DivX Pro 5.02 Features

**Enchanter** · 2 Jan 2003, 03:01 PM

The default values are Max = 12 / Min = 2, but I usually set mine to Max = 2 / Min = 2 - is this a good idea? What should I set this to?

No. Leave it at the default values, or use a max quant of 8. The higher the quantisizer value is, the lower the quality is, but the smaller the filesize will be. A min quant of 2 will ensure that it can achieve the best quality where possible.

2) Should I enable Quarter Pixel or not? I've read differing opinions of whether or not it's a good thing. Keeping in mind that I only use DivX for computer and not for tv, would it be ok to enable Quarter Pixel to save a few MB of filesize for extra video?

That's a good question. I've not used the QPel feature of DivX 5 much, so I can't comment on it. However, I've tried the QPel feature of the XviD codec and it worked wonders (a 700kbps video looks more like it is 1500kbps). However, it comes at a price and that is FAR higher CPU power requirement. My P4 had no problems decoding it, but my P3 800MHz simply cannot keep up with the task.

I would like to hear comments from other users (DivX 5's QPel) as well please.

3) Are there any plugins for FlaskMPEG such as de-interlacing filters or any sort of quality filters or even a bitrate calc? Even a GUI would be nice that has these features, as I find Flask's native de-interlacer to be lacking in that, well, it doesn't really do anything!

No. Flask's deinterlacer has always been horrible and I am glad there are better options out there, such as GordianKnot.

But really I seem to be able to get great quality videos (640x288 and fullscreen!) out of flask and all of them end up around 697mb or so, so they fit on a single cd. Maybe I'm just a perfectionist, but after comparing my DivX work to some others' out there, I can honestly say that my videos are of much higher quality and production value (i.e. cropping, size, etc.)

I don't know. I found Flask's resizing algorithm to be horrible as well. I did Monsters using Flask some time ago and even using the Bilinear mode resulted in video that looked jagged and hence hard-to-compress. The final result looked good, but it could be better...

**khp** · 2 Jan 2003, 04:33 PM

Originally posted by Enchanter
No. Leave it at the default values, or use a max quant of 8. The higher the quantisizer value is, the lower the quality is, but the smaller the filesize will be. A min quant of 2 will ensure that it can achieve the best quality where possible.

I would like to add that setting the Max quantizer to 2 locks the codec into useing maximum quality, durring the entire movie. So if you do this you might as well use qualitybased encoding at 100%.

Originally posted by Enchanter
I would like to hear comments from other users (DivX 5's QPel) as well please.

I did some tests when divx5 first came out. Of course my test were based on filesize since I consider myself to be a horibble judge of quality. You can read about it here.

DiVX 4/5 comparison test - Doom9's Forum

http://forum.doom9.org/showthread.php?s=&threadid=19306

DiVX 4/5 comparison test MPEG-4 ASP

In short Q-pel seems to improve compression by about 5%. But it might introduce artifacts at low bitrate.

**HDRed** · 2 Jan 2003, 04:55 PM

I would like to add that setting the Max quantizer to 2 locks the codec into useing maximum quality, durring the entire movie. So if you do this you might as well use qualitybased encoding at 100%.

Well now this can't be true, because I've used Max 2 / Min 2 for 5 DVDs (Stir of Echoes, Taxi Driver, The Untouchables, Sum of All Fears, Gladiator) and each has come in at under 700mb filesize on a single cd with bitrates of 633-724 (can't remember exactly for each one). So honestly I disagree with you there khp.

That's a good question. I've not used the QPel feature of DivX 5 much, so I can't comment on it. However, I've tried the QPel feature of the XviD codec and it worked wonders (a 700kbps video looks more like it is 1500kbps). However, it comes at a price and that is FAR higher CPU power requirement. My P4 had no problems decoding it, but my P3 800MHz simply cannot keep up with the task.

Enchanter, does this mean the machine that's viewing the film will need to be this much more powerful as well, or just the encoding machine? Encoding speed is no problem (1.6a @ 2.56Ghz), but I'd like for lesser comps to be able to watch the movies as they normally would without QP on.

No. Flask's deinterlacer has always been horrible and I am glad there are better options out there, such as GordianKnot.

Agreed, but I simply cannot "get into" GordianKnot. I don't like how it tells you that you need to crop your movie to 330x222 etc. just to get a 60% compression quality with 654 bitrate. Flask works fine for me, but I really do dislike its de-interlacer.

I don't know. I found Flask's resizing algorithm to be horrible as well. I did Monsters using Flask some time ago and even using the Bilinear mode resulted in video that looked jagged and hence hard-to-compress. The final result looked good, but it could be better...

