new 3ivx version on tuesday,...

[Selur]

Just visited: www.3ivx.com and read:

3ivx D4 4.5 to be released on Tuesday.

Thought some of you might be interested in this. ;)
(It really did improve a lot since last official 3ivx D4 4.4)

Cu Selur

[Doom9]

there are a great many Tuesdays ;) But if they mean next Tuesday I really need XviD 1.0 for another codec comparison.

[Neo Neko]

Excelent! I hope this has the AR flag support! If so 3ivx could become my decoder of choice for a while!!

[Tuning]

3ivX splitter's bug must have been removed.:rolleyes: (For AAC play back..;) )

[Selur]

@Tuning: yup, it has been fixed

@Doom9: afaik it should be next Tuesday :)
About the comparinson: Xvid devapi4 should 'win' since (at the moment) 3ivx doesn't support b-frames (but b-frames will come in later versions). With disabled b-frames dev4 and 3ivx are on par from what I can tell from some small tests.

Cu Selur

[SeeMoreDigital]

Come on then.... Where is it?

Cheers

[Stux]

Originally posted by SeeMoreDigital
Come on then.... Where is it?

Cheers

It just went GM ;)

We're currently building the release packages

Please Hold :)

[SeeMoreDigital]

Blimey... That was quick!

[SeeMoreDigital]

Are you guys experiencing problems?

Cheers

[Selur]

HAR HAR just downloaded D4 4.5 :)

Cu Selur

[lalo]

woooooooooowwwwwwwwww:p :p :p

[Tuning]

Quote from download page:

Note: The Audio Encoder and Audio Decoder are restricted to 50 encodes or a maximum usage of 30 days.

Does this implies Multichannel AAC decoder expires in 30 days?.
Then
:( :( :( .......

EDIT:
If it does not expire, then we don't need CoreAAC DS filter further. 3ivX audio decoder can handle multichannel LC/HE AAC.

[Atamido]

CoreAAC should handle those fine so I don't know why you would want to get rid of it.

Personaly, I usually detest all-in-one solutions, prefering to keep everything modular.

[Tuning]

What I meant was no seperate installation of CoreAAC is required. I haven't tested yet, but it seems all profiles that coreAAC able to decode is supported by this audio decoder.

Can anyone confirm the decoder expiry in 30 days?:)
Thank you.

[shitowax]

The splitter can still be "opened", so you can use the "all in one" solution or you can customize things to the hell using coreAAC and/or ffdshow or tons of other filters if you want ... :)

The AAC filters are time and use limited due to vialicensing ...

Originally posted by Pamel
CoreAAC should handle those fine so I don't know why you would want to get rid of it.

Personaly, I usually detest all-in-one solutions, prefering to keep everything modular.

[Tuning]

Originally posted by shitowax
The splitter can still be "opened"....
Thanks shitowax, this is what I was looking for.:)
So Re-registering CoreAAC....:D....:p

Who said coreAAC is not required ?... :mad: ..:p

[bond]

great that 3ivx finally arrived!

at least for mp4 i can say it is the best splitter available :)
let the codec testing begin...

to the AAC vialicensing issue:
3ivx' AAC encoder is based on faac, which isnt bad, but there are anyways better aac encoder already available, like quicktime and nero, so...

[Selur]

So waiting for a nice gui which does DVD=>mp4 conversion ;)
(I know how to do it all manually, using Graphedit&Co but a combined Gui would be fine)

Cu Selur

[phrentec]

How do I encode audio to aac using 3ivx 4.5? Can I do it through virtualdub or what other program should I use.

[Ramirez]

GraphEdit > source>3ivx audio encoder>3ivx24 muxer>filewriter.mp4

Seem kind of buggy though, when muxed in an mp4 container (A/V) it crashed every dshow player I tried.

Edit: a simple audio2mp4 conversion also doesn't works (/graphEdit dies at 85% till finish)

[superdump]

Tested the video codec briefly on the matrix revolutions ultra video. I resized it to 672x352 and aimed at 1000kbps (125kbytes/s). This is not a formal test but I found that it was about 1.2dB less in terms of overall PSNR when compared to XviD dev-api-4. On the other hand, it took about 15 minutes to do two passes I think and XviD took about 30.

