A word on DivX / XviD capable DVD Players

[Darrius "Junto" Thompson]

Doom9,

I saw your home page post today about that MPEG-4 hardware player you bought and think it really shows why we think its important for hardware manufacturers or consumer electronic manufacturers to work with us (DivX) to ensure that content actually plays back well on devices. Unfortunately there are a few companies who don't seem to understand or care about high quality, stability and performance. It really irritates me that many of these companies are marketing how their devices play DivX or XviD content well and this is just not the case. Because we (DivX) knew this would happen we created our certificaton program to try and prevent companies from trying to confuse and misinform users into believing they are getting a DivX player. We decided we had to make sure these devices work the way they should otherwise users would have a horrible experience. People keep asking why devices haven't come out sooner supporting DivX and this is why, we don't want them to unless it works the way we as users would expect it to! And the good thing is we have quite a few partners very close to full certification. KISS-Technology as a partner is a great example. They understand how important it is to ensure their product performs very well and as they would expect it to if they spent the cash on it. Stayed tuned for quite a few more in the coming months.

A few things on why these companies are experiencing problems. These companies are use to creating cheap MP1\VCD, and MP2 Players that can't deal with data rate spikes you see in DivX from using Variable Bit Rate. They can't afford the memory required to handle this influx of data during peaks (overflows the memory bucket) and this and weak processors are the cause for most of these problems. It's combination of weak hardware and little memory yet they don't seem to care that they are selling these devices emphasizing DivX and XviD playback as a key feature when in fact the experience is horrible. That's why on DivX 5.0.4 we created the new Rate Control Algorithm which has something called VBV (Video Buffer Verifier). This ensures that data can still have spikes but in a way that will not create problems on hardware. VBV defines the memory size. For each hardware manufacturer working with us they must be able to handle content within this VBV range ++. This ensures they can handle the memory. We of course also verify performance (frame rate) and most importantly that quality is acceptable. This is a challenged since even though our partners use the authentic DivX Hardware SDK's we created they still have to build hardware so that it can handle the performance required for playback (fast hardware with enough memory). So regardless of who the partner is we don't allow our name to be used on the box stating DivX Compatible or Certified unless they meet these requirements. Of course companies will say plays DivX or XviD but unless you see our ugly stamp saying certified then I would caution trusting it.

I think many people haven't understood the importance of why we created a DivX Certification program. This is exactly why we did this. We knew this would happen. This ensures that when a DivX user user see's an authentic DivX Certification logo on a product they would know that we in fact certfied the product in being able to handle DivX content. If it doesn't have it then I can't comment on its ability to be a good reliable product. If it can't handle DivX it can't handle XviD or possibly many other flavors of MP4 codecs. It's unfortunate there are companies doing this as I'm sure plenty of people are going to be spending money on devices that can't do what they expected them to do when they payed for them. So please make sure either you get good honest feedback on a product claiming it can playback DivX or XviD and of course if it has our Certification Logo you'll know we spent quite a bit of time with our hardware partners making sure we that this works the way it should.

Damn that was a long post! Sorry this subject has just irrititated me lately seeing all these devices and knowing that they would not work making me feel as if these companies are knowingly taking money from people selling something that doesn't work.

Darrius Thompson
Director R&D DivX

[avih]

thanks for the post Darrius.

cheers

[BlackSun]

Basically this is what happened with Sigma Design :o

[Doom9]

Junto: is there a list of officially certifed players somewhere on your site? I'm sure that would help people make the right choice.
Also, can you tell us a little bit more about what is required to get the certification.
For instance, I was told be an elta employee, that a DivX3 capable firmware would be released as soon as DivX Network certification would be complete. This got me wondering, after all, I don't see DXN having a particular interest in playing anything but their own codecs. I also heard from mecotek that their DivX3 supporting firmware would be out pending a DXN certification.
Also, about your profiles. It looks like the home theater profile only allows for bitrates up to 4mbit. What's the reason behind this? In certain scenes, codecs take 6mbit or more if they can, certainly not a bad thing when the VBR algorithm has been properly implemented. And, the elta appears to play Quantizer2 DivX5 clips with GMC just fine (I obviously had to deactivate any profile to encode using those settings).
Unfortunately it appears that if XviD testing were even less important for manufacturers, and since XviD offers a more complete implementation of the advanced simple profile (multiple b-frames for instance), the full use of those features hardly works. I hope that one day there'll be an industry standard for MP4 standalones because the way it looks now, months after the first device has come out, neither codec is fully supported and there's no date for a firmware that can handle all the advanced simple profile features.

[kxy]

Any chance that the DivX / XviD capable DVD Players will support XCD(MODE2_FORM2: 2324 bytes) in the future?

[Darrius "Junto" Thompson]

Great questions. Hopefully I can answer them all adequately.

Originally posted by Doom9
Junto: is there a list of officially certifed players somewhere on your site?

Not yet, but we are working on creating a section for this immediately. Seeing everyone's interest both on our forums and here in the doom9 forums has caused us to accelerate and see the importance of creating a section on our site that lists certfied products.
I'm sure that would help people make the right choice.
Also, can you tell us a little bit more about what is required to get the certification.For instance, I was told be an elta employee, that a DivX3 capable firmware would be released as soon as DivX Network certification would be complete. This got me wondering, after all, I don't see DXN having a particular interest in playing anything but their own codecs. I also heard from mecotek that their DivX3 supporting firmware would be out pending a DXN certification.

We are requiring that for products to be DivX Certified that they have the ability to decode DivX3 since their is so much legacy content in this format. Let me clarify what we have done.
There are 2 levels of requirements for manufacturers working with us. 1. DivX Compatibility and 2. DivX Certification. We plan on requiring everyone to be DivX Certified soon but one of the problems has been the DivX3 compatiblity issue. The 2 Requirements are identical however "Compatiblity" does NOT require the ability to decode DivX3 but DivX Certification does. We created DivX Compatiblity so that we could get devices available to users as soon as possible. DivX3 required quite a bit more work since it is not MP4 so it is technically very challenging and it pushed the timeline substantially for getting DivX\MP4 devices into the hands of DivX users. We decided to create a DivX Compatiblity Requirement so we could certify products that meet a mininum set of requirements minus DivX3 for at least the very near future so there are at least products that can decode DivX4,5 well and at the same time we will move toward ONLY having a "DivX Certfication" of which DivX3 decoding is a requirement.

Also, about your profiles. It looks like the home theater profile only allows for bitrates up to 4mbit. What's the reason behind this? In certain scenes, codecs take 6mbit or more if they can, certainly not a bad thing when the VBR algorithm has been properly implemented.

Very good question. Home Theater profile requires that bitrates with an AVERAGE bitrate of 4000kbps be supported with a max peak bitrate requirement of 10,000kbps for at least a 3 second period. So this is not a problem :) However going back to DivX3 we know there is not a VBR algorithm for DivX3 that implements any type of acceptable buffer model that would well with DivX3 since the peaks can be substantially higher than DivX5 since it does not have video buffer model built into it. As mentioned we built this buffer model into DivX5 so we could ensure high quality playback with hardware requirements that could actually be implemented. So again Home Theater requires you can support 4000kbps with a 10,000kbps max peak for 3 seconds. And as noted on all our certification the only Advanced Simple Profile feature required is B-frames.
And, the elta appears to play Quantizer2 DivX5 clips with GMC just fine (I obviously had to deactivate any profile to encode using those settings).
Unfortunately it appears that if XviD testing were even less important for manufacturers, and since XviD offers a more complete implementation of the advanced simple profile (multiple b-frames for instance), the full use of those features hardly works.

