How to get the best TMPG encode, IMHO :)

[DDogg]

First, a quick disclaimer: All this is in the vein of IMHO. At this time, I think these are the best settings for getting the best TMPG encode. Much of this is anecdotal (to use somebody's favorite expression ;) ) and comes from the UseNet A.B.VCD newsgroup. I cheerfully admit these setting are based upon subjective interpretation, that they may be wrong, and reserve the right to change these suggestions completely. However, all that said, they should provide a good starting point. :)

I expect we will continue to refine these settings via more testing and discussion. That is the main reason for the new thread and it being placed in the advanced section.

You must use Build 3 or higher as many more TMPG settings were added.

OK, very importantly, set the max to 2520 and the min to 755 in the D2S bitrate tab. In the frameserver tab select Bicubic (NOT simpleresize)and set B=0.00 and C=0.55.

Set audio to 192. [224 should be ok to use] That should keep total bitrate within SVCD spec, or close enough, given some muxing overheads.

In the encoder tab with TMPG selected choose:
Rate Control Mode=Constant quality (CQ)
Motion search precision=Highest quality

Advanced:
If not mentioned leave unchecked
DC component precision=10
Use floating point DCT=checked [If your machine is fast, else leave unchecked]
No motion search for still picture part by half pixel=checked
Soften block noise=unchecked (see note below)


Notes:
1> Soften block noise may be very useful for 2 hour 2CDR encodes where the video is bright with high action. If you get macroblocks try it on a second run. We need more testing on that
2> Speed is similar to CCE - TMPG CQ mode uses a completely different methodology than multipass and uses ONLY one pass. Because of this and even with the above settings that add quite a bit of time to the encode, the speed advantage of CCE is negated and the speed of TMPG CQ encodes will be competitive with a 4 pass CCE encode, if not faster.

Replies
I invite anybody to reply to this thread that has done a full length encode using the above settings. Please make note of the total length of the video as well as the CQ value used by TMPG in your reply. Also, the final size of the BINS or MPGs rendered.

If, after doing an encode using these settings, you feel you have a better suggestion please share those settings with us. We want to arrive at a consensus via real testing instead of theory.

Oh, one last thing, this is not the place to get into a subjective CCE versus TMPG dialog. The subject is "How to get the best TMPG encode" [via dvd2svcd]. :)

[marks430]

Using a dual P3 1GHz with 512 MB PC133, I encoded the NTSC region 1 anamorphic DVD "Joe Dirt" from VOBs previously ripped using internal routines. The length of the video was 1:30:53.

Total time for SVCD creation, 8 hours 33 minutes 23 seconds.

Encoded at aspect ratio 16:9
Force Film was activated.
Audio set to 192 with no downsample.
Resize method set to Bicubic with b value 0.00 and c value 0.55
Max bitrate set to 2520, Min set to 755 and Max average set to 2230
Between 61 and 100 minutes use 2 CD of 800 MB.
TMPGEnc Rate Control Mode set to Constant quality (QC)
Motion search precision set to Highest quality
DC Component Precision set to 10
Use floating point DCT and No motion search... both checked
Soften Block Noise unchecked

Video Quality Value Used was 85.51

Ended up with 2 bin/cue files.
CD1 was 825,644 KB and CD2 was 796,515 KB

I burned the images using Daemon Tools with NTI and the video was very good quality when viewed on my 48" TV.

[DDogg Edit: I am taking the liberty of editing in the same encode results done with CCE that Marks430 posted in another thread. Note the times of encode are nearly identical. Kudos to Marks430 for the clear and concise posts! A perfect example of how to do it.]

The CCE encode is now finished. I encoded the same VOBs as above. NTSC Region 1 "Joe Dirt" anamorphic.

Total Time for SVCD creation, 8 hours 47 minutes 48 seconds.

Encoded at aspect ratio 16:9
Force Film was activated.
Audio set to 192 with no downsample.
Resize method set to Simple Resize
Max bitrate set to 2530, Min set to 300 and Max average set to 2230
Between 61 and 100 minutes use 2 CD of 800 MB.
Video Encoding Mode set to Multipass VBR 4 Pass
Image Quality Priority set to 17
Anti noise filter checked and set to 2
Linear quantizer scale checked
Upper field first checked
Zigzag scanning order checked
Progressive frames checked

Average Bitrate 2190

Processor Usage during CCE was around 65 percent

Ended up with 2 bin/cue files
CD1 was 825,575 KB and CD2 was 826,577 KB

While this one was cooking, I burned the images from the TMPGEnc encode and I have to admit, they looked pretty good, very little difference from the ones I had done before at 3 pass VBR. TMPGEnc could be a viable alternative to those who experience the CCE crashes.