I've also run a couple of 1 pass quant 2 encodes to see what the max quality is like. I found that XviD was about 1.4dB better in terms of overall PSNR but again 3ivX was considerably faster. 16 minutes for XviD, 5 minutes for 3ivX.

I used the defaults for 3ivX except I changed the force keyframe after value to 250 frames. HPel, 4MV and ASP were checked but NOT adaptive quantisation as I found this lowered PSNR (naturally :)). The dual pass goal was of course best quality.

For XviD I was using a dev-api-4 build from today (2003/11/18) using QPel, GMC, B-frames (1/1.50/1.00), VHQ4, h.263 + trellis, and chroma motion. I used VHQ1 for the first pass to speed things up a little.

I also found something interesting about XviD, I know I'm going off-topic (sorry :)), but I found that when doing the quant 2 encodes, h.263 with trellis was better than MPEG and HVS Best. Go figure. :)

I always use h.263 and trellis anyway.

Have fun!

[Tommy Carrot]

Originally posted by superdump

I also found something interesting about XviD, I know I'm going off-topic (sorry :)), but I found that when doing the quant 2 encodes, h.263 with trellis was better than MPEG and HVS Best. Go figure. :)


Better only psnr-wise or do you really think it's better visually?

[Enigmax]

Good codec.
Direct competition for DivX.
More express, and its quality without B-Frames is similar to DivX with them.

Greetings

[Selur]

Personally I would say if you don't use b-frame in Xvid dev4, 3ivX and Xvid are on par, but enabling b-frames gives Xvid the edge.

@superdump: you did take into account that Xvid 1000kBit != 125kByte in 3ivx since for Xvid k = 1000 and for 3ivX k=1024 ?
(just wanted to check since I first forgot to do this in my tests)

@all:
about adaptive quantizers: for me it looks like they improve image quality (not really caring about PSNR), what do you think ?

Cu Selur

[Stux]

Originally posted by Selur
about adaptive quantizers: for me it looks like they improve image quality (not really caring about PSNR), what do you think ?


Adaptive Quantization should improve average image quality for data rate controlled encodes, but in a Constant Quality encode it will probably subtly damage the average quality, but will result in a large bitrate saving.

Adaptive Quantization is not used when doing a Dual-Pass encode

[Selur]

Adaptive Quantization is not used when doing a Dual-Pass encode
thx for reminding ;)

btw. are there other fetures that are automaticly deactivate during dual pass encoding?

Cu Selur

[superdump]

Tommy Carrot: Just PSNR. I'll have a look with my eyes in a bit.

Selur:

3ivX filesize - 18.4 MB (19,388,416 bytes)
XviD filesize - 17.8 MB (18,767,872 bytes)

I will make an encode with the filesize as the target for XviD later. Got to go to lectures now. :)

[shitowax]

Could you be a bit more precise about what source, encoding and muxing parameters you used ?

Originally posted by Ramirez
GraphEdit > source>3ivx audio encoder>3ivx24 muxer>filewriter.mp4

Seem kind of buggy though, when muxed in an mp4 container (A/V) it crashed every dshow player I tried.

Edit: a simple audio2mp4 conversion also doesn't works (/graphEdit dies at 85% till finish)

[superdump]

Originally posted by superdump
Tommy Carrot: Just PSNR. I'll have a look with my eyes in a bit.

Selur:

3ivX filesize - 18.4 MB (19,388,416 bytes)
XviD filesize - 17.8 MB (18,767,872 bytes)

I will make an encode with the filesize as the target for XviD later. Got to go to lectures now. :)

125kbytes/s in 3ivX then same filesize for XviD. Settings as above.

XviD:Filesize: 18.5 MB (19,406,848 bytes) Overall PSNR: 40.6270 (*)
3ivX:Filesize: 18.4 MB (19,388,416 bytes) Overall PSNR: 39.1327

Constant Quant 2.

XviD:Filesize: 58.8 MB (61,710,336 bytes) Overall PSNR: 46.2148
3ivX:Filesize: 69.8 MB (73,230,336 bytes) Overall PSNR: 44.8619

Again bear in mind the signficant speed difference so as always use the codec according to you purpose.

