("C:\Captures\capture.avi")
Letterbox(10,10)
TemporalSmoother(2,3)

Whatever you do, do not delete your capture until you are satisfied with the disc in the player. I can't tell you how many times I've had to recapture, because of a glitch, and I had already deleted my source files. Make sure you clean your VHS heads too. A head cleaner can be found at most electronics shops. If quality is paramount, consider upgrading your VHS to SVHS, and using the S-Video outputs to capture with. I didn't go the SVHS route, since my VHS is on the way out, but you may want to.

[yamato72]

glm1051, you're getting a lot of good info here but it's all over the place depending on what YOUR needs are. DJRumpy mentions capturing uncompressed which you can't do AFAIK with the ADVC100. DV is all you get. I personally use an ADVC1394 (same thing but an internal card) for capturing my Hi8 home videos. I've spent months on this and I come to the forums here routinely for tips. I think the reason you don't like your results is the noise in your VHS source. It all comes down to the need to feed your mpeg encoder the cleanest possible source. Do some filtering prior to encoding (I still use VDub) and you'll get better results. My biggest problem was mosquito noise (I use TMPGEnc) before I started filtering. I'm archiving footage of my daughter's first years and will not settle for a crummy finished product.

Currently I'm using the following filters in VDub:
1) Steven Don's DNR at setting 12
2) Klaus Post's Smart Smoother 2.11 at dia 5, thr 50, amt 200

I frameserve to TMPGEnc, 2-pass VBR 1000/5000/7000, highest-quality motion, 10-bit DC and I MASK (not crop) the noise on the sides and bottom of the frame (under the advanced tab). I've done an A/B comparison on my TV with the original tape played through the camera to the finished DVD in my DVD player and I think the difference is too close to call.

I personally didn't like the Temporal Cleaner filters when I tried them but I'm not capping VHS.

You need to be prepared to do a lot of tests. Pick a 20-30 sec scene in which the quality problems you're having are most noticeable to you and filter it a dozen different ways before you encode. Heck, I authored 10 test clips to a DVD+RW and watched them on the TV with my wife to determine which one we liked best. The settings above gave the least amount of noise while still preserving the fine gradients on skin (I look closely at foreheads and cheeks, and any solid-color areas to determine if I've gone too far with the filtering).


It's important to be familiar with each step in your "post-production" process and know what you will/will not be happy with, then move on to more advanced techniques. I'll try my hand at Avisynth on the next footage I work on, but I'm still not ready. Currently a 30 minute clip takes me 8-10 hours to encode this way, and I'm on a 2GHz P4, 512MB Ram. I don't mind the time because I think the reults are fantastic. And since like most people I work/sleep normal hours, I have the encoding running while I'm doing other things.

Just be patient, as you will not get good results without a lot of trial and error, and that will directly affect the quality of your finished product. My method above is just a suggestion, keep reading these forums and trying different things and you'll be amazed at the quality you can get from a standard PC.

[yamato72]

Yikes I reread my post and it sounds like I'm trying to disagree with you DJRumpy but I'm not -since I'm using similar hardware to glm1051's I just wanted to mention what I do....

Sorry DJ, I l know you're one of the most knowledgeable (and patient) posters on here....

[DJRumpy]

No worries 8) yamato72. I wasn't suggesting capturing uncompressed though. Huffman actually compresses your video, but it is a lossless compression scheme. If you choose the 'Convert To YUY2' option, then the codec is described as being 'slighly' lossy due to the conversion from RGB, to YUY2, but I can't see any difference. The compression ratio is about 1.3:1. Capturing uncompressed isn't a good idea. It's extremely disk intensive, and likely to cause more dropped frames than capturing to Huffman.

DV Video is also compressed as well, although I never work with that codec. I believe it has a compression ratio of about 5:1 (don't quote me on that though, as I don't have the specs in front of me). I don't know if the compression it uses is lossless. Maybe someone with more experience in that arena can comment. I believe it is slightly lossy. Adam, Snowmoon, Baldrick, anyone??