I do like the fact that with CCE, I still have some processor left over to surf the web, check email, etc., without significantly slowing down the process. When TMPGEnc was running I had 100 percent processor usage so any other use of the computer would be stealing time away from the encoding.

I will keep these files as well (for a while) so if you want more info, let me know.

Mark

[DDogg]

marks430 or anybody else, It would be great to see a 2hour or even 2hr 10 minute encode done on two disks in a similar fashion as above.

[Mozart]

ddogg's quote:
(...)Note the times of encode are nearly identical.(...)

hehehehe. Note also that he is using dual processors and 4 pass CCE. He said also that tmpeg encoding quality was the same of others 3 pass that he made in the past:rolleyes:

[marks430]

As I get the time, I will do an encode of "The Family Man" which is 2:05:31 and post results here for both methods, however, I will do this one with CCE 3 pass VBR for comparison. I will posts results as the encodes are completed.

The original reason I started the testing was to determine if TMPGEnc would give similar quality and be quicker than CCE on my dual processor system. From the results of my first post, the quality was acceptable, but the speed was not notably faster. And I used 4 pass VBR instead of 3 pass that I normally use. This next encode could be more difficult to maintain the quality. Should have some results tonight.

EDIT: Forgot to change some settings.:( Results tomorrow.:)

[DDogg]

hehehehe. Note also that he is using dual processors and 4 pass CCE.

By using highest and floating point DCT we increase the time of the tmpg encode by about 50%. My thought was we should compare "equal time working" encodes to get a fair judgement.

"2> Speed is similar to CCE - TMPG CQ mode uses a completely different methodology than multipass and uses ONLY one pass. Because of this and even with the above settings that add quite a bit of time to the encode , the speed advantage of CCE is negated and the speed of TMPG CQ encodes will be competitive with a 4 pass CCE encode, if not faster."

....He said also that tmpeg encoding quality was the same of others 3 pass that he made in the past

(Shrugging shoulders) That's pretty subjective and could be effected by a half dozen criteria.

The funny thing about this subject is there seems to be a need for people to line up behind a tool as if it was a matter of loyalty. I think it is all quite silly really.

I subscribe to the concept of looking at the data, via some reasonable attempt at a scientific method so the right tool can be used for the right job.

At the moment I am thinking there may be some validity to thinking TMPG MIGHT do a better job on lower bitrate encodes but more data is needed. There does seem to be a remarkable difference when encoding a DV-AVI. A startling difference actually. Until we figure that one out we will not have a full understanding of the strenghts and weaknesses of both TOOLS.

Let's just keep plugging away with testing and continue to understand the differences between both encoders.

[chainsaw135]

Well dogg i've done another encode with tmpgenc all settings are what has been put forth in the tmpgenc thread. I should post my logs of dvd2svcd to give you some time ideas, but i deleted them its been a long ass day so:( N E ways it took about 7 1/2 hours to complete evolution. Which is Widescreen anamorphic: 1.85:1 length is 102 minutes. If I do a 4 pass with cce it takes like 6 hours roughly. So time really isnt an issue with Tmpgenc at that level. Like I said I delete my darn log files that i wanted to post, but i'm doing shrek now so hopefull i'll keep the log files ect to post:).


Point being so far from tests, If I do the high bitrates cce has so far been better in quality, with a high bitrate of 2,000 or better. I have done tmpgenc with lower then 1,600 and cce also and have found that tmpgenc actually looks nicer, as a whole even tho 1,600 bitrate svcd looks like "poo" in my opinions i'd rather do a vcd if i have to go that low:). Well hope this gives some insight on a different perspective.

[marks430]

Completed the TMPGEnc portion of the longer DVD.

Encoding the NTSC Region 1 Anamorphic DVD "The Family Man". Running time is 2:05:31. VOBs where ripped to HDD using internal routines prior to starting encode.

Total time for SVCD creation, 9 hours 31 minutes 44 seconds.

Encoded at aspect ratio 16:9
Force Film was activated.
Audio set to 192 with no downsample.
Resize method set to Bicubic with b value 0.00 and c value 0.55
Max bitrate set to 2520, Min set to 755 and Max average set to 2230
Between 100 and 130 minutes use 2 CD of 800 MB.
TMPGEnc Rate Control Mode set to Constant quality (QC)
Motion search precision set to Highest quality
DC Component Precision set to 10
Use floating point DCT and No motion search... both checked
Soften Block Noise unchecked

Video Quality Value Used was 83.42

Ended up with 2 bin/cue files.
CD1 was 825,461 KB and CD2 was 826,350 KB

I am now starting CCE encode, will add results when finished.

EDIT: CCE encode now finished.

Total Time for SVCD creation, 9 hours 49 minutes 59 seconds.