Tommy Carrot: I had another look at the videos using a stackvertical() in avisynth. Same conclusion as always... h.263 is slightly more blurred and has less blocks, MPEG is crisper but has more blocks. I looked at (*) above and one aiming for the same filesize but using MPEG quants and no trellis. I'm not sure if I'd be able to tell the difference at quant 2 but I'll have a go and edit this post in a bit.

[BoNz1]

Is the news page on the doom9 homepage correct? If I am not mistaken it says that 3ivx 4.5 can use qpel. I think there is only half-pixel and full pixel, ie unticking half pixel enables full pixel.

[shitowax]

We chose to not release our qpel implemetation, but it was part of the package available to beta testers...


Originally posted by BoNz1
Is the news page on the doom9 homepage correct? If I am not mistaken it says that 3ivx 4.5 can use qpel. I think there is only half-pixel and full pixel, ie unticking half pixel enables full pixel.

[SeeMoreDigital]

Hi 3vix guys,

I've generated a few short 2pass VBR anamorphic encodes with 3vix, taking care not to select half pel etc, but I'm confused!

Assuming that it is a compliant Mpeg4 codec. Why does my Xcard not recognise the video stream?

Cheers

EDIT: Sorry, my encodes do appear to work now. However, the Sigma GUI does report that the format is initially unrecognised.

[shitowax]

I cannot help you without more input : what file format did you use (avi, mp4, ...) ? What fourcc (the default 3ivx one is "3IV2", it may be unrecognized by that card) ? what encoding parameters did you use ? I also have no idea if this card supports MPEG-4 PAR ...

Originally posted by SeeMoreDigital
Hi 3vix guys,

I've generated a few short 2pass VBR anamorphic encodes with 3vix, taking care not to select half pel etc, but I'm confused!

Assuming that it is a compliant Mpeg4 codec. Why does my Xcard not recognise the video stream?

Cheers

EDIT: Sorry, my encodes do appear to work now. However, the Sigma GUI does report that the format is initially unrecognised.

[SeeMoreDigital]

Sorry, I was a little vague. I'll check my settings tomorrow.

But what I can answer straight away is that the Sigma Xcard by all accounts the can support MPEG-4 PAR!

Would you have a short 720x480/576 anamorphic PAR MP4 encode I can download to test this card for sure?

Many thanks

[Enigmax]

Changes the FourCC to DivX with the AviC from Nic.
It works without problems.

Great codec 3ivx. I have compressed several films with results phenomenal, better than in the tests.

Greetings

[Selur]

little question:
are the quantizer matices (h.263 and mpeg) the same used in 3ivX ?
(if not constant quant comparisions wouldn't give much clue about quality)

Cu Selur

[Stux]

Just a little point, there are no matrices for H263 Quantization

[Neo Neko]

@shitowax/stux or any other 3ivx crew reading here. I have some MP4 files created with ffmpeg recently. But when I try to play them I get the audio stream only. In graphedit your splitter does not even show a pin for the video! It is there though. And it plays fine on both my linux PCs in Mplayer. Any thoughts on this?

[Stux]

Originally posted by Neo Neko
@shitowax/stux or any other 3ivx crew reading here. I have some MP4 files created with ffmpeg recently. But when I try to play them I get the audio stream only. In graphedit your splitter does not even show a pin for the video! It is there though. And it plays fine on both my linux PCs in Mplayer. Any thoughts on this?

Well, send us a copy of the clip and we'll see what's wrong and try and fix it. Most likely ffmpeg is writing some sort of malformed MP4, which the splitter is not handling gracefully.

Please contact bugs@3ivx.com, thanks :)

[ChristianHJW]

Am i too stupid to find it, or is there no such thing as a 3ivX video DShow encoder filter coming in the package ? I thought as you guys make real nice DShow stuff, that there would be one coming with the package.

I am still trying to find a way to make 'native' MPEG4 MKV files and Tim Jansen doesnt have the time to add a new 'raw MPEG4ES' output mode to his XviD DShow encoder filter unfortunately :( ....

[Stux]

Originally posted by ChristianHJW
[B]Am i too stupid to find it, or is there no such thing as a 3ivX video DShow encoder filter coming in the package ? I thought as you guys make real nice DShow stuff, that there would be one coming with the package.