Although GMC is not a requirement it does not mean that some partners will not include suppport for it, however I would bet that GMC will work during decoding for many files using a standard Simple Profile decoder since although you may have GMC turned on it will only be activated in frames where GMC may have a benefit. Many times this will be very few frames if hardly any frames. So we recommend that companies include GMC and Qpel however it is not required since the benefits may not be worth the cost of implementation, especially for Qpel. Qpel is very processor intensive which would require more powerful and thus more expensive hardware to support it even squeezing in GMC adds of course more cost. So we have to make sure what we require the absolute minimum requirement for high-quality video which can also be realistically implemented into hardware at acceptable prices. All these decisions have been more complex than it may at first seem. (oops just came back to "edit" or add this) We *recommend* our partners not only include GMC and QPel but also Ogg Audio support.

I hope that one day there'll be an industry standard for MP4 standalones because the way it looks now, months after the first device has come out, neither codec is fully supported and there's no date for a firmware that can handle all the advanced simple profile features.
Yes I really hope there will be a true standard. It's always interesting in what features are proposed into a standardized codec and what features actually create an improvement in quality that is worth the additional complexity and thus cost in hardware. You would think that more thought would go into this but there is that complex landscape of companies wanting *their* certain features\patents included in a codec. Also I wonder how much thought went into using MP4 for real full resolution full frame rate video when it was created. If much thought wasn't placed on real movie type of video maybe less emphasis was placed on these things. And if you look at MP4 you can see full res, full frame rate, >300Kbps video wasn't emphasized. I need to list at least one example of the pains we see daily in trying to get DivX devices with good support on the market.

One good example of some of this oversite is the MP4 file format. It sounds great and could be great. Try implementing it in hardware :) So as many of you know it is based off of QT. The MP4 file format keeps all of it's important timing data in one location\chunk, etc.. at the beginning of a file and must be indexed to achieve proper audio\video\data synch. A movie a few megs could have a couple hundered kbytes of index date. AVI although old, has simple interleaving which does not require this. So imagine having a 2 hour movie in the MP4 file format and trying to support it in a DVD\DivX type of player. What happens is for proper synch you have to keep all this index data stored in memory during playback PLUS have enough memory to handle playback (video data). The index data alone can be > 8MB and then you add in what is needed for video playback, menus, etc.. and you have a device that is going to be expensive to build because of a large memory requirement. So for the near future there is a cost problem of even building consumer devices at reasonable costs that support everything that is required for the mp4 standard. Many seem to be moving very slow and not necessarrilly thinking of the big picture which includes cost of implementation not including licensing cost and not also thinking about what users need\want at a minimum. Sorry such a long description I just wanted to list *one* example of how complex this has been and how it is not just a basic technology issue we have to try and overcome. This is just ONE example. Damn this has been tiring.... yet fun.

We've tried our best to push things foward more rapidly so that good products are available soon which are *useable* rather than wait for what could be forever before we see good technology put to use. So as you can see we could only wish there was a standard that everyone could agree upon and that not only seems to work well but can actually be implemented in software and devices at acceptable price points with latest technologies.

Anyways we're at least doing one thing.... we're helping to push technology foward given all the roadblocks we've had to work with our partners and the DivX community to help overcome :) There are some great people out there trying to make good things happen. When I say "we" I also mean the hardocore video enthusiast out there also pushing us and others to make things happen.

Hope this helps.


Darrius Thompson
Director R&D DivX

[Doom9]

about QPel and GMC: some would argue that first offering those features, then making them optional in the profiles introduced with 5.04 is a step backward and does not take care of the existing userbase of those features. While theoretically I'm inclined to agree that those features don't do much good, at least in the GMC case, QPel in XviD isn't so bad anymore - it's more a matter of getting used to a different look.
Also, what about XviD? Theoretically a decoder should be able to handle both, but since XviD's implementation is more complete (multiple b frames for instance - I've just verified yesterday that using 2-3 b-frames can lead to slight jerkyness in some cases whereas 1 b-frame, as in divx5, works just fine in my standalone). What does your certification require towards that end? How many consecutive b-frames have to be supported?

And while we're at the subject, do you do any testing for GMC and QPel support? And if you're in the market for a player that does it all (full MPEG-4 advanced simple profiles support, full DivX3 support), what kind of logo does he have to look for? I suppose DivX home theatre profile isn't enough since that only covers DivX5 with 1 b-frame.

Last but not least, the KISS player seems to be having problems with DivX3 clips in some instances.. is it DivX certified?

[snooty]

I don't think the Kiss fully deserves certification at present. It plays all DivX content in the wrong aspect ratio and 3.11 playback is flawed (dropped frames and awful sound).

[Darrius "Junto" Thompson]

Originally posted by Doom9
about QPel and GMC: some would argue that first offering those features, then making them optional in the profiles introduced with 5.04 is a step backward and does not take care of the existing userbase of those features. While theoretically I'm inclined to agree that those features don't do much good, at least in the GMC case, QPel in XviD isn't so bad anymore - it's more a matter of getting used to a different look.

I agree it's unfortunate GMC and Qpel will receive very little support for the near future. Adding these features to hardware unfortunately is to expensive for manufacturers right now and there are very few hardware solutions that can handle these features, especially Qpel, and those that are availabe are definitely not within a price range anyone would find acceptable. Yes Qpel can offer improvements but again it goes back to the cost of supporting it in hardware vs the benefits. Even in software Qpel adds quite a bit of computation. When porting and designing for hardware it gets much worse. Also these are not just DivX products these are our partners products with our endorsement that the product meets a minimum set of requirements. They can decide to support even less than we recommend and not get certification or they can decide to also support GMC and Qpel as we have recommended along with other features. If a full implementation of GMC and Qpel were required and supported I would imagine you would be waiting a minimum of a year before ever seeing support and even then the device would be quite expensive. When processors become available I do hope these features will be supported as it only adds value to the product.

Also, what about XviD? Theoretically a decoder should be able to handle both, but since XviD's implementation is more complete (multiple b frames for instance - I've just verified yesterday that using 2-3 b-frames can lead to slight jerkyness in some cases whereas 1 b-frame, as in divx5, works just fine in my standalone). What does your certification require towards that end? How many consecutive b-frames have to be supported?

Certification does require support for B-frames. However support more than 1 B-frame has not been explicity stated and so then is not a minimum requirement. We were just recently looking at support for multiple b frames and what we have found is a very similar issue.... more complexity when in hardware. There may have to be substantial optimization for some hardware along with different memory requirements for referencing additional b-frames. It's interesting since I would have expected that when vendors built their generic MP4 decoders this would have been taken into consideration. Anyways, support for multiple b-frames is currently not required but we are investigating support. Did you try multiple b-frames on the Kiss Player? Or has anyone else? I would expect it to almost always work but depending on hardware exhibit the "jerkiness" you mentioned due to memory and indexing of frames in hardware. So *currently* 1-Bframe must be supported however that doesn't mean a manufacturer might not explicity support more than 1 properly.

And while we're at the subject, do you do any testing for GMC and QPel support? And if you're in the market for a player that does it all (full MPEG-4 advanced simple profiles support, full DivX3 support), what kind of logo does he have to look for? I suppose DivX home theatre profile isn't enough since that only covers DivX5 with 1 b-frame.

When we certify hardware we have a set of test cases and test clips that we push through the hardware for hardware verification\testing. And so yes there is testing for GMC and Qpel especially if these features are implemented. If they are not this becomes a test for how the prdouct will handle these features (error conditions). We currently do not have a logo or program that designates full MP4ASP+DivX3 because of the reasons described above. It may not technically be feasible for a considerable time on a non-pc device within a price range most of us could afford within the consumer price range that can support these features. Yes this sucks but right now it's a limitation of not having hardware that can handle the power needed for all these features.