Encoded at aspect ratio 16:9
Force Film was activated.
Audio set to 192 with no downsample.
Resize method set to Simple Resize
Max bitrate set to 2530, Min set to 300 and Max average set to 2230
Between 100 and 130 minutes use 2 CD of 800 MB.
Video Encoding Mode set to Multipass VBR 3 Pass
Image Quality Priority set to 17
Anti noise filter checked and set to 2
Linear quantizer scale checked
Upper field first checked
Zigzag scanning order checked
Progressive frames checked

Average Bitrate 1530

Processor Usage during CCE was around 65 percent

Ended up with 2 bin/cue files
CD1 was 825,438 KB and CD2 was 825,481 KB

Will update post once I have a chance to burn cds and watch them.

[sundance]

Is TMPGEnc paying more attention to what you set as max. bitrate (e.g. 2530 kbps) as CCE?
I'm asking this because my recent encode with 2530 kbps using CCE 2.50 resulted in bitrate peaks up to 2912 kbps (verified with Bitrate Viewer) that caused problems with my standalone player...

-sundance-

[Holomatrix]

>>Marks430,
Just wondering about your resize methods. Why are you using Bicubic for TMPEG and Simple for CCE? Should you not be using Bilinear for downsizing? or are you going to try Bilinear on later tests?
Thanks

[mordant]

I don't want to confuse the issue here at all...but...

I'm playing around with a DivX file and AVI2SVCD and TMPGEnc.

I use the settings suggested above except that I choose Simple Resize.

The Divx is about 45min long and is 576x240. I did this once with aspect ratio set to 4/3 and ended up with a fullscreen encode (as I mentioned earlier). The CQ was about 90 and the 2 disks were the right size, but I had wanted a widescreen copy.

On a lark I tried setting the aspect ratio choice in AVI2SVCD to 16:9 instead and it looks like it is going to give me a widescreen encode. I tried doing this with the bicubic resize but the CQ went up to 200 and AVI2SVCD was unable to find certain files. When I changed to Simple Resize I got a CQ of 100 and the encode is about half done.

Why would I get a CQ of 90 on 4:3 setting, and a CQ of 100 on 16:9? Is it because TMPGEnc puts the black bars in at top and bottem and doesn't have to encode as much? Why would using bicubic send my CQ up to 200+ and Simple Resize give me the 90-100 range?

I also did this same movie with TMPGEnc without DVD2SVCD and 2 pass average at 2250 (as per my bitrate calc). It worked and gave a good result (source aspect was 1:1 VGA and gave proper letterbox encode).

Well, sorry if this realy shouldn't be here, but seemed to fit somewhat the discussion.

[mkanar]

IMHO, 'Soften Block Noise' should ALWAYS be set to 99/99 or 100/100 (however high it goes...I forget...). This feature is ONLY used when a bitrate shortage occurs and of course, softing the video to however much is required to prevent macroblocks is preferable over simply seeing macrovision blocks. If no bitrate shortage occurs, then it doesn't soften the video. With "The Emperor's New Grove" I kept reencoding over and over with different settings trying to get rid of the macrovision blocks in a particular 'dancing scene.' The quality was identical in this scene whether I had 'soften block noise' set to 0/0 or 99/99; I discovered that the source video itself contained macrovision blocks!!! This is certainly sad that Disney is unable to encode a macrovision-free mpeg2 stream for a DVD when such high bitrates are allowable. I do wonder if the decoding engine in DVD2AVI is lacking however....

MKanar

[m3taPT]

It would all be fine, it it werent for the fact that i can hear a LOT of difference between 224 and 192. 192 is a no-no for me. Also for a lot of ppl i know, who "caress" a minimum quality.

[mordant]

Well, it worked. Using the 16:9 seemed to give my divx the correct aspect ratio. I would have to say that I was surprised to find my 2 pass VBR beat out the CQ100 encode. I figured 100 meant it was as good as it could be, but there was much more noticible pixilation on the CQ100 than on the 2pass. If this had been from a DVD rather than an AVI it might have been a lot harder to tell.

I'm going to give CCE a chance again and see how it goes.

[marks430]

@ Holomatrix

Just using the settings that were suggested by DDogg for TMPGEnc. The ones I used for CCE are typical settings for my encodes. Since DVD2SVCD included the Simple Resize option I have been using that method and haven't noticed any detrimental effects yet. I haven't had the chance to take a look at the final products for my latest tests, but I am willing to do additional tests if it will help anyone. Please realize that my observations about the video quality are based on my own opinions, which may not be up to the same standards as others, but I will try to give an accurate account.

Let me know if additional testing is desired.

[DDogg]

marks430, I did not want to break your train of thought but it would be better to use the same resizing methods.

Also, since you volunteered, trying a short chapter encode with Simple, simple W/TS 2,1 and also bilinear, using the same cq and bitrate settings as were used in the full encode might be interesting.