Not yet ;)

Although you can use the Filter Configurator filter to configure the VfW codec once its in the graph (Video Compressor category)

[bond]

well the sense of using a directshow encoder filter, would be to get rid of vfw ;)

[Neo Neko]

Originally posted by Stux
Well, send us a copy of the clip and we'll see what's wrong and try and fix it. Most likely ffmpeg is writing some sort of malformed MP4, which the splitter is not handling gracefully.

Please contact bugs@3ivx.com, thanks :)

Hmmmm. Well just to let you know I solved my own problem. I am not 100% sure how ffmpeg handles it's bframes. But it gives every decoder fits unless it is being decoded by libavcodec ala Mplayer etc. I was having problems with explorer crashing each time I clicked an AVI encoded with ffmpeg derived codecs where bframes were used. In fact every once in a while ffvfw was throwing divide by zero errors when Bframes are turned on. So I can say with some authority now Bframes and ffmpeg is a no no. Getting to the point I did some encodes sans bframes and viola no problems. Explorer does not crash and 3ivx splits and displays the MP4 with the correct AR. I may still submit the problem sample though. Maybe you guys will be able to make the decoder even better for the next version. :D

[winman]

@Neo Neko

From my experience with ffvfw, ffmpeg tends to generate 2 empty frames at the beginning of the avi when you turn on bframe. This will cause problem for a lot of decoders.

If you delete those 2 empty frames, the avi will play back perfectly fine.

[Neo Neko]

FFVFW sure it relies on VFW to work as it does. But such hacks etc should not be needed in the standalone binary. It should be able to take in and put out frames at it's leasure as there is no external program placing restrictions on it.

[KpeX]

I'm having a problem trying to mux anything with the 3ivx media muxer lately - whatever graph I try to create in Graphedit, avi + aac > mp4, avi > mp4, aac > mp4, GraphEdit crashes as soon as I hit the play button. (Graphedt.exe has encountered a problem and needs to close...) The muxer worked fine when I first installed, I have no idea why it suddenly stopped working. I tried uninstalling and installing 3ivx. Any help or things that can cause GraphEdit to crash like this would be appreciated.

[Neo Neko]

@Shitowax; Stux; or any other 3ivx lurkers ;)

Ok I have gotten the 3ivx filters working on more mundane content created by Xvid or Divx. Still having random issues with ffmpeg created content. Other than that the filters are now working great. But I guess this falls under feature requests. When encoding with 3ivx or ffmpeg you can set the AR in the video stream. You are already doing quite a bit of processing with Divx streams to repack them. Would it be possible for you to add the option for the muxer to set the AR flag for generic divx and xvid content as well. In this way all movies even ones encoded in the past with a slightly off 1:1 AR can be viewed peopperly. Also the format supports author information etc. Will it be possible to set this in the future via your muxer?

[shitowax]

@Kpex: That sounds very strange ... Does it work with other directshow graph (like an avi/mkv/ogm remux) ? Because for the moment, the only real bug we found in the muxer occurs when we try to mux very large bitstream (> 8MBits)... If it worked before and doesn't anymore, you may have an HDD problem ...

@Neo Neko: We've already hesitated to include such feature (PAR overloading) in the 4.5 release, I doubt it will be in 4.5.1 (this one will be released really soon), but maybe in the next one. It doesn't require anything that hard, just a VOL rewrite. For the meta data, we already parse them in the splitter (it works nicely with quicktime files or itunes .m4a in all directshow players), so we propably add their creation via our muxer ... But this would need a real muxing GUI to work easily ;)

[Neo Neko]

Well the sooner you can get PAR overriding etc going the more I am gonna use the filters. I have a mess of stuff that could benefit from it right now. And I am only gonna be getting more. Encoding with square pixels just does not make much sense. Especially when you plan on playing back on devices that can't benefit from square pixels anyway. ;)

[shitowax]

Basically, it would be much more logical to ask that other MPEG-4 encoders support custom PAR settings that asking us to overload their PAR ... Anyway, it's on our TODO list with low priority.

Originally posted by Neo Neko
Well the sooner you can get PAR overriding etc going the more I am gonna use the filters. I have a mess of stuff that could benefit from it right now. And I am only gonna be getting more. Encoding with square pixels just does not make much sense. Especially when you plan on playing back on devices that can't benefit from square pixels anyway. ;)