Ok, I have been making SVCD's for a while now, and I do like their quality. But I have friends and family that I cannot share my video’s with because their DVD players will only play VCD. So I have started to play with creating VCD's, however since they have lower quality output then SVCD its much more critical to find the right encoder and the right settings. To this end I need to find a short test Clip that has a variety of tough to encode scenes.

Does anyone have such a beast?
If Not what kind of clips should I try to assemble?

Thanks. Dave

Footage that included some of the following would be a good test:

- rippling water (or similar fast-moving detail)
- smooth pans (to test for jerkiness)
- slow fade-outs (for temporal artifacts)
- brightness gradients across a plain backgroud, such as a desklamp shining on a wall (for spatial artifacts)
- scrolling white text on a black background

Easy enough to create youself if you have a camcorder....although be aware that ripped DVD footage is MUCH easier to get a good encode from than home-recorded camcorder footage.

Tecold site has such a videoclip
You will need to install/have canopus dv codec
http://www.tecoltd.com/enctest/enctest.htm
(see "how did we run the test")

You might consider trying to make a XVCD so you get a Bit better Quality, Usually if you raise the Bitrate up to about 1500-2000kbs the Quality will But Much better than at the VCD standard Bitrate and Most DVD Players should have no problem playing them, ...

Try the opening credits for the third season of Farscape. That is my test clip because it has lots of light and dark and extreme fast movement.

When you say you need a VCD encoder though, you should be aware, if you are not, that MPEG1 realtime encoders aren't too good unless you pay major bucks for them.

In the consumer arena some are better than others, but they all suck.

I recommend a good MPEG2 encoder or one that creates AVI/DV and encode at max settings, and then transcode using a good transcoder on its maximum settings. That will get you a better result in my opinion.

I do this a fair bit. I have virtually all of the transcoders, and I prefer MCE because it is about as good as procoder, and much faster, and the application is much cheaper. TMPGEnc is okay, but even on max settings I find that TMPGEnc is a bit fuzzier while handling fast movement slightly better.

CCE is out of the running on VCD, an I seldom use Ligos, so I can't say there.

mark

I think it's better to use some of your own video material as test clip because that is what you are going to encode in the end.

I used to Use Music Video"s as test Clips Cuz some Music Videos have a Lot Of Scene changes and Fast motion and Color changes and it is these Properties that Encoders have Problems with, so the More variation in the Video the better test clip it will be for mpeg encoders......Cheers

Thanks for all the suggestions, right now I am testing using the test clip from and apparently it’s a ball buster.

Initially I am going with 1150 bitrate, 29.97, and 224 audio for my tests. I decided to start with TMPGenc right now, because I am most familiar with it. However so far I think the water scene is killing it.

I decided to sneak a peak at how well the Panasonic encoder would go right out of the box. I tried it and it did better on all three scenes in their test clip but the water still was all boxy.

I am planning on trying the Canopus, Mainconcept, and the Panasonic encoder as well, but I am unfamiliar with them. Could someone recommend settings for these to encoders?

Thanks DaveC

This test clip is in the PAL format. So you should encode to PAL format which for VCD is 352x288 resolution and 25 fps, unless you want to test the PAL to NTSC conversion made by the encoder.

It would be interesting if you could measure the SSIM index by comparing original and encoded clip ny using the ssim plugin in avisynth. Read more about ssim here:

http://forum.doom9.org/showthread.php?s=&threadid=61128

Oh you are right I did mispeak, I have been encoding it in to 352x288, 25 FPS. I normally don't, but I didn't to convert it as part of the test.

DaveC