At last; Great SVCD from DV with TMPGENC!!!

[Elco]

Thought I should share this info, 'cause it really made me very happy!

So "simpel" and never thought of it;
After some struggling not to loose to much quality from my own made DV material, I discovered how to eleminate blocks and interlace lines in my svcs-mpeg2 files.

This is what I do to get a nice DV-PAL ;
1 Capture my DV material with DV Studio Plus,
2 Edit with Premiere, keeping the same settings on the time line like fields, audio/video bitate etc.
3 Save as DV-AVI
4 Encode with TMPGEnc-i

BUT... a lot of us noticed that TMPGEnc can make or destroy your work, depending on its settings. Next settings gave me a VERY nice mpeg, especially on scenes with a lot of movement;

In TMPGEnc(i) use the standard SVCD template,
unlock (load the unlock template),
best use Constant Bitrate (+/- 2500 kbps),
use non-interlace (encode mode), AND................
use ---DOUBLE DE-INTERLACE--- in the advanced settings.

Man... what a difference!

Some tests with all sorts of VBR didn't make the result any better.

Hope it will help some of you making a SVCD of your home-made DV-material.

Rgrds, Elco.

[NoPlayBack]

YES...some days ago i did also post a question, how to improve quality (reduce blockiness). If i compare the SVCD-image which i downloaded from this site with that, what i can produce with DV-AVI -> Premiere -> Avisynth -> TMPGEnc, there is a big big difference in quality. I also detected the De-Interlace-filter.
Nevertheless...it is better, but still far away from the SVCD-image.
First thing i wonder... i do not change any settings in Premiere when exporting through Avisynth (for example 720*576 picture-size, field-order, audio-bitrate a.s.o.) because i thing they have no influence. Is this right? Or do you set specific things in the export-settings?
Second thing i wonder... can i look with some tool into the mpg-files to check, which bitrate is used? I would like to compare the SVCD-image on this site with my files...

[Elco]

Hi!
Perhaps you can try not to frameserve, but export your time line to an avi-dv file, and THEN encode the new file with TMPGEnc. Just to see the difference, if there is any.
Then use the highest FIXED bitrate to be sure of the quality! Of course your movie will expand some MB's, but on the otherhand, VBR will do one thing for sure; lowering your bitrate on certain moments...
And, it will help you to experiment with other settings, knowing that the bitrate is NOT the point when quality differs.

Rgrds, Elco.

[NoPlayBack]

I must frameserve... if i create DV-AVI, then TMPGEnc shows an error when opening. I tried that already, although i could not believe that should be a different, i wanted to give that a try.
I am using CBR (about 2500), i tried already 2pass-VBR to have same quality at lower filesize...i definitly had not same quality, the result was worse. I found out with using "Highest quality" at motion search precision again the output was less quality in comparison to "motion estimate search".

[Elco]

Hmmm... strange TMPEnc can't handle your AVI-DV file.
I had the same experience with the VBR.

About the "motion search precision", I had a very little bit better result using "High Qality".
The "motion estimate search" was worse using the "double- deinterlace" setting.

[NoPlayBack]

Good hint...
I did use even field for deinterlace... gonna try double and try high quality against motion search precision.
Found some hints on TMPGEnc-BBS concerning error "Cannot open file"...they say that changing priority of DirectShowMultimediaReader to 3 and AVI-VFW to 2 helps. Gonna try that, if it works i will check the difference between frameserving and file.
Did you get a quality with your settings that is comparible with the SVCD-sample on this site???

[Elco]

Hi!
Downloading the vcdcom svcd sample now. Will take about 50 minutes, so will come back one that one later.
This moment I'm encoding some very hard AVI-DV material, like my son playing in the waves in the ocean on a warm and sunny holiday.
You can imagine the movement of the picture all the time, a lot of contrast and me shakin'the cam because of all those beautiful girls coming by...

But the result of the mpeg is, concerning the orignal AVI still VERY good. Some small blocks, but only visable on my pc-monitor. After putting this on a SCVD, the picture on my TV (Philps 100hz!) is smooth, and comparing with SVHS.

[DVman]

ELCO:
Could you compare the difference when you de - interlace and when use
proper interlaced - bottom field first + interlaced output settings?
I thougt that de-interlacing is not a good thing when your source is interlaced.
It also takes longer to encode.

I agree on CBR vs VBR. The file size is a bit larger, but you anyway need 2 CDs
when encoding 1hr tape. CBR encodes faster and at the bitrate of 2500 VBR does
not make much sense.