Bilinear? I've never even used the bilinear filter in flask. Being a Photoshop and 3d gaming nut, I learned long ago that Bicubic is much better than Bilinear at rendering anything. Try HQ Bicubic and I think you'll like it. The picture quality I get with Flask 0.78.39 is amazing to fit on a single cd. Or maybe in comparison to some really bad rips out there, I don't know.

There really needs to be a standard of quality out there to help one go by. Also the same thing with size.

**HDRed** · 2 Jan 2003, 05:28 PM

Another quick question:

Double-Buffering - I know it's something to do with using more video ram, is that right?

Should I enable it for encoding?

**Enchanter** · 2 Jan 2003, 07:16 PM

Well now this can't be true, because I've used Max 2 / Min 2 for 5 DVDs (Stir of Echoes, Taxi Driver, The Untouchables, Sum of All Fears, Gladiator) and each has come in at under 700mb filesize on a single cd with bitrates of 633-724 (can't remember exactly for each one). So honestly I disagree with you there khp.

I believe khp meant that the test was carried out at 100% quality (without regard to what the final filesize is to be). With a min/max quant of 2/8, you will get smaller filesize compared to 2/2. In your case, where you manually input the bitrate, having a max quant of 2 will force the codec to compress all frames at this quantisizer value. This is fine until you get to fast-motion scenes where a large amount of bitrate is required. A higher amount of max quantisizer will allow the codec to "smoothen" (whatever I should call "making the video easier to compress") the video. At he end of the day, the final result of 2/8 should look better than 2/2, because of less amount of artifacts.

Enchanter, does this mean the machine that's viewing the film will need to be this much more powerful as well, or just the encoding machine? Encoding speed is no problem (1.6a @ 2.56Ghz), but I'd like for lesser comps to be able to watch the movies as they normally would without QP on.

It won't be a problem with the encoding machine (it will just encode slowly). However, it all boils down to viewing it on the machine (which is one of the main objectives of DivX/XviD) and we all want it to play smoothly (without dropping any frames).

I personally love the XviD codec (I found it even better than DivX 5), but the heavy decoding CPU requirement is holding me back (I have sub-GHz systems too and I view my shows on them as well).

Agreed, but I simply cannot "get into" GordianKnot. I don't like how it tells you that you need to crop your movie to 330x222 etc. just to get a 60% compression quality with 654 bitrate. Flask works fine for me, but I really do dislike its de-interlacer.

I believe it only recommends, not forces. Even if it does not comply with what you want, you can always manually edit the AVS script GK creates and input your preferred parameters.

Trust me when I say everything about GordianKnot shrinks anything about Flask.

Bilinear? I've never even used the bilinear filter in flask. Being a Photoshop and 3d gaming nut, I learned long ago that Bicubic is much better than Bilinear at rendering anything. Try HQ Bicubic and I think you'll like it. The picture quality I get with Flask 0.78.39 is amazing to fit on a single cd. Or maybe in comparison to some really bad rips out there, I don't know.

I did try HQ Bicunic before going to Bilinear. The resulting picture was even worse. The stair-case effect looked so real it hurt my eyes. I thought going to the softer Bilinear would solve the problem. It did alleviate the problem but did not remove the problem altogether.

**khp** · 2 Jan 2003, 07:28 PM

Originally posted by Enchanter
I believe khp meant that the test was carried out at 100% quality (without regard to what the final filesize is to be). With a min/max quant of 2/8, you will get smaller filesize compared to 2/2.

What I meant was that a two pass encoding done at max/min quant 2/2, would be the same as doing a quality based encoding at 100%, as logic would dictate, and I'am pretty sure it worked like this the last time I tested this (with divx4). But after the first pass divx5 apparently decides that the setting is invalid and defaults back to 12/2 (or something like that). This means that the first pass is preformed at maximum quality (I'am quite sure about this, Check the first pass logfile) and then the second pass is done as a bitrate controlled encoding, this is pretty bad since divx5 does not preform bitrate curve compression in the same manner as nandub or xvid.

**khp** · 2 Jan 2003, 07:40 PM

Originally posted by HDRed
Agreed, but I simply cannot "get into" GordianKnot. I don't like how it tells you that you need to crop your movie to 330x222 etc. just to get a 60% compression quality with 654 bitrate. Flask works fine for me, but I really do dislike its de-interlacer.

If you don't need to preform a compressability test to decide which resolution to use when you encode with flask, why do you need to do so with GordianKnot ?.

DivX Pro 5.02 Features

DivX Pro 5.02 Features

Comment

Comment

Comment

Comment

Comment

Comment

Comment