Last but not least, the KISS player seems to be having problems with DivX3 clips in some instances.. is it DivX certified?

The current KISS player is NOT certfied. There is not a single device that is DivX Certified but there are sure plenty of devices being sold that claim DivX playback. KISS however will likely be certified shortly although I'm not at liberty to give the approximate date for this. They are working very hard to make this happen. We've held our ground on what the minimum requirements are for certfication so several companies have decided they would just create and release mp4 products since they could not meet certification and now we see that many of these products just don't produce a good user experience. So Kiss is DivX Compatible but NOT certified *yet*. We are getting very close to having hardware available for certfication. I"ll be very honest with a timeline I'm about to give you. Although we are working with just about all major chip vendors there is only 1 major chip manufacturer we are working with that has about an 85% probability of meeting certification in August ...Yes only 1 chip in that time frame... but of course many devices can and will license this chip if they succeed for use in their products. These product would hit shelves around end of summer or early fall. If this does not in happen in August then the next date would be Oct/November with improved hardware. Wish us luck on the August Chip! Oh and of course it is possible that other chip vendors will pull a rabbit out of their hat and also have a solution. But for now there is only 1 that I consider having a probability in the timeframe mentioned. Just by hearing this you might start to realize that this is *really* an issue of hardware not being advanced enough to handle what we are requiring as the minimum for certification. We are working with just about every major chip vendor yet only 1 is currently *close* to having a solution although all are working VERY hard toward a certified chip. We could have certainly chose lower requirements but then the products would have been provided a horrible experience.

As soon as devices are certified you can be sure we will be making announcements since it will really be milestone for us after all this time and work. We acually have a meeting today to discuss getting this information on DivX.com as soon as possible along with a section that list products that are certfied. The logo looks very similar to the badges seen on divxnetworks.com. http://www.divxnetworks.com/certified/about.php

One question I have for everyone. Of course we (DivX) will certfify the product to ensure that DivX certfication is met, however there are other features that are also important in a DivX\DVD\XviD player such as Menu's\GUI, Navigation, Audio Support, Network Connectivity, Qpel, GMC and multiple B-frame support. It could be a conflict of interest if we were to do a review that also gives everyone feedback on all these features. It would be best if this feedback came back from a neutral and reliable third party. This type of review may happen on its own or we could help find the right people, site, or company to start doing 3rd party reviews. This is alread somewhat happening via the forums here and at divx.com but not in a manner that makes it easy for eveyrone to quickly see. Thoughts on how this might be approached?

Darrius Thompson
Director R&D DivX

[Doom9]

however there are other features that are also important in a DivX\DVD\XviD player such as Menu's\GUI, Navigation, Audio Support, Network Connectivity, Qpel, GMC and multiple B-frame support. Well, imho, your certification should include whatever is required for DivX playback. Video without audio isn't so much fun which makes me wonder if there's any audio requirements in the certification. MP3, VBR MP3 and 6ch 448kbit/s AAC should imho be supported on a certified device. And since your codec does Qpel and GMC that's something that you should have a certain interest as well.

as for the rest, this is indeed something that should be handled by a 3rd party, though how exactly I do not know. I have found myself critisizing major websites for their hardware testing because they didn't use the right divx encoding tool, or didn't publish all their parameters (and sometimes their results simply didn't make sense). The community was once working on common speed test sets but that never got anywhere either so I'm not sure if a common test case can be established at all. It takes a strong entity to pull this off and I don't see any that has enough respect where it matters (community and the industry).

[Darrius "Junto" Thompson]

Originally posted by Doom9
Well, imho, your certification should include whatever is required for DivX playback. Video without audio isn't so much fun which makes me wonder if there's any audio requirements in the certification. MP3, VBR MP3 and 6ch 448kbit/s AAC should imho be supported on a certified device. And since your codec does Qpel and GMC that's something that you should have a certain interest as well.

as for the rest, this is indeed something that should be handled by a 3rd party, though how exactly I do not know. I have found myself critisizing major websites for their hardware testing because they didn't use the right divx encoding tool, or didn't publish all their parameters (and sometimes their results simply didn't make sense). The community was once working on common speed test sets but that never got anywhere either so I'm not sure if a common test case can be established at all. It takes a strong entity to pull this off and I don't see any that has enough respect where it matters (community and the industry).

Sorry I wan't clear enough when mentioning testing audio support. What I meant was review of audio playback ONLY in devices that can also be standalone audio players such as KISS for listening to your music collection. So this is seperate from the audio within a multimedia\video file. We indeed certify for audio video synch and reliability. Currently the minimum requirement for support of audio within a DivX file is CBR and VBR audio with the Max being 320Kbps due to the audio spikes in VBR audio. Ogg is recommeneded. AAC is not required. This then goes back to requiring addditional licensing cost for requiring AAC audio by the manufacturer who will already be paying for other video and audio patents. Ogg does not add to cost in terms of licensing so it manufacturers may be more open to implementing it. This adds up and gets fairly expensive. However if the manufacturer has AAC support then I would imagine they would ensure it works with a video file.

Thanks for the feedback on the review issue. So we will only be able to state if the product is certfied and hopefully there will be a reliable source for a true product review. What if we setup a system\site or payed for someone else to do this that allowed for user feedback\ratings similar to how cnet works?

Darrius

[bond]

Originally posted by Darrius "Junto" Thompson
We were just recently looking at support for multiple b frames and what we have found is a very similar issue.... more complexity when in hardware.I dont like this argument :p I think any additional feature will cause more "complexity" in hardware! So were to draw the line?

And a standard for both software and hardware developers already exists:
either the standard mpeg-4 advanced simple profile is supported by both software and hardware developers or not!
I think it would be better to push open standards like mpeg-4 to enable competition and support for all standard compliant codecs, which would be the best for most of us :p !
sorry, but i am sceptical about things like inventing new standards like a "divx certification". what about insuring support for xvid or other mpeg-4 standard compliant codecs, which can be better than divx5?

[Darrius "Junto" Thompson]

Originally posted by bond
I dont like this argument :p I think any additional feature will cause more "complexity" in hardware! So were to draw the line?

This is absolutely the problem and thus the line has been drawn otherwise it could be forever before there are any products available.

And a standard for both software and hardware developers already exists:
either the standard mpeg-4 advanced simple profile is supported by both software and hardware developers or not!
I think it would be better to push open standards like mpeg-4 to enable competition and support for all standard compliant codecs, which would be the best for most of us :p !
sorry, but i am sceptical about things like inventing new standards like a "divx certification". what about insuring support for xvid or other mpeg-4 standard compliant codecs, which can be better than divx5?
Sure, MP4 ASP exists and is a standard but no one can afford to support it or have technology that can support it. MP4 simple profile is good but b-frames adds substantial value and has a low cost for implementation. You have to ask yourself what is the minimum requirement you are willing to live with to have a product within the year vs years. Also remember the decision is not just up to us, there are multiple hardware companies that also have to take a large risk in producing and cost for producing these chips and so there input is absolutely vital and we must agree on solutions. DivX Certification is NOT a standard it's purpose is to easily convey what is supported by the product in regards to DivX and even Xvid video as it exists today.

Would you rather have NO ONE creating any sort of certification so that you might go to the store and buy one device that only support MP4 SP and then another that supports MP4 SP but only at 800kbps, and then another that supports MP4 + Only GMC, and etc...? Look at some devices being sold today.. you'll see in small print that it says DivX\MP4 ASP\Xvid and so you think it's great. You then spend $500 and notice that the small print says MP4 ASP (does not include GMC, Qpel, DivX3, MP4 File Format). This really bothers me that this is happening where consumers are in a way tricked into buying something that might not be what they expected. With certfication you will at least know what to expect.