I should also mention that I've never used the TemporalSmoother in VirtualDub, but I do use the one included with MPEG2DEC. It does excellent work for VHS sources. It will virtually eliminate the 'mesquito' noise that yamoto72 is describing, with no noticable loss in detail. Try it on a clip with blue sky in the background if you have problems seeing the difference. You should give it a try, although you will have to become at least slightly familiar with AVISynth, or download one of the tools to write the scripts for you. They are a handy crutch for getting used to it, until you learn the ropes.

I tend to avoid 2-d filters. They tend to bleed out details (like the skin tone problem yamato72 is describing. You should definately try them though. Quality is such a subjective thing. What I find as unacceptable, you may find visually appealing

[yamato72]

I should mention that I've used Huffman (whatever happened to "Huffyuv?") with the settings DJRumpy mentions and that method is likely the highest-quality capture method available to a home PC user (on a budget anyway). I was using it with a Pinnacle PCTV/generic drivers etc. I switched to a DV cap card mainly because huff-encoded caps were tough for me to work with in a non-linear editor. Real-time playback was impossible. The convenience of real-time previews and smaller disc-space requirements is nice too. Right now I'm cutting together a movie of my oldest daughter's first year. I cap in Vegas to DV, then load the cap in VDub, crop out the junk, apply my filters, and do 3 things: 1) save the audio as wav (for my DVD authoring) 2) save the filtered avi to disc (panasonic dv codec) to use in my movie timeline, and 3) frameserve to TMPGEnc and encode as m2v (again for DVD). This takes 10-15 minutes in front of the computer and 9-12 hours encoding. Then I've got finished "raw footage" in mpeg2/pcm audio for posterity and also clean DV video to use in my movie timeline. I stick to 20-30 minute clips at a time and delete the original cap after this step (only have 100GB to play with).

The only drawback right now is that after the movie timeline is exported from Vegas, the video clips contained within are 3rd generation (capped/filtered/exported). I have to filter the video seperately from the finished movie because I use titles and panned stills and the filtering IMO was degrading the quality of those parts at the same time it was cleaning up the video segnments. Vegas claims that when exporting the timeline only transitions etc. are re-rendered, though, and I can't really see a deterioration in quality anyway.

Once I finish this movie, I'm considering capping the "posterity footage" with Huffman. Of course this would mean capping the same footage twice. Yikes! Now that I'm comfortable with filtering my video in VDub, I may try my hand at huffman-compressed footage in Vegas. And Avisynth. I also will work on tweeking my TMPGEnc settings (or use AC3 instead of PCM) so that I can get a full 2 hours on a DVD. Currently I can fit 80-90 mins of what I consider near-perfect quality copies on one disc.

[ktnwin]

I am converting my VHS tapes to DVD as well, and time is a major factor as well as quality. I do not want to use software to filter noise out because it takes forever.
I also noticed some of the tapes when converted to DVD tends to be noisy especially in the dark scenes. Bright scenes always look perfect. The main trouble is when playing the AVI or the resulting VOB files on the PC, they all look perfect.
Someone has said that by using the SIMA COPY MASTER to filter noise during capture will help to improve the capture quality.

[yamato72]

I guess there are 2 things to consider. One is that anything you can do to improve your source before capture (such as cleaning video heads, using a VCR with a TBC, cleaning corrosion from connectors, using the Sima) will considerably shorten your post-capture work cleaning your video. Two is that by all accounts you can use alternative software, such as Avisynth and CCE, to encode quicker than VDub/TMPGEnc can.

Heck, just using CBR instead of VBR turns an 8-hour encode into a one-hour encode in my experience. It's just that your file's 3 times as large.

Also, the more time you spend dotting your I's and crossing your T's BEFORE you encode to mpeg, the less lkely you are to find out that the results of your hours-long encode job doesn't look right. That's what causes people to throw in the towel.

At this stage in the game, with less than perfect source material, good results just take a long time. Look how quick/easy a real-time hardware (even software)-based encode from the output of a satellite dish can be.

