Quality of converting video formats

[TwoJ]

The question in brief is that there are lots of guides on how to convert various video formats to other formats but I haven't seen one that talks about the video quality of the converted video.

I realize there is no way to determine converted video B is xx% as good as video A but has there been any tests/studies on this that gives an idea about the relative quality.

The whole point of this was some music videos in RealMedia that I would like to convert (can't stand propriatary codecs like RM) but there are 2 constraints - one is that the file size has to stay realtivly small to the original, 2- the resulting video has to be similar to the original in quality.

I realize any conversion of video will lose quality but are there any rough ideas of what kind/how much loss we are talking about and at what file sizes/settings?
I did a test conversion of the 3MB RM file and the software converted it to a 60MB avi (DivX,mp3)! So the video quality was ok but the file size was just too big!

Thanks

[teegee420]

The whole point of this was some music videos in RealMedia that I would like to convert (can't stand propriatary codecs like RM) but there are 2 constraints - one is that the file size has to stay realtivly small to the original, 2- the resulting video has to be similar to the original in quality.

If your desired outcome is the same quality and file size, why bother converting at all? Don't get me wrong, I hate all Real Networks products too, but nowadays you don't have to use them anymore. There's Real Alternative. 'Twas a good day for me when I was able to uninstall all that Real crap.

[TwoJ]

could never decide which was worse Real or Kazaa!

But anyhow its because i have to send it to some people that are not very knowledgable about computers (click the start button in the lower left part of the screen ....) on dial-up, and they have WMP and if at all possible I don't want to have to download/install Real or Real-alternative (so happy since i found it ).

Anyhow I realize converting is not recommended but has there been any tests or recommendations on conversions?

I am not saying that the file has to be the same size but as I said the default conversion program I used did a 3MB RM file to a 60MB avi. So a 2000% increase in file size is too much. What I am wondering is does anyone know if a 200% increased file size is comparible in quality?

[teegee420]

How long is the movie? File size doesn't tell the whole story.

[TwoJ]

The file I was testing is 4:46 m, but I have quite a few of various lengths.

I guess no one has really done any tests in this area to determine correlations between conversions?

[teegee420]

Most people know better than to use real media as source files.

I would say the best you can do is convert it to VCD mpeg-1, which will be about 10mb per minute. At least that way the person you send it to does not have to install anything to get it to play. You should retain most of the quality too.

On the other hand, you can try converting it to wmv and remain under 10mb but quality will definitely be worse than the original rm file.

[TwoJ]

So you are saying that the best I'm looking at is to get a nearly 50MB VCD file from a 3MB RM video?

I realize that RM is an overly-compressed piece of crap but why does a conversion have to ballon in size!
I'm a little more familiar in transcoding audio files which you can expect some increase/decrease in file size but nothing like the order of this for video!

What about if you drop the bitrate to minimum how do you think the file would turn out?

[junkmalle]

If the recipients don't want to install any codecs then your only real (heheh) alternative is Windows Media 9 codec. With that you can probably get a file about the same size without losing too much quality.

[teegee420]

So you are saying that the best I'm looking at is to get a nearly 50MB VCD file from a 3MB RM video?

40 tops if you use DivX witha 96kbps audio bitrate.

I realize that RM is an overly-compressed piece of crap but why does a conversion have to ballon in size!

In your case, yes. You want to maintain as much quality as possible and not force the recipient to install anything, in addition to keeping roughly the same file size. This simply isn't possible with non-proprietary video compression.

What about if you drop the bitrate to minimum how do you think the file would turn out?

In my opinion, the numbers I have already specified are the minimum bitrates needed.

[TwoJ]

Well I still don't understand exactly why it is necessary for a 3MB file to turn into a 40MB file, because i can't believe a RM uses a 96kbps audio stream.

I also don't understand why a proprietary codec can achive such a vastly superior (in compression) in comparison to a non-proprietary.

Just going by Nyquist it should be sufficient to use 2X the video/audio bitrate, or I am way off track on this?

Anyways thank you for all your input.

[junkmalle]

Well I still don't understand exactly why it is necessary for a 3MB file to turn into a 40MB file

VCD quality mpeg is fixed at 352x240 (NTSC) or 352x288 (PAL) -- that's probably twice the dimensions (4 times the data) of your RM stream. VCD bitrate is fixed at a constant 1150 Kbps because it's meant to be compatible with VCD players -- a rather old standard from the day of 1x CDs. And although VCD quality isn't great it's probably far better than your RM stream. Your RM stream is probably less than the 30 (NTSC) or even 25 (PAL) frames per second of VCD.

I hate recommending a Microsoft product, but given your requirements, I really think your only hope is to use Windows Media 9 at the same frame size and frame rate as your RM file. Use a variable bitrate and adjust it to get the size file you want.