The hardware is the limitation and so we have to decide what are the best features to squeeze into that hardware. Someone has to try and get everyone on the same page so we can understand what to expect out of these products and what the best features are for the product given the hardware limitations otherwise it will get quite confusing. This is the problem today.

Instead of continuing to go into any more detail here I'm going to spend my time creating an FAQ and Page on DivX.com that should contain all this information and more. This thread has been very useful. Thanks everyone. If there are other open questions please send them to me or post them here and I'll get them added to the site. We will detail exactly what our certification process is on the described page along with many of the questions that have been asked here and of course list the compatible and certified products.


Darrius

[Doom9]

What if we setup a system\site or payed for someone else to do this that allowed for user feedback\ratings similar to how cnet works?
I'm not familiar with how cnet works but I know the system at vcdhelp.com (now dvdrhelp.com) and there's one problem with it.. reports are not always reliable which is kind of a problem. I was once discussing the possibility of a DVD-R compatibility database with user ratings and we got stuck at the point where it came to verify the reliability of a report. I suppose if you can get a couple of guys who really know what they're doing, and give them the tools they need, that could work out. You can still have multiple opinions that way, and the grassrot comments can be made through a forum.

[bobololo]

Originally posted by Darrius "Junto" Thompson
The hardware is the limitation and so we have to decide what are the best features to squeeze into that hardware. Someone has to try and get everyone on the same page so we can understand what to expect out of these products and what the best features are for the product given the hardware limitations otherwise it will get quite confusing. This is the problem today.


The hardware cost / limitation compromise you pointed out at several times appears to be a central concern for player manufacturers. I obviously understand that they all try to optimized the cost to the detriment of mpeg-4 advanced tools especially when they don't bring significant efficiency. However there's still a point that remains unclear to me : the price of these devices ! Here in Europe an elta mpg4 costs about €260 (the same price applies to other mp4 capable players). That's damn expensive compared to a xbox (€200) or a low-end minipc (~ €220) that provide all the cpu and memory (1Ghz+ / 128 MB SDRAM) resources required for honestly decoding MPEG-4 ASP (and much more than EM84xx). Moreover I have the opportunity to see the main board of a KiSS player and there're exactly _4_ components, the Sigma Design chip, a flash and two SDRAM chips. So how much does the sigma design chip cost to justify such a contrast ?

-- bobololo.

ps: btw do you plan to fix the qpel chroma rounding in DivX 5.x and comply with N5546 ? ;)

[BlackSun]

What about Matroska support as well as MP4ff ? You were talking about menus system, matroska will support all of this and even more, is patent free and a system like control track would be very interesting on a set top box player.

[n9801904]

Originally posted by bobololo
Here in Europe an elta mpg4 costs about €260 (the same price applies to other mp4 capable players). That's damn expensive compared to a xbox (€200) or a low-end minipc (~ €220) that provide all the cpu and memory (1Ghz+ / 128 MB SDRAM) resources required for honestly decoding MPEG-4 ASP (and much more than EM84xx).

Remember that the business model for a games console is somewhat different. MS have been willing to make a massive loss on the hardware since they hope to make up for it by selling / licensing games for it.

I guess some of the cost of the MPEG 4 hardware codecs has to be down to the quantities they must be selling. It costs a heck of a lot to develop a chip. I am sure as time goes on and the market grows that the price will fall.

Very interesting thread. Thanks to Doom9 and Darrius.

[zag2me]

The current KISS player is NOT certfied. There is not a single device that is DivX Certified but there are sure plenty of devices being sold that claim DivX playback. KISS however will likely be certified shortly although I'm not at liberty to give the approximate date for this.

Hmmm, does the current K|i|s|s HARDWARE meet the specific requirements of the Divx certification program?

The firmware is still at a very early stage and kiss seem to be spending more time creating new models rather than getting the firmware sorted out. 3.11 support is nice but I would like to see a specific set of encoding tests for divx5 (maybe you could release these so we can test them on our own players). Ie proven encoding methods that will have no problems with playback.

[zag2me]

Some problems that I can see with the kiss players and Divx certification:

1) VBR audio is still not at a stage where you can say it will stay in sync every time. Something to do the number of key frames i think.

2) The search functions still dont work too a high enough standard for every day use. (only timesearch and very unreliable)

3) Divx Aspect ratio's are all over the place.

Three very important features of a playback machine I would think.

[Darrius "Junto" Thompson]

Originally posted by zag2me
Hmmm, does the current K|i|s|s HARDWARE meet the specific requirements of the Divx certification program?

As mentioned already it currently does NOT meet certification requirements *yet*. You mentioned some of the hiccups in your other message and

3.11 support is nice but I would like to see a specific set of encoding tests for divx5 (maybe you could release these so we can test them on our own players). Ie proven encoding methods that will have no problems with playback.
Coming soon on DivX.com. I"m trying to spend more time on getting all this information into one place so everyone can easily find detailed information on what our testing involves for certifying the final product. This information would include properties\settings of the test clips we use.


Darrius

[Darrius "Junto" Thompson]

Originally posted by BlackSun
What about Matroska support as well as MP4ff ? You were talking about menus system, matroska will support all of this and even more, is patent free and a system like control track would be very interesting on a set top box player.

The problem is before a consumer electronics manufacturer spends millions of dollars implementing a new technology in a device or hardware platform they must see that a large consumer base exists before the implement a new technology. They won't implement anything unless they see that enough users are already using the technology in large numbers unless they have VERY strong confidence the technology will gain massive adoption. They have to see how it makes them more money worth their expenditure in resources (cash & people) It's good that it may be patent free but the cost of implementation must be easily offset by an increase in sales. Remember that if you move software into hardware it always requires some type of porting or rearchitecture regarless of how portable we think our software is. It must be redesigned and mapped for hardware, it will require additional memory even when not used (even a 100Kb is a big deal), and then it must go through testing. The cost is high so they always require that their is proof that the technology will help them sell more product. If Matroska does well and there are millions supporting it in a few years I'm sure they would be interested.

Your last roadblock would unfortunately be having Open Sourced code. Yepp it can be an issue. Sorry it's true. I've had to deal with myself more than once! They have to be absolutley sure that your software is patent free. If a patent attorney has not verified this they won't just take your word for this and will fear that if they use your technology that they will infringe on patents. They aren't usually willing to take this risk. So some validation must happen that you are in true patent free by a source they approve of. Sure this happens in a closed source model however it's also a perception issue. The number 1 question asked will be... how do you know that one of your contributors didn't take code from where they work or use to work and contribute it to your project. They are scared because they perceive that you have no idea who that person might be. Anyways, OSS is incredibly great at pushing technology rapidly foward since knowledge propogates quickly, unfortunately once you try and get it into any major commercial endeavor that require major commercial partners there are some roadblocks you have to overcome which is definitely possible. So the first step is make them come to you by creating a great product or technology that demonstrates in can increase sales and these roadblocks are a *little* easier to overcome.

Darrius Thompson
Director R&D DivX
Author of messages that are way to long but hopefully useful :)

[bond]

Originally posted by Darrius "Junto" Thompson
The problem is before a consumer electronics manufacturer spends millions of dollars implementing a new technology in a device or hardware platform they must see that a large consumer base exists before the implement a new technology. They won't implement anything unless they see that enough users are already using the technology in large numbers unless they have VERY strong confidence the technology will gain massive adoption.so to keep away users from using advanced features like qpel or gmc isnt really the right way to create a large consumer base...
its like a vicious circle: consumers wont use these features because manufacturers dont support them (divx5 profiles) and manufacturers dont support them because consumers doesnt use them
but hey, wake up everybody could gain better quality with these features (although they are only usefull if they are fully functional but programmers wont tune the features because nobody use them ;) )

[Charbax2]

I own the Archos JBMM 20. It was once displayed on the divx.com frontpage as the "First DivX consumer electronics". That dates to August 2002.