[DJRumpy]

ktnwin, the 'noise' your describing in the dark areas is due to luminance compression. The PC monitor supports 255 levels of luminance. The tv only supports 235. If you capture in 255, or it gets converted to 255 anywhere during the capture/conversion process, and then it gets compressed to 235, you get the sparkles in your dark gray/black/green areas. Without knowing what capture/conversion/encoding methods your using though, I can't be of much assistance there.

Curious about the noise filter. They generally do not take up that much time (unless your filtering in TMPGenc of course). You should prefilter in VirtualDub, or optimally, in AVISynth. With AVISynth, you can do everything on the fly, and view your results in realtime. Just make sure your working in YUY2, and not RGB colorspace. Even filtering in VirtualDub will generally net you a new avi at about 1X speed (20-30 fps). It may add to your conversion time, but the result is usually well worth it.

[tommyoz]

I hope i'm not off the subject with this post. There is a lot of good info on this page more than i can chew. I just wanted to include that having an upgrade on your PC is very helpfull as well, especially on time. I don't know what some are using but I just built a new machine P4 2.5Ghz and i'm finding a lot better results with my video encoding and transcoding.
I haven't tried it with capturing yet but that will be next.

I also have to ask DJRumpy if he is or was an actual DJ? or is it just for the name?

[tphaggerty]

Like yamato72 said, the ADVC-100 can capture only to DV, its a hardware encode (I have an ADVC1394 as well, its the same basic idea). These Canopus products do an hardware encode of analog to DV, which is then passed to the "capture" program as a DV AVI stream. They appear to the system and capture programs as DV cameras. So Huffy just isn't an option.

I too am doing exactly the same, converting 8mm and Hi8 to DVD. I have settled on AVISynth (which is REALLY not that hard to learn at all, in fact, since you don't have to "start" a frameserver at all, once the script is put together, it is much easier to use than VDub). I use AVIsynth to do noise reduction (using NoMoSmooth right now, still experimenting), cropping (to center a slightly off center video stream), masking to remove overscan (which is a crop() followed by an Addborder() in AVIsynth) and finally BiLinearResize to 352x480 in order to give my encoder all of the bitrate needed. I then encode in Mainconcept and author using Ulead MF2. This gives me a GREAT looking 2 hour DVD. I plan to try out a few of the noise filter suggestions on this thread though. The only problem I have is that LONG encodes wind up with a few spots of spurious artifacts (it seems to be some problem with MainConcept, but not sure).

I also have CCE basic and have done the same, but am having problems with the fact that CCE outputs to elemetary streams in 2 pass VBR (have to multiplex for MF2 to recognize correctly) and stuttering in the final output. Still working on it.

I use VBR 2200/4700/7200 for encoding (again to a 352x480 output resolution) and this just about fills up a 2 hour DVD when using Layer 2 224/kbps audio.

Print this thread out and try this stuff, you should get better results.

[tphaggerty]

One thing I forgot. There is a thread here and on Doom9 about using the Canopus DV codec for decoding the captured AVI. I recommend trying this. It really seems to provide a much cleaner starting point than DirectShow or the Panasonic DV codec (both of which I've used).

You can download the Canopus DV codec (decode only) from canopus site. In order to have AVISynth use this codec, you have to change the fourCC code in the AVI to 'cdvc'.

[DJRumpy]

tommyoz, just a digital home DJ. I mix digitally with Samplitude. Compilation type stuff. Mostly techno/progressive, with a smattering of Trance. Samplitude is an excellent tool for mixing/fades.

tphaggerty, you should try a higher end software product that can support elementary streams. I like SpruceUp, as it also supports 5.1 AC3 streams and it's simple, but check out any of the higher end products. Many programs will accept MP2's renamed as .MPA. Dunno why, but you should try it. Maybe save yourself some work

The only suggestion I would have would be a lower MIN, and a higher MAX for your DVD bitrate. They won't affect the final size, but they will allow the encoder more freedom to meet bitrate needs. I always use 0, and 9800 respectively. Some will tell you your player may have problems. Think of it this way. Your audio always has some bitrate going on, as does your video (even black frames take bits to encode). I've burned DVD's for as many as 10 different people using a 0 MIN, and have never had one turn out as incompatable.