It would matter when doing DVDs at 8000

<font size=-1>[ This Message was edited by: DVman on 2001-08-15 07:11:47 ]</font>

[Elco]

Hi DVman,
As I learned from experiments with non- and de-interlace in TMPGEnc, this did NOT harm the end product at all.
But the source is AVI-DV, I don't know the result with other AVI material.
The SVCD I make is mostly to be used on a standalone DV player and normal (HQ) TV set.

The combination of Non-interlace and Double Deinterlace did the trick for me. So much better in especially home made DV material; great reducement of those ugly blocks AND interlace strips.

I couldn't notice much difference in encoding time, but why not give it a try?

[Elco]

Hi NoPlayback,

Just made a test with non-interlaced and double-deinterlace for an avi from a disco; very dark and continiously colored spots moving on the dancefloor. My little daughter and girlfriend jumping around on kiddo-disco music. (can you imagine?

Now I tried the "motion estimate search" again, and must say.... not bad at all!! In the colored moving spots, even better. A very little bit more "rustle", but not to much.
Still have to compare it on TV, so not all tests done yet.

The connection with VCDHELP broke up after 87% download of the samplefile. Great!
Can you mail me a very short svcd-mpg2 (2-3 mb) so I can reply you with one so we can compare the quality?

[NoPlayBack]

Right now "playing around".
Will send you a 2.4Mb file.
I'm doing parachuting-videos on mini-dv...so i have also a lot of movement.
I tried your non-interlaced and double-deinterlace...well, that seems not to fit for me. I set encode mode to interlace, and the deinterlace-filter i set to even field. That's a little better for me. I will know try Highest quality.
The settings in TMPGenc did work, i can now read an AVI-file from disk...but no difference to frameserve.
Also i did upgrade to the i-version.
There is one difference.... i am only watching results on monitor, not on a DVD-player and TV-screen.

[Elco]

O.k., played around aswell and made 5 little test movies with non- and interlaced, double-deinterlace and evenfield deinterlace.
Must say that the evenfield gives better sharpness, but somehow much more blocks as well. Especially in the shining water parts.
On standalone DVD player & TV in both ways the blocks were absolute less, and further same result in sharpness difference.
The "fastmotion option" gave a good result in "busy" scenes. During the more "easy ones" parts the HQ looks better at my side.

Soooo.... conclusion SO far;
The de-interlace option is helping A LOT!
However, depending on the material (source and what is filmed) it is to recommend to play around with the sort of deinterlace option (even filed, double or whatever)

Just noticed your mpg comming in, so we'll meet again soon!

[Elco]

Now my eyes are getting red and burning after several hours of starring to the testmpgs, but I really CAN'T see the fifference between the NON-interlace and interlaced setting in TMPGEnc. Somebody knows the trueth about it; when to use one of them?

[NoPlayBack]

Well, i think our mpeg's are comparible...quality is the same. Hope you managed to download the sample-file. If my tools are telling me the right things....the samplefile has lower bitrate, also uses CBR, same framesize, same fps. Also if i simply calculate the relationship between length of video and filesize, the bitrate in the sample must be lower....and still the quality is much better than the mpegs i can produce. Is there somebody who can explain that?

[Elco]

Hi!
This is what I found after searching teh TMPGEnc support BBS about good SVCD's versus out own material.
Someone asked the same question about blocks and other artefacting noise;

PhilPugh ( Mail ) 03/05 (Mon.) 09:45 ( IP:207.18.199.151 ) [ Edit / Delete / Reply with quotation ]
They use expensive hardware compressors for a start - but the main difference is that they use input video that is shot in good light and from a tripod. This leads to no extraeneous movement info that needs to be allocated bandwidth. Also panning and zooming at anything other than very slow rates is out. Just using a tripod will make a significant difference to the final SVCD. Also choice of background etc is crucial. Lots of fine detail (like trees) or reflections off water will kill your final MPEG2. In professional production they will pre-filter this sort of stuff in the video processing - look at commercial DVDs carefully and you will spot that they also have artefacting sometimes.

FWIW why don't you take your archives onto mini-DV/D8 tapes - it's a lot cheaper than CD bit for bit and you can re-edit whenever you want. Just try editing MPEG2 and see what happens to the quality.

[Elco]

This is probably the reason why our shots (waves in water, high movements of trousers hanging out of a plane) gives us a hardtime to encode to acceptable SVCD's.

Enlarging to 720*576 does NOT make much difference.
Encreasing the bitrate DOES, about 4500-6000 solves the problem for 95%. But of course, this is FAR above the SVCD standard.
I really getting closer to the idea that I should burn to DVD... If only the costs...