Though the ArchosJBMM20 still only playsback DivX 4 or 5 at 352x288@25fps or 352x240@30fps.

Sending an email to DXN, I was told the DSC21 and the hardware inside the ArchosJBMM20 probably could be pushed to attain better resolution and framerate for both decoding and encoding (because the Archos can also record 320x240@12fps!). Though no firmware upgrade yet has improved any new resolutions nor DivX3-support on ArchosJBMM20.


Today Archos releases a new generation of portable DivX players. The ArchosAV140 and ArchosAV340. With supposedly the Texas Instruments DSC25 inside. Now it is said the ArchosAV340 can playback up to 640x304@30fps (http://www.hotmp3gear.com/av300review.htm). Though on Archos.com it says 640x272@25fps and 352x288@30fps (http://www.archos.com/products/prw_500542_specs.html). All still onlyu DivX4/5 and still no DivX3.


This current limitation kind of hurts me cause over a year ago, I started reading on http://www.divxnetworks.com/press/pr_detail.php?pr_id=26 that the Texas Instruments DSC25 should be able to playback "full-motion, high-quality video". And after emailing DXN I was told that probably means at least DVD-resolution and actually even more than DVD-resolution.



So. When I invest 600 dollars in a portable DivX player, I would hope for it to playback at least 640x480@25fps and DivX3/4/5.


What I miss is some reliable information source that can inform me of what to expect from the different available processors. A reliable source that could inform me if I have any hope one day see a higher resolution support on ArchosJBMM20 or ArchosAV340 through better firmware updates.


And then I see on Texas Instruments website there is a new processor called DM310: http://focus.ti.com/docs/apps/catalog/general/applications.jhtml?templateId=1010&path=templatedata/cm/general/data/vidimg_digstillcam_tisol_dm310proc

And now that processor is said to do "MPEG-4 video decode at VGA resolution (640 × 480)" and "achieve real-time MPEG-1 and MPEG-4 video encode at CIF resolution (352 × 288)".


So please, one reliable source, inform me if Archos has any good reason releasing the AV340 today, if they shouldn't instead have waited a few weeks or a month and integrate a faster processor into their big-LCD Mpeg-4 portable player/recorders.


I definately miss an awesome independant DivX-hardware reviewing and information website. With latest clever news and reliable estimation of which hardware on the market is COOOL and which is a CAPITALIST-VAPOR-CRAP.

[ChristianHJW]

Hi,

i better apologize in advance for posting here, but please allow me to make some comments about matroska support in hardware devices :



Originally posted by Darrius "Junto" Thompson The cost is high so they always require that their is proof that the technology will help them sell more product. If Matroska does well and there are millions supporting it in a few years I'm sure they would be interested.

1. Currently a lot of time and work has to be invested to make hardware devices read and play DivX/XviD AVI files. This will change significantly if next generation's hardware devices will be out, as already announced by all major Japanese hardware manaufacturer's, as those will have firmwares based on a Linux version like µC-Linux that is used for SIGMA's EM 8500 already. This will open the door for custom and tailor made firmwares offering a lot of new possibilitites, the incredible speed it took KiSS to add Ogg Vorbis audio playback to the DP 450 is a good example for that.

2. I am very realistic about matroska support in earlier devices like the DP 450, its not going to happen. I honestly believed they would run into major problems with handling VBR MP3 in AVI files, and considered matroska as an alternative to them to handle service requests with out-of-sync audio better ( in my idea KiSS and others could have offered a AVI2MKV tool to their customers, and these files would have played in perfect sync then ). The fact they didnt even reply to me this time was a clear sign that the REAL problem for them was the missing DivX3 support, and not VBR3 MP3 sync problems. And as -h couldnt finish his DivX3 to MPEG4 ISO conversion tool before he got 'trapped' by a lovely girl, there was no chance for a DivX3/AVI to MPEG4ISO/MKV/mode2 tool also .... :( . As Junto says, maybe if matroska gets real widespread support due to its advanced editing/subtitles capabilities ( SSA/ASS/USF ) they will think of adding MKV support to their devices. But maybe by that time the first full µC-Linux devices with open firmware will come to market ;) ...


Your last roadblock would unfortunately be having Open Sourced code. Yepp it can be an issue. Sorry it's true. I've had to deal with myself more than once! They have to be absolutley sure that your software is patent free. If a patent attorney has not verified this they won't just take your word for this and will fear that if they use your technology that they will infringe on patents. They aren't usually willing to take this risk.

matroska's framework ( EBML ) is based on XML ;) .. anybody trying to prove to us we are violating his patents would have to face a pretty hard time doing so .... :D ....

[Atamido]

Personally I am a little more optimistic about hardware Matroska playback. Because in most cases you are still decoding the same content, being able to parse it from a Matroska file is relatively simple.

It just comes down to widespread use. If there is a need, they will support it. Gabest was able to create his own parser from scratch in a couple of days, so I am sure they could pull it off the same type of thing relatively easy. (Especially with two example filters)

[outlyer]

Originally posted by ChristianHJW
2. I am very realistic about matroska support in earlier devices like the DP 450, its not going to happen. I honestly believed they would run into major problems with handling VBR MP3 in AVI files, and considered matroska as an alternative to them to handle service requests with out-of-sync audio better ( in my idea KiSS and others could have offered a AVI2MKV tool to their customers, and these files would have played in perfect sync then ).
The idea behind a hardware player is to be able to play your files with the least of work, it's totally counter-productive as a company to say to your costumers they've to transmux their files to play them, so it's way more sensible (again, from a business perspective) invest in workarounds to play in sync those files.

@Pamel:
I'm also optimistic but Matroska parsing is not _that_ easy, and you should now this :devil: Current hardware players are really low on power compared to a full-fledged computer. Of course, the day they want to implement Matroska reading, players will be more powerful.

@Junto:
As I see it you're leaving a lot users aside with your emphasis on hardware player compatibility, I can understand your interests in it, but it would be way nicer to implement some more advanced features in the codec and left users decide if they want the file playable in standalone players (you could pop some bloody red dialog to make newbies understand it :P). Now to make this decision we switch to XviD, I've been using DivX for ages, but after seeing the quality that XviD's alphas with advanced settings give I'm using it for my rips.

Anyway, as said before, I understand your choice and your market focus.

[Doom9]

chris: current players are already based on an embedded linux.. just look at the firmware. But no company will release their firmeware as open source, so the number of users is the only factor that matters. And here we're talking about real numbers, not a few dozen Ks like with tools that we tend to use around here. If 1000 of 50'000 VDubMod users care about Matroska, that is enough for integration. But 1000 out of millions, that's something else entirely, especially when that functionality does cost real money. I'd even go as far to say that it would be foolish to even think about such things while devices don't even fully support all codec features.

@junto: I believe that if you had a certification for all advanced simple profile features it would encourage manufacturers to try and push the envelope. Their customers will see that all devices do not qualify for the best certification, and start asking questions. That will put some pressure on manufacturers to differentiate from their competition and try to create a device that gets the "super certification" which can then be used as a selling argument.

[Darrius "Junto" Thompson]

Originally posted by Doom9
@junto: I believe that if you had a certification for all advanced simple profile features it would encourage manufacturers to try and push the envelope. Their customers will see that all devices do not qualify for the best certification, and start asking questions. That will put some pressure on manufacturers to differentiate from their competition and try to create a device that gets the "super certification" which can then be used as a selling argument.