Even if your recipients don't have the WM9 codec it will automatically download and install when they try to play the video (assuming they have an internet connection) with Media Player.

Divx, Xvid, 4ivx, etc can probably match the size/quality of your RM file but will require manual download and installation of the codecs.

[TwoJ]

Sorry - i should have been more specific - i can understand why a VCD would have to be that large. I was thinking of xvid,divx, etc.

I understand that wmp would be a good choice but i really don't want to go down the same road of propriatary codecs with little support for tools and lots of support for DRM.

Lets say, at least for argument's sake, that i install the divx or xvid codec on these machines - what kind of settings would be necessary to maintain quality and keep size down? any idea if I would be right about the 2x sampling (encoding) rate?

[teegee420]

I do a lot of DivX and XviD encoding. For example, I record "Family Guy" off of Cartoon Network and convert to DivX. I'm not particularly interested in picture quality, but rather file size. My goal is to get a 22 minute episode down to under 60mb. To do this I must use a video bitrate of 255kbps and an audio bitrate of 96kbps. I decided upon these numbers because this was the most I was willing to compromise video quality.

To give you an idea of what it looks like, here is a 15 second sample. Just right click the link and select 'save target as...". It's about 1mb in size.

Now here is the same clip encoded with enough compression to make it aproximately a quarter of the original size.

Bear in mind that animation is much more forgiving than live action. Do you think you would be content with the results of test2 for your clip? I sort of doubt it.

[junkmalle]

Lets say, at least for argument's sake, that i install the divx or xvid codec on these machines - what kind of settings would be necessary to maintain quality and keep size down?

Use the same frame size and rate as your RM files (you never gave any details about your source files, but I'm guessing you're talking about 160x120, 15 fps). Use 2 pass viariable bitrate at about 100 k bits per second. Examine the results and decide if it's good enough for you. It probably won't look much worse than the original RM file.

any idea if I would be right about the 2x sampling (encoding) rate?

Nyquist tells you that you need to sample at 2x the highest frequency you want to capture. But that refers to the original conversion of an analog signal to digital format. It has nothing to do with compresssion bit rates.

[Alan69]

Well I still don't understand exactly why it is necessary for a 3MB file to turn into a 40MB file, because i can't believe a RM uses a 96kbps audio stream.

I also don't understand why a proprietary codec can achive such a vastly superior (in compression) in comparison to a non-proprietary.

Just going by Nyquist it should be sufficient to use 2X the video/audio bitrate, or I am way off track on this?

Anyways thank you for all your input.

You're mixing apples and oranges, lets bring some precision back into the thinking..

"Just going by Nyquist it should be sufficient to use 2X the video/audio bitrate, or I am way off track on this?"

As guessed you are way way off. Roughly 2x the END bitrate, not the compressed, has nothing to do with what they compress to. Say your RM video is 3MB. Say it's good compression and decompresses to 200MB while it plays. Then you need 400MB to capture it well, and MPEG takes it back down to 60MB for around the same quality. Not MPEG's fault you are calculating off some one else's compressed data and not the original signal you're trying to capture. That's not really everything correct but it's a rough idea why your idea of what it should be is so far off.

There was a big source file. RM threw out a lot of data from the original to make it's file, favoring it's compression. With the same source file, MPEG could likely do well with the file although bigger because of it's different compression. Instead, you're using the RM file. Since MPEG doesn't have lots of good data to make good decisions for it's compression, it has to have tons of extra bitrate to capture what few details are left in the RM file with any accuracy, since what's coming from the RM is geared to it's compression, not MPEG.

RM achieves it's 'superior' compression by throwing out almost all the data and generating a video stream that has a very low faithfullness to original. It's almost impossible for something else to efficiently use it's own compression techniques to reproduce something approaching a good comparison to the original when what it's starting from is such a poor comparison to the original.

Do the reverse. Use a source file, and make a low bitrate MPEG from it. Then use that MPEG to make a RM file. Then compare how crap that looks compared to a RM file of the same size made from the original source. You're leaving out a major part of what happens in transcoding like this, and not paying attention to the fact it'll work just as badly going the other way, simply because you aren't doing it in the other direction. RM isn't magic, you're just feeding crap to the MPEG encoder instead of the good file the RM had to start with..

And as junk said the VCD is generally higher resolution so you're wasting even more bitrate encoding smaller dots than the picture now contains.

"I also don't understand why a proprietary codec can achive such a vastly superior (in compression) in comparison to a non-proprietary."

Might as well wonder why a Ferrari is different than a VW when they're both 'cars'. Assuming they should be anywhere near equal just because they're both codecs is a bit naieve, they're tunded for very different operation. Try and encode a full 720x480 movie in RM and see how well it does at that.

Alan