Edit: You can do that easily by using crash recovery once you have ran the full size projection routines and have got the CQ. Then just pick a chapter and manually input the CQ. You will not need to use the projection routines again.

First we need to find what looks best in tmpg. Then we can compare that to CCE.

[chainsaw135]

Well i've done more tests with the settings that i guess most of us having been using for tmpgenc, i'm all burned out at this point. My findings still and this is my view could be different from eye to eye but i think tmpgenc at this point in the game is better at the lower bitrate encodes, but the lower bitrates i dont want n-e ways. I guess that leaves me out because i'm all about quality i pay for theses dvd movies "GRIN" so i wanna have a good back up of my stuff:). Well maybe in a week or 2 i'll start more test but for now i'm done and tired.

[Holomatrix]

@ marks430
The only thing that I'm not sure about is what resize method to use, so I hope that you and others that are willing to keep up all this testing which is fantastic can at somepoint post a sticky thread on what your opinion is on what looks best, that includs LowBitrate Bicubic, Bilinear, simple, and Highbitrate Bicubic, ect with TMPEG and the same with CCE.
Thanks again for all the input you guys provide :)

[marks430]

No problem, just have to take care of a few more pressing matters first. I should be able to complete this possibly by sometime this weekend.:) Of course, by that time, DVD2SVCD will have released a new improved version.:D

[DDogg]

I wish some of the folks on UseNet that claim much experience with TMPG would post and advise here. I feel like we are reinventing the wheel sometimes. So, if you are one and need an invitation, here it is :)

Questions and muses:

The only difference here over a straight TMPG encode is the use of AviSynth rather than the direct D2V. I wish I knew what TMPG does if it gets a frame served to it in the same size as its output is set for. Is it smart enough to know not to mess with it? Or, does it reprocess (I don't think so)?
What method is the internal resizing based upon?
What AVS resizing filter provides the best source to work with TMPG (this is THE question, of course)?
Does use of a direct D2V instead of an avs provide any difference in the final encode?
Is the internal resizing faster or slower than avisynth?

Anybody fill in the blanks?

[mrbass]

old review of TMPGEnc
http://tangentsoft.net/video/mpeg/reviews/tmpgenc.html

read this guys conclusions
http://tangentsoft.net/video/mpeg/reviews/conclusion.html

[DDogg]

Here is a huge long post from UseNet. I am a little leary about the correctness of copying his post here without direct permission, but it a nice piece of work and it is very thought provoking so I thought I would risk it. I'll pull it if he has any problem with me sharing it here.

Again, beware of what some call a slight TMPG bias on the poster's part, but this gent (bugs bunny) seems to spend a lot of time testing and really seems to be trying to get deep in the guts of both tools. The last few paras may be of interest for the fast reader. His point in the 8th para about size control being done via a combination of reduction of bitrate and CQ variance might well prove interesting and is something to test more.



After doing some rather intensive analysis of frame-by-frame bitrates over the last several days, I've found some interesting things in addition to the comments posted along with these samples. The realized actual video bitrates,both the maximums and the overall bitrates throughout the film, are effected differently with a CQ render and a multi-pass render.

With multi-pass, using a total bitrate of 2756, audio of 224 and video of maximum 2532 and minimum of 900, just set it and go. Multi-pass uses the specified maximum video bitrate as a hard ceiling and the minimum bitrate as a hard deck. No matter what, it will never get higher than or lower than what is entered for the min/max bitrates. The fewer the number of passes, the closer the high and low bitrates are to the ceiling and hard deck, but the midrange bitrates are all over the place, almost non-existent. And fast bitrate frame changes are not picked up very well. Which is why too few passes with multi-pass produces bad results. Once you hit about 5 passes, the compression is most efficient and the midrange values are more in line, and the peak values become closer to the extremes, high and low.

Regardless of the number of passes, you can see how the top and bottom bitrates are "forced" to get to those points, however, because the goal of multi-pass is a desired filesize. Some bitrates that can and should remain high are forced lower, and some low bitrates that should remain low are forced higher, in order to achieve a specified average bitrate. Beyond 6 passes, at 7 or 8 or more, and it starts pulling back from the extremes and the compression becomes a detriment, a point of diminishing returns, to the degree where frames can get dropped in order to achieve the desired average bitrate, or the fast bitrate changes start getting ignored. This is plainly visible on a frame-by-frame analysis. So, generally speaking, if you use DVD2SVCD with CCE, use a minimum of 4 passes and a max of 6, with 5 being the ideal.