Yeah I thought exactly the same thing. The problem is again pricing\availability of the hardware. And they are already pushing the envelope just with MP4 SP. I know it sounds as if it should be simple but it really isn't. They aren't just building MP4 they are adding on top of DVD software already occupying space. Many chip companies found out late in the game that their latest and greatest chips didn't have the horsepower they thought was needed for MP4 after they built them so they have had to go back and improve,
modify, etc.. their hardware. That's how hard it's been and this is not uncommon. If there is product differentiation where a product has support for full MP4 ASP and cost >$500 vs one that has SP + B and cost $250 the difference in this one feature may not be worth it to them. This is not our call to make. We just state the minimum for DivX Cert and provide the technology to allow them to meet the minimum plus ASP and more and then verify the quality of the device according to certfication with rigid testing. The risk a company will take over the market because of full ASP support is likely to be low to them and they'll go for volume with the right price point. P.S. I of course could be completey wrong. As hardware does become available we should re-examine having a "Super Certfication". That's an excellent idea.

Let's than assume there is one manufacturer willing to spend the money on full MP4 ASP Compliance. Iwould expect the device would cost a MINIMUM of $500 US Dollars but likely cost much more that this if they were to do this today. Certainly it's possible that 1 company can do MP4ASP soon but IMHO it won't push other companies into also supporting MP4 ASP unless that product demonstrates volume at that $500 pricepoint that would give them better volume\revenue when compared to a $250 device. If they can build a product just with MP4SP for half the price, with less risk, and make more money with higher volume they certainly will. Of course many of us could just let these companies know how important we think these features are but we the technophiles in this forum, although we are considered important, will only amount to very little loss in
volume for them if we don't buy these devices vs the millions of general consumers who know nothing about the intricacies of video technology who will buy these devices and they know this. Ask random people off the street if they like Qpel and see if they look at you funny. We're just a pimple on an elephants ass :)

Darrius

[ChristianHJW]

Originally posted by outlyer @Pamel: I'm also optimistic but Matroska parsing is not _that_ easy, and you should now this :devil: Current hardware players are really low on power compared to a full-fledged computer. Of course, the day they want to implement Matroska reading, players will be more powerful.

Programmable devices, when being constructed for some specific purposes, can sometimes easily outperform even the fastest CPU on the market, just look at DSP's. They are extremely powerful for the specific purpose of crashing huge amounts of data using some form of algorithms ( digital signal processing ). The problem we have here is that the decoder chips of the devices have to care about stream processing also, while in the first instance they should normally concentrate on the decoding ( iDCT, etc. ).

I am convinced next series of devices will have separate chips to care about stream handling, menues, etc, and there will be other chips doing the decoding only. If this architecture is realized, those units wont have a problem parsing matroska files, as they will be built for the specific purpose of shifting data, using big buffers, etc. ( they need this for MP4 also, matroska is only a little bit more complex than MP4, think of MP4 atoms ).

We'll see ;) ...

[fccHandler]

Originally posted by ChristianHJW
think of MP4 atoms
Boy that makes me mad; it's the second time I've seen that mentioned in these forums. For the record, "atoms" are purely a QuickTime construct (they're not too different from "LIST" chunks in AVIs), and they are not part of the MPEG spec. I hate the idea that people are beginning to equate QuickTime MP4 with MPEG-4. Aaargh!

...
Sorry guys, just had to get that off my chest... ;)

[bobololo]

Originally posted by fccHandler
Boy that makes me mad; it's the second time I've seen that mentioned in these forums. For the record, "atoms" are purely a QuickTime construct (they're not too different from "LIST" chunks in AVIs), and they are not part of the MPEG spec. I hate the idea that people are beginning to equate QuickTime MP4 with MPEG-4. Aaargh!

I'm not sure to well understand what you mean but AFAIK, the MPEG-4 system part (14496-1) defines the mp4 file format that is strongly based on QuickTime mov format.

-- bobololo.

[fccHandler]

I'll try to explain. There is a new file format going around called MP4, and there is (in my belief) a common misconception that these are MPEG-4 files. From what I've seen, they are simply .mov files containing a video stream with the "mp4v" FOURCC. I suppose this is some (proprietary?) QT variant of MPEG-4, however other "compliant" MPEG-4 decoders (like DivX and XviD) won't decode the content. I'm questioning to what extent the "mp4v" stream is really MPEG-4 compliant.

My copy of 14496-1 doesn't mention the word "atom" anywhere, and the syntax it describes doesn't look anything like a QuickTime .mov. Do you have a link to more info? How does MP4 == MPEG-4? I'd like to understand where the "misconception" is coming from.

Sorry again for being way off topic. This MP4 thing has been bugging me for some time...

[Darrius "Junto" Thompson]

Originally posted by fccHandler
I'll try to explain. There is a new file format going around called MP4, and there is (in my belief) a common misconception that these are MPEG-4 files. From what I've seen, they are simply .mov files containing a video stream with the "mp4v" FOURCC. I suppose this is some (proprietary?) QT variant of MPEG-4, however other "compliant" MPEG-4 decoders (like DivX and XviD) won't decode the content. I'm questioning to what extent the "mp4v" stream is really MPEG-4 compliant.

My copy of 14496-1 doesn't mention the word "atom" anywhere, and the syntax it describes doesn't look anything like a QuickTime .mov. Do you have a link to more info? How does MP4 == MPEG-4? I'd like to understand where the "misconception" is coming from.

Sorry again for being way off topic. This MP4 thing has been bugging me for some time...

bobololo is correct. The MP4 file format is based off of QuickTime and it certainly is mentioned within the spec as so. And Atoms are a key part of this. It is 14496-1:2000 and not 1999. It is in section 12 and 13 all over the place. In MP4 a meta data structure (object) is called an Atom in QT terminology. These .mp4 files are certainly MP4 files. In .mp4 files the atoms define the meta data structure of the file. As per the spec the meta-data is contained within the meta-data wrapper (the movie atom) the media data is contained either in the same file, within the media-data(atom)s, or in other files. The media data is composed of access units; the media data objects, or media data files, may contain other unreferenced information. Oh and of course it doesn't playback with XviD or DivX. DivX and XviD ONLY decode the frames after they are demuxed from the file format. Decoding is a separate function from parsing. Neither of us distribute an MP4 file format demuxer. If we did it would work. So when someone says a video file is MP4 you must clarify if they mean just the video stream or the entire file to include file format. In AVI the DivX video stream is certainly MP4 but the AVI file format does not make it a fully compliant when the file format is taken into consideration. Ok don't expect me to answer this type of question often since this info can be found on plenty of places on the internet. MANY! However it's Friday and I'm ready for the weekend. Here is what came up in Google from Apples site:

http://developer.apple.com/techpubs/quicktime/qtdevdocs/whatsnew.htm
7/13/2000 -- A new version of the QuickTime File Format is available in PDF format. This is a major update and revision, including documentation of new QuickTime atom types and many new code samples. This is the version of the QuickTime File Format which is incorporated in the MPEG4 specification.


Darrius "Junto" Thompson
Director R&D DivX

[fccHandler]

Thank you Darrius for taking the time to reply in such detail. Seems I'm stuck with an old version of 14496-1 after all.

Originally posted by Darrius "Junto" Thompson
Oh and of course it doesn't playback with XviD or DivX. DivX and XviD ONLY decode the frames after they are demuxed from the file format. Decoding is a separate function from parsing. Neither of us distribute an MP4 file format demuxer. If we did it would work.
Well, I tried something like that a while back. Essentially what I did was I used GraphEdit to attach the outputs of the "QuickTime Movie Parser" filter to the "AVI Mux" filter, to see if I could remux the movie data into an AVI container. That method does work with older common codecs like Cinepak, and it kinda seemed to work in this case. The AVI showed the correct "mp4v" FOURCC from the QT file, but I couldn't get it decoded by DivX or XviD after trying all the common AVI MPEG-4 FOURCCs I could think of. (IIRC even QuickTime wouldn't play it back.)