[tphaggerty]

The only suggestion I would have would be a lower MIN, and a higher MAX for your DVD bitrate. They won't affect the final size, but they will allow the encoder more freedom to meet bitrate needs. I always use 0, and 9800 respectively.

Yeah, that's what I was doing originally. I tried changing to a more "balanced" min/avg/high bitrate based on other threads to see if it would eliminate my spurious artifacts, I'll probably go back to what you said. I almost always check all of my encodes using BitRate and noticed that MainConcept (VBR 1-pass) and CCE Basic (VBR 2-pass) never went above about 7000 (my original max was about 9000). BitRate also says that CCE consistently produces an encode with much lower Q values than Mainconcept (but I am wondering if I used closed GOPs my CCE encode, which would affect this, no?).

CCE basic produces .mpa files for audio, I should try renaming to .mp2, maybe that will work! I tried SpruceUp but was a little concerned about the driver installation issue (I can't remember what they're called). MF2 is really not too bad, much better than MyDVD etc. And its CHEAP (I didn't see much reason to go to DVDWS).

I have noticed that CCE produces a much "softer" encode than MainConcept even with the exact same AVISynth setup. Anyone else notice this? At first I didn't like it, but if you are far enough away from the tube, it seems to improve the picture somewhat.

[yamato72]

tphaggerty, is this the thread you're referring to?

https://www.videohelp.com/forum/viewtopic.php?t=155686&highlight=canopus

fingernailx wrote the following:

Couldn't pass a chance to complicate it more. If I 'capture' or transfer to an avi wrapper file, a dv clip from a camcorder to my hard disk and open in VirtualDub I can be decoding it with the canopus codec, the panasonic codec or the mainconcept codec depending on whats installed as a vfw codec on my machine. If I open the same clip in Vegas video I'm using their direct show filter under wdm. If I open the clip in Movie Maker or WMP I'm using the M$ built in filter. If I open it in tmpgenc, mainconcept or cce do I open with vfw or directshow? What are my priority settings in tmpgenc? What are my import module settings in MainConcept? What if I frameserve to cce using avifilesource or directshowsource?

And never got an answer. Can anyone discuss the effect using the Canopus codec vs. Panasonic codec just for decoding? I thought the decoded file (opened in VDub for example) would be the same regardless, and the codec only really comes into play when you re-render (for example if you filtered and saved a new avi in VDub).

Thoughts?

[tphaggerty]

tphaggerty, is this the thread you're referring to?

https://www.videohelp.com/forum/viewtopic.php?t=155686&highlight=canopus

Yes, but if you look in there, I think it points to a more detailed thread on Doom9. I got this to work by changing the fourCC code and using AVISource() in AVISynth (not directshow). The only issue is that the first frame is GREEN, then everything is OK (same for MC Encoder, CCE and TMPGenc). To my eye, the file looks cleaner.

I did some test encodes (with MC) using the same AVIsynth file, just changing the fourCC code of the source to force the use of different Codecs for decode. It did look better to me PLUS BitRate showed a lower Q value on the resulting files. Note that on the Doom9 forum, it says that the Canopus Codec has the "chroma upscaling bug" and you need to add the FixBrokenChromaUpscale() command to your AVIsynth file to correct.

[tphaggerty]

Here is the doom9 thread on using the Canopus DV codec for decoding an AVI file.

http://forum.doom9.org/showthread.php?s=bf6fb0c5d62b3d1a421442505c3ba43d&threadid=33526

[yamato72]

Thanks tphaggerty I'm going to try the Canopus codec and editing the fourcc in the avi. This will be handy if I move to Avisynth, too bad you need to fix the chroma bug. Nothing is ever simple when perfection's what you're after....

[EddyH]

Haven't heard much buzz on these lately (flopped?) and certainly no specs/results, so...

How good a quality do standalone DVD recorders (yknow.. the VCR replacement things) do in comparison to all our favourite programs, especially for applications such as this?