The same exact settings with a TMPGEnc CQ render will result in things that are contrary to common sense, depending on the CQ setting. Depending on the maximum bitrate value entered, the 100 CQ setting forces a higher bitrate than is specified, at least insofar as the peak values are concerned. The maximum, minimum and CQ settings affect each other more directly, and dramatically, than
they will with a multi-pass render. The CQ setting is sort of like the
equivalent of the forced average bitrate in a multi-pass render. So a CQ of 100 will force a high average bitrate. But instead of controlling the average bitrate directly, it controls the level of quantization (compression). The higher the CQ setting, the less compression. More on that in a bit. :)

Think of the max and min bitrates, especially their peak values, as connected to each other with a rigid slider. That slider is the CQ slider. It would be nice if the CQ slider acted in a more precise mathematical manner, say, a CQ setting of 50 would "weight" the average bitrate exactly half way between the minimum and maximum bitrates specified, and raising it towards 100 would raise the
weighted average bitrate. That really IS the way it works, but unfortunately, 50 doesn't correspond to middle of the road bitrate. It will vary with each movie's actual bitrate instead.

With a CQ render and a CQ setting of 100, if you use the max video bitrate for a given audio rate, like 2532 for 224 audio, or 2564 for 192 audio, instead of the video peaking at or near the 2532 setting, the peaks are more in the 2900-3100 range. This in no way runs the render out of specs or anything, as these peaks figures don't actually represent a higher sector rate, but instead the peaks are
realized upon the decoding of the MPEG. But, what it can do is actually degrade the video in certain respects. Rendering a video with high peak values like this, and some of the faster, high bitrate frame changes are not picked up as cleanly as they can be. Changes like a fast moving dark object against a light background, or a movie with lots of very rapid music video-like scene cuts. The highest setting for motion detection should be used with CQ=100 and high max
bitrates in order to help pick up these changes.

Common sense says a CQ of 100 will produce better video than a CQ setting that is lower. But that's not necessarily what happens. With most movies (except slow moving, conversation type movies) a maximum allowed bitrate and a CQ of 100 doesn't give the "variable" part of the VBR enough room to be variable. Turns out, with most films, a setting of 90-95 will give the peak values a little breathing room. Also, by dropping the maximum video bitrate down to about 2500
for 224 audio (2530 for 192 audio) and a CQ setting of 90-95, you'll get cleaner high bitrate changes. These changes are small, and might not even be noticed on a regular TV, but you sure can see them on a high end DVD player output to a HDTV, and you can also see them in a frame-by-frame analysis.

Another thing is, and this is very important to good quality, (Doc, you might want to take note of this for The Guide, too) if you are looking to cut down on the file size, lower the maximum video bitrate and leave the CQ at 90-95,instead of lowering the CQ figure, THEN start lowering the CQ if you need to. At 2300 and below, crank the CQ up to 100. You can drop the max bitrate down to like 2100-2200 even. A higher CQ at a lower bitrate will give you MUCH better video than a high bitrate and a lower CQ setting. For example, for an average movie, a bitrate of 2532 and a CQ of 65 will not look nearly as good as a bitrate of 2200 and a CQ of 100. The lower bitrate and higher CQ setting gives the variable factor more room to be variable on the high end, resulting in better quality.

Compression differences between multi-pass and CQ: A multi-pass render uses compression very dynamically. Meaning, the quality of
the video will move up and down depending on the compression. Viewed as a graph, the bitrate of a movie looks like a very volatile stock graph. Very jagged, up and down, all over the place. A multi-pass render will cause the compression graph to move along this bitrate graph at a different level, but in tandem with the bitrate. The more passes, the closer the difference between the two graphs on the chart (if on 2 passes they are 5 inches apart, on 5 passes they become 2
inches apart, and on 8 passes they lay on top of each other as one).

A 2 or 3 pass render at a max bitrate of 2532, specific points on the graph the compression/bitrate ratios might be something like, 4/2, 7/4, 10/6, 13/8, etc. With more passes, the differences between them get narrower, as the lower bitrates are forced higher and the higher bitrates are forced lower. The same points on the graph 4/2 becomes 3/3 (less compression/higher bitrate), 7/4 becomes 6/5 (less compression/higher bitrate), 10/6 stays at 10/6 (already middle ground), 13/8 becomes 18/5 (more compression/lower bitrate).

With a CQ render, the bitrate stays relatively flat from beginning to end. It's the compression graph is what jumps all over the place. Assuming a CQ of 95, the same 4 points as above might look like 2/6, 4/6, 2/6, 15/6.

Which is better? You decide. Internally, the graphic differences are striking. The end results of both will produce great video. But, it should be noted that, in the same manner that the multi-pass bitrate and compression graphs follow each other in tandem, so do the actual graphs of the CQ render and the original VOB files from the DVD. This means that a CQ render comes closer to representing the actions and changes of the video on the original DVD by a more accurate percentage basis. Interesting.