I'd love to hear any comments you might have about that, since you obviously know a great deal more about the subject than I do. But I already feel bad that I've dragged your thread off topic, and anyway it is Friday! Enjoy your weekend. :D

[outlyer]

Originally posted by fccHandler
Well, I tried something like that a while back. Essentially what I did was I used GraphEdit to attach the outputs of the "QuickTime Movie Parser" filter to the "AVI Mux" filter, to see if I could remux the movie data into an AVI container. That method does work with older common codecs like Cinepak, and it kinda seemed to work in this case. The AVI showed the correct "mp4v" FOURCC from the QT file, but I couldn't get it decoded by DivX or XviD after trying all the common AVI MPEG-4 FOURCCs I could think of. (IIRC even QuickTime wouldn't play it back.)
I also tried what you did some time ago (before QuickTime had an MPEG4 codec) and never tried again, I can't help here.

Anyway I've been writing some code to transmux from mp4/mov (as said, almost identical) to avi. A file using QuickTime's MPEG4 codec once transmuxed is perfectly playable by both XviD and DivX.

[fccHandler]

Originally posted by outlyer
A file using QuickTime's MPEG4 codec once transmuxed is perfectly playable by both XviD and DivX.
Wow! You mean you've actually done this with your code?

[outlyer]

Yes, and almost all of what is needed is part of mpeg4ip ;)

[fccHandler]

Thanks for that tip; I'll check it out. And thanks to everyone for setting me straight. I'm not mad anymore and I've learned a lot today. I'll leave your thread now (I've got a lot of catching up to do apparently...) ;)

[Doom9]

If they can build a product just with MP4SP for half the price, with less risk, and make more money with higher volume they certainly will. Of course many of us could just let these companies know how important we think these features are but we the technophiles in this forum, although we are considered important, will only amount to very little loss in
volume for them if we don't buy these devices vs the millions of general consumers who know nothing about the intricacies of video technology who will buy these devices and they know this. Ask random people off the street if they like Qpel and see if they look at you funny. We're just a pimple on an elephants ass
Of course they don't know, but then they download a movie off kazaa they cannot play on their $250 standalone and then another one, and after the 3rd one the device goes back to the store and they'll get an XBox with a modchip or a DVD recorder for their PC ;) People don't have to know about QPel, they just need a clip they can't play to realize something is wrong. And since you're already testing what happens with GMC and QPel clips it wouldn't require additional effort for the super certification.

I don't think I overestimate the importance of this community on the large scale, but then again, right now those devices are only sold to people who know what they're doing. If early adopters tell others to hold out, they most certainly will (I got my whole circle of friends to get a DVD player within half a years time, they'd still be using VHS if I told them DVD sucks... none of them is very tech savy), and the ones that just pick up a player because they know they can get DivX from their friends or off kazaa will sooner or later run into those files that will make them return their player because a regular player only costs half the money, but does what its supposed to with all the discs it's supposed to play. I guess I shall make it an objective to push the proliferation of QPel to put additional pressure on device manufacturers so that they get their act straight and release products that actually do what they're supposed to do. I was pretty pissed when I could only play one out of 13 movies I tried out the first day I got my player. And I knew what I was getting myself into.. if I hadn't, the player would've gone back the very same day.

[Darrius "Junto" Thompson]

Originally posted by Doom9
Of course they don't know, but then they download a movie off kazaa they cannot play on their $250 standalone and then another one, and after the 3rd one the device goes back to the store and they'll get an XBox with a modchip or a DVD recorder for their PC ;) People don't have to know about QPel, they just need a clip they can't play to realize something is wrong. And since you're already testing what happens with GMC and QPel clips it wouldn't require additional effort for the super certification.
You are right, the experience will be horrible if there are enough users that have content using Qpel or GMC and a newbie has no idea their file has this feature and the product just doesn't work. The additional effort on testing is already done it's the lack of hardware that exists so again we'll have to wait unless we can convince someone to build a high-end device with these features. Let's give it yet another try. Your right, if I find that a small percentage of my files can't play then I'm going to be very irritated if I expected them to playback. Support will happen just not any time soon though. But the newbie is going to have it worse because they won't know a thing about the file and why it won't work. That's why we create the DivX Profiles to help alert the user of what features will and will not work with devices.
What percentage of files would you guess use GMC and Qpel? As you mentioned once they are on the market they will also want to find a way to differentiate themselves and by saying they can support all features they likely will. What would Super Certfication require? Keep in mind a hardware company must also see the value of this since it would take some time and be priced at the high end. MP4ASP Full Res, Full Frame Rate, MP3 Audio, AAC, ... and? Anyways this is why we have DivX Profiles it ensures playback.

Darrius

[Jordan "L0g05" Greenhall]

Hey guys, I've been sitting next to Darrius watching him type in all of these long posts and had to jump in on this discussion. I've actually been up to the elbows in this excrement - so I'm sensitive to all of the thoughts that have been brought-up on the board.

Let me set-out our thinking. First-off, we've been working with the chipset companies and the CE OEMs for more than 1.5 years now to get this stuff out there. They were all going to go to market with MPEG-4 Simple Profile - we held firm on full ASP for a long time (and suffered great pain at their hands as a result) - but at the end it turned out that getting Qpel and GMC onto DVD chips was just a non-starter. The cost of silicon was way too much and no-one was going to be able to do it. So we looked at the bang-for-the buck of GMC and Qpel, the amount of content actually created with these tools and pulled the plug on requiring support for those features.

For X-mas of 2003, a DVD chipset costs about $6. This leads to about a $45 bill of materials for a DVD player and this is why you can buy a low-end DVD player for about $50 (i.e., they are sold at zero margin and this *doesn't* include license fees to the DVD forum that cost about $20 a piece). Right now a chipset that can support Qpel would be about $40. Each dollar in a chip = $4 - $5 in a player, so you are increasing the BOM by about $160. This leads to a >$200 DVD player out the gate - and that doesn't include full GMC for which it appears nothing short of a Pentium can do the job. So to include Qpel would require a $250 price point - just to get to zero margin and no-one will sell a DVD player at that high a price to get no margin (i.e., the cheap ones are loss leaders). Realistically the price point of a player with that kind of a chip would be in the >$400 range.

Flip the economics around. The chip guys sell to the box guys - and it costs $1 million dollars just to run a fab cycle (and that doesn't include the cost of the silicon, chip design, QA, etc. So they don't make a chip unless it will move *lots* of units. The bigger the chip guy, the larger the number of units they want to sell before they even think about a feature. So if a chip guy is thinking about adding a feature (say, DivX) to their chip, they have to think about how much more expensive it will make their chip and compare that to how much more a CE box maker would be willing to pay. Qpel is *big* on-chip. So a Qpel-enabled chip will cost more than a non-Qpel chip. Will the box maker be willing to pay that much more for it? The answer appears to be a firm - no. We tried, and pretty much everyone shot it down. Heck, we are *still* fighting with major OEMs who want to dumb-down the existing Certified requirements

Our goal is to get players out to people for a reasonable price - you should be able to see DivX Certified players on the market before Xmas at below <$200 - maybe even below $150. Network-connected players (which rule, by the way) might come in below $250. That meant convincing DVD chipset makers (who are extremely conservative, do one chip every 1.5 years and do the *absolute minimum* to sell into the market) to support DivX. And that meant dropping Qpel and GMC.

Some things to think about: even if we had been able to include Qpel and GMC, there still would have been a reasonably large % of DivX and XviD content that wouldn't have played on these devices.