--
Bugs

[DDogg]

Damn, I thought this would really get some discussion going. Perhaps everybody is off testing :)

[marks430]

Actually, I was doing a little testing.:cool: But my DSL connection also went down.:( You don't realize how much you use it until you can't.:D

So.....the optimum CQ should be between 90 and 95, unless the bitrate falls below 2300. Since I'm already testing, I'll play around with the max bitrate with CQ set at 90 and 95 on my test chapter to get about the same file size as I've been getting. Interesting.

[Slick]

Originally posted by DDogg
Damn, I thought this would really get some discussion going. Perhaps everybody is off testing :)

Maybe there are not as many Tmpgenc fans as was first thought.

either that or they have found out that CCE makes a better job :)

IMHO I would say the latter.:D

[Hook]

DDog-

I tested your settings and am posting results:

Using a dual Athlon MP @ 1.2GHz each with 1024 MB DDR. I encoded the NTSC region 1 DVD "Deep Blue Sea” from VOBs previously ripped using internal routines (should probably add the HDD’s are 100 ATA 75 gig also using Win2k SP2 so I get the true data transfer rate). The length of the video was 1:44:32.

Total time for SVCD creation, 4 hours 57 minutes 13 seconds.

Encoded at aspect ratio 16:9
Force Film was automatic.
Audio set to 192 with no downsample and autodetect for Azid.
Resize method set to Simple (default) with b value 0.00 and c value 0.75
Max bitrate set to 2520, Min set to 755 and Max average set to 2230
Between 100 and 125 minutes use 2 CD of 800 MB.
TMPGEnc Rate Control Mode set to Constant quality (CQ)
Motion search precision set to Highest quality
DC Component Precision set to 10
Use floating point DCT and No motion search... both checked
Soften Block Noise unchecked
No Subtitles
DVD Chapters = PBC Use Selections
Include XML on CD
Include Title Pic and Change CD pic



Video Quality Value Used was 83.88

Ended up with 2 bin/cue files.
CD1 was 825,578 KB and CD2 was 775,063 KB

I burned the images using Nero and the video looks fine on my 55" TV. I’m easy to please when it comes to the quality so I rely on forums like this to give me the best quality / best performance encodes. I do want to point out that since I’ve been using Tsunami the encoding time has decreased dramatically due to the processor usage. TMPGenc uses 100% as opposed to CCE which would use about 50% total.

[DDogg]

Thanks, Hook, great report.

@all
BTW, a couple of people have said the 2.52 version og TMPG was flawed and had a blockiness problem. I don't know if that is true but you might want to upgrade to 2.53 which was said to be better.

[m3taPT]

@ddogg: yes i heard some ppl complain of the same "blockiness" (i myself dunno what the exact problem is, but that was also the word used to describe it. 2.52 is apparently a no-no.

@hook: since u have a 55" tv, you should get some good sound equip and rip at 224.

[marks430]

I finally got done with the testing. As suggested by DDogg, I did a single chapter of the NTSC Region 1 "The Family Man". In my full movie encode to 2 CDs, TMPGEnc used a video quality of 83.42 so I used a Fixed CQ value of 83 to give about the quality as the full movie to the test encodes.

Single Chapter length of 9:06
16:9 Aspect Ratio
Max = 2520, Min = 755, Max Avg = 2230
Rate Control Mode = Constant Quality (CQ)
Motion Search Precision = Highest Quality
Fixed CQ Value = 83
Use Fixed CQ Value checked
DC Component Precision = 10 bits
Use Floating Point DCT checked
No Motion search for still picture part by half pixel checked
Soften Block Noise unchecked
No Deinterlace

I did 4 different test encodes using 4 different resize methods

1. Simple Resize
2. Bicubic Resize b = 0.00 c = 0.55
3. Bilinear Resize
4. Simple Resize with TemporalSmoother Strength = 2, Radius = 1

5. After reading the long post, I decide to set the Fixed CQ to 100 and play around with the bitrate to yield about the same size file as the above methods. After a few tries, I ended up with the maximum bitrate = 1650 and maximum average bitrate =1350.

I took the 5 mpg files that I had and created one bin/cue file using TSCV. I burned this to CDRW using the Daemon Tools copy CD method.

After viewing the 5 different files, the last one (#5) was clearly the worst. At the bitrate I had to use to get the filesize, the picture looked terrible.

Of the 4 different resize methods, #4 had the best quality, but none of them were really good enough, especially on a 48" TV. The chaper I encoded had a facial close up in which the face covered about 40% of the screen. The colors in the cheeks kept changing. The best word I can think of to describe it is blotchy. It kind of reminds me of the way some things look when the video mode is set to 256 colors.

After reading the posts about version 2.52, I will upgrade and try one more test and see if the new version changes anything.

Any questions or suggestions?

Mark

[EDIT] Added number 5 above for clarification. Sorry DDogg:D

[DDogg]

At first I was a little confused what you were saying until I realized the para below the numbers should be number #5 and yes, at those bitrates I am sure it looked pretty bad.

1-4 all should have looked pretty good with the settings you gave them so if you can run 2 or 4 again with 2.53 with the exact same setting we can see if there is a difference.

Then if you could do 2 and 4 with a max of 2350, min of 900, CQ of 95?

[SiliconSoul]

any updates on this?

do the tmpg encodes look the same as the cce?

this is a interesting topic! please do not end it!

now that i got my 80gb hd now i can encode more svcd... i stopped for a while because of no hard drive space!

[SiliconSoul]

dude u suck do u even read! DAMN! he pointed you to this post and u do not read and just repost the same damn thing from your POST! use your brain and eyes!


and when i asked the last time if there were any updates i was wondering if anyone is still testing on this and if so what are the results...

im going to do Riding with boys in cars with cce then tmpg in the next few days... and i hope to have the results... and yes i am slightly comparing them... i hope tmpgenc is close to cce cause CCE bombs on me at different times and stuff like that which sucks... and i hope to use it as primary if its good enough or second if cce bombs

[svcd_neophyte]

can someone test this setting??
Min. bitrate: 900
Audio: 224
Max bitrate and Max Ave:2496
Bicubic Resize 0,70
DC component 10
CQ 100

id tried this setting: seems like the best for me compared to the other one i've done before using tmpgenc. Again this is subjective in my part. Id like to get some feedback from all the dvd2svcdiers that uses tmpgenc. Thanks :)

tmpgenc version used: 2.53

svcd_neophyte

[Slick]

Originally posted by SiliconSoul
do the tmpg encodes look the same as the cce?



IMHO TMPG does not cut the mustard compared to CCE.
even CBR in CCE does a better job.

But if CCE crashes on you then you don't have much choice.:D

but then it is only my opinion.;)

good luck.

P.s. if you are waiting for a reply from DDog, you will wait a long long while.
He resigned about 3days ago.

just read the stickies.

[svcd_neophyte]

> IMHO TMPG does not cut the mustard compared to CCE.
even CBR in CCE does a better job.

With the right setting of TMPGENC i think it can equal or surpass CCE's 4-5 mutlipass vbr. Im not saying that my setting is the best. Ive done a couple of encodes with different setting using tmpgenc and so far i like this setting better. Again, the threads the title is how to get the best out of tmpgenc not which one is better tmpgenc or cce. If you like cce thats fine with me ;)

About DDogg, although he already take a new path in life (which i wish him all the best) there's still lots around that's been doing tmpgenc testing.

Oh one more thing i like about tmpgenc its free :))

svcd_neophyte

[Slick]

Originally posted by svcd_neophyte
> Again, the threads the title is how to get the best out of tmpgenc not which one is better tmpgenc or cce. If you like cce thats fine with me ;)



Yes svcd_neophyte I do know what the thread is about. :rolleyes:

and i'm sorry, but if you bothered to read my post properly instead of jumping to conclusions it was a reply to silicon soul not you!
he asked the question "Quote": "do the tmpg encodes look the same as the cce? "

My reply was that in "MY" humble opinion & "MINE" only "it doesent cut the mustard compared to CCE."
and because it is my opinion neither you nor anyone else will change it. O.K.

I dont have a problem with anyone else using Tmpgenc. it just dosent suit me.:)
when someone comes up with some settings that work (for most movies) I will consider using it.

About DDogg, although he already take a new path in life (which i wish him all the best) there's still lots around that's been doing tmpgenc testing.
yes I know. I am one of them. :D

Oh one more thing i like about tmpgenc its free
So is CCE :D

[SiliconSoul]

anyways im starting my encodes of riding in cars with boys and im going to encode with cce 5 pass and multiple encodes with tmpg with the options recommended here first and newest one and then a few that i think would work. like 2 pass vbr and stuff.

ill be back witht the results

im sad to see dogg leave but i think we can figure out how to make tmpg encodes that are better than cce... will take a lot of testing BUT OH WELL!

[SiliconSoul]

New plan riding in cars with boys does is not worth of a encoder test. im going to use kiss of the dragon or another high action movie like the matrix. there was a thread that recommended certain scenes... ill have to look that up. that was for a encoder speed test but im sure it will work for a encoder comparision and to get the best looking tmpg encode.

[Tomas]

Hi Someone who could help me?
When I encode I use this values:

Resize method set to Bicubic with b value 0.00 and c value 0.55
Max bitrate set to 2520, Min set to 955

TMPGEnc Rate Control Mode set to Constant quality (QC)
Motion search precision set to Highest quality
DC Component Precision set to 10
Use floating point DCT and No motion search... both checked
Soften Block Noise unchecked
TS 2.1
Fixed CQ value=100
edit the avisynth script. Edit as part of video encoding
No Deinterlacing

The picture is fine when slow scenes appear
But when it's high action the picture is very blocky.
I have tried soften makroblocks but it doesen't work.
What are I'm doing wrong?

I hope I'd posted at the right spot this time.

Thanx TOmas

[ShinerSpoon]

Tomas, acording to this quote from above is your CQ set too high? Maybe you might want to try doing the same encode with a CQ of 92 and check "soften block noise"?

Common sense says a CQ of 100 will produce better video than a CQ setting that is lower. But that's not necessarily what happens. With most movies (except slow moving, conversation type movies) a maximum allowed bitrate and a CQ of 100 doesn't give the "variable" part of the VBR enough room to be variable. Turns out, with most films, a setting of 90-95 will give the peak values a little breathing room. Also, by dropping the maximum video bitrate down to about 2500 for 224 audio (2530 for 192 audio) and a CQ setting of 90-95, you'll get cleaner high bitrate changes. These changes are small, and might not even be noticed on a regular TV, but you sure can see them on a high end DVD player output to a HDTV, and you can also see them in a frame-by-frame analysis.

[Tomas]

Thanx! I give it a try:)
/tomas

[Tomas]

HI Shinerspoon, and everyone reading this!
I did the encoding you tiped me.

bicubicresize 0,55
CQ=92 and soften block noise

TS=2.3
max bitrate 2500
Min = 955
DC Component 10 bits
Use floating point DCT
No motion search for still picture...
edit Avisynth script

The result became much better but not good
it is still big blocks in very fast scenes.
In slow its perfect.

I'm going to try to set the min bitrate to 1500
and the max to 2500. TS=2.5, and see if the result became better.

I dont care how long time the movie takes to encode
it's the best result i want!

If I do a bilinear resize do I get rid of the blocks in fast scenes?
Should I change the matrix values? would that help?

Anyone got a tip?

Thanx Tomas

[SiliconSoul]

the ts setting is that the temperal smoother and ur using the top option as 2 and bottom as 3?

i thought defaults were ts 3,2 ....
or is yours at the default and your just posting it as 2,3 instread of 3,2.... that could be a problem too

[m3taPT]

Originally posted by Tomas

I dont care how long time the movie takes to encode
it's the best result i want!




So...., just do a 2pass VBR, set motion search precision to high or highest quality.

CQ is for ppl who are not willing to wait for the best results.

[Tomas]

I have tried the 2pass mode
but it doesent work on my computer
The encode doesent start it just says 0
but the %increase.

I have a Thunderbird 850, 352mb RAM

please help me out.

[m3taPT]

"preview wise", you will only see anything when the 2nd pass starts at 50%. But, if it REALLY stays at 0%, seems like a tmpg problem.

Try this: run tmpg in "standalone" mode, and try to play around with the 2 pass, using ... any mpg, and see if it actually does anything similar, and post the results here plz.

[ShinerSpoon]

CQ is for ppl who are not willing to wait for the best results.

No, I do not think this statement is correct. From my readings on many forums, CQ mode gives the best encode that TMPG is capable of. The problem with CQ mode is file size predictability so many inexperienced people will opt for 2 pass vbr so that they can get an exact file size. This must have been why the author of dvd2svcd wrote all those file size prediction routines into dvd2svcd. It seems those routines do a good job solving the problem of file size projection.

[m3taPT]

Originally posted by ShinerSpoon


No, I do not think this statement is correct. From my readings on many forums, CQ mode gives the best encode that TMPG is capable of.


Although reading the forums is encouraged, maybe you should use more your eyes to watch the real output mpegs instead.

[Tomas]

Thanx for the answer! m3taPT :)

Hi. I have tried to run 2pass in the TMPGEnc 2.53
but i doesen't work.

I think CQ=92 max bitrate 2520 min 955
looks very good when It's not happening
anything in the movie. But when a fast car drives throw
the car get very blocky.
"Yes I'm usin soften makroblocks!"

Doesn't anyone here had tis problem?
Maybe I should test CCE?

How do you encode m3taPT?
Or anyone else to get the absolute best
looking encode in DVD2SVCD?
:D

/Thanx TOmas

[m3taPT]

Yes, ive been testing with a small 15 mns chapter using CQ back and forth, forth and back, and its STILL very very below the quality of a 2pass vbr.

Assuming your tmpg (or OS??) has an undetected problem with 2pass vbr, this is what i advise you:

a) get a diff tmpg version? reinstall?
b) try CCE with at LEAST 3 pass. For perfect quality, i'd go for 5, but 3 does the "trick" on most cases, even with fast action scenes.