They don't support ADPCM audio - and certainly not "DivX Audio". Which flavor of Qpel did you use (at last count there are three - and it isn't easy to handle all three in hardware). Bad AVI encodes or SMR 3.11 spikes into the 10 Mbps ranges are going to break even the most robust silicon. And to the extent that XviD forays outside the MPEG-4 specification . . . all bets are off.

So, the bottom line is that there is going to be some level of background noise of files that just won't play on a "CE" device. Qpel and GMC will fall into that noise - and the cost/benefit cuts against those features.

You can certainly get players that will play almost everything. I would recommend a 1.5 Ghz PC running Linux. A modded Xbox is also good (I hear). But for the bajillions of people who are looking for CE (and don't own a soldering iron) - I hope that DivX Certified will make them very happy.

Jeez - "super certified". There are a couple of chip companies who are targeting Qpel and they have been hitting me on just this issue. This is incredibly tough. Is the addition of what ultimately is a minor feature enough to justify a completely new profile (and all of the consumer-education that would require)? You also have to take into account testing. We have people dedicated full time to testing players that are looking to be "Certified" to ensure that they can do what they say they can do. "Super Certified" means an expanded test profile.

I guess here is the question: how much more would you pay for Qpel?

Interestingly on the audio side, it is all over the freaking board. Some guys support Ogg, AAC and the whole list. Others bite your head off if you even mention supporting another audio standard. What you have to know is that it is a bitch to get this stuff into firmware and tested for CE players and the chip guys absolutely hate supporting anything they don't have to (and in the case of AAC, they really hate paying the additional $$ required for the license). Believe me, I'd love to switch to either OGG or AAC - but then you'd be in even a bigger quagmire. At least every DVD player can play MP3 these days.

[Doom9]

To come up with an idea what a full ASP certification would include I'd first have to know what the existing certifications involve... but basically everything that is in the advances simple profile (maybe I should try to get a copy of the standard document one day just to get a full overview of what's there).. GMC, QPel and multiple b-frames.

As for the price point, I paid $370 for the elta player, which is more than I've spend for a CE device in a long long time. Okay, the store was somewhat overpriced but still.. that is an expensive player. In fact I believe they only had like one or two players that were more expensive (and that didn't play divx). Looking around, some of the announced DivX capable players in Europe are available for 200€ and that's considered a bargain.

What is the difference in price between chips that support QPel and those that don't, for instance the Sigma 85xx versus the 86xx series? In any case, some chips are available, it's merely a question of using them instead of the cheaper flavor.

"Super Certified" means an expanded test profile. according to Junto I thought GMC and QPel was already being tested so the extra effort would be trivial.

I guess here is the question: how much more would you pay for Qpel?for QPel alone probably not that much, but for the capability to play ALL DivX and XviD files that could be worth a sizeable premium.

I don't know how many clips use the advanced simple profiles features. I know that when you're doing XviD 2 b-frames is standard and with QPel getting pretty good I think we'll see much QPel clips in the future. You're more interested in DivX.. there I can only say that in my guides I've always suggested the use of GMC, but not QPel. I thought you had a poll on divx.com about the use of the asp features once, what was the result of it? Also, I just had a quick look in your DivX hardware forum, and I saw a lot of "can't play this can't play that" which isn't very encouraging.

BTW, XBox modding doesn't require any soldering skills anymore, there's modchips that you can just plug in. If it weren't for the 2 GB ISO limitation and the noise I'd have gotten myself a dedicated XboX as DivX player a long time ago.
Last but not least, at least my player has XviD written all over it which will be yet another noisance when people try out those clips and find they don't work either. I'll have to talk to the XviD developers to know what they think about that.

[shlezman]

"Super Certified" = divx certified

Its unfair that DXN take advantage of their popularity to get into a "new" market such as certification. MPEG4 defines profiles and levels, and DXN takes advantage in a "hole" of the spec to define their own mid-profiles, which can be called also "Super Certified" Simple Profile.

Hopefully some HW company will release an SDK for their "video playing dedicated machine" and the OSC (Open software community) will take it from there (just like with xvid). Or even better ...

Maybe DXN can collaborate with the community and reconstruct their relation by forming and "Open Certification Programs" that will supply tools and a common firmware development environment for such devices. Or even better ...

The OSC can do that by itself.

Reading back my post, makes mw wonder ... did I have too much coffee today :confused:

[CruNcher]

@shlezman

i agree what DXN is trying now is ridiculous for me it looks like they try to get a wider support and for the moment they only get this from Sigma, to move sooner or later to DivX 6 which will not be based anymore on the Mpeg4 standard what Gej said at the Cebit on the SAD both. For me it is unbeliveable that DXN at the moment has the power in the Mpeg4 Industry Forum to actually reform the standard that where set along time ago and this was Mp4 + ACC and nothing else sorry Darrius "Junto" Thompson that i can't share your opinion about how you see the manufactures of standalone players sure it's right that they will first support things which are cheaper but in first place what the customer want's is the most important and i can't belive that the customer want's it your way and i think other Mpeg4 Industry Forum Members as Ahead/Ateme will show you that this can also be achived without this "DivX Certified campaign"


Maybe DXN can collaborate with the community and reconstruct their relation by forming and "Open Certification Programs" that will supply tools and a common firmware development environment for such devices. Or even better ...


i think thats what the M4IF was created for which also DXN/Sigma are members of


The MPEG-4 Industry Forum is a not-for-profit organization with the following goal:

To further the adoption of the MPEG-4 Standard, by establishing MPEG-4 as an accepted and widely used standard among application developers, service providers, content creators and end users.


so if you want to bring it to the "end user" do it the right way in cooperation with the other forum members or do your own thing and leave thats my opinion.

[shlezman]

a few cups of cofffeeeee later ...

@ CruNcher
(there are members only in "egged" - Israeli joke)

Being a member means you paid a few bucks, that's about all. You can be a member and fight the group ... that's called buisness. No one can kick you out as long as you pay.

As for the MP4IF its an excelent organization with a small problem called politics. That's why (I'm only guessing) the "standard" dosnt support MP3 and OGG and many other great stuff. DXN are fully aware of that (being members) and they also know that the amazing amount of video content that already exists in "their" format will never fall into the standard's boudaries, that's why DXN make their own certifications, cause no one else will certify MPEG4 SP with B frames and MP3 audio packed in an AVI but everyone wants it. Thats unfair that DXN will get royalties (paied by consumers) for boxes that carry a nice "Divx-Inside" sticker, when all they actually done was printing that sticker.

If DXN had certified boxes that play Divx 6 (as you said), then I'd say go for it, you earned the sticker (and the royalties that come along with it), but after all we are talking MPEG4 not some propriatry codec.

[Doom9]

@cruncher: I believe you're barking up the wrong tree here. DivX certified means just that.. it doesn't mean "supporting MPEG-4 level xyz profile abc". It's their right to certify hardware that supports their implementation. Right now I don't know how the certification works so I rather no comment on it, yet they don't claim that they're certifying any generic MPEG-4 compatibility.

To my knowledge, the MPEG-4 standard does not contain any delivery mechanisms. For DVD you have the DVD forum which has defined how a DVD should look like and how it should be played. In order to put a DVD logo on your DVD player, your player must adhere to a standard that goes way beyond an MPEG-2 level and profile. Just to give you an example: There's different audio formats (MP2, PCM, AC3, DTS), there's the definition of the VOB files, their size and how they are to be put on a disc, there's the filesystem of the disc, etc. etc. Now as I said I might be mistaken on this but I haven't found any similar standard for MPEG-4 delivery on CD/DVD media. Thus, everybody is cooking his own soup so to speak. Right now there's chaos when it comes to storing MPEG-4 video. Take XviD as an example. It pretty much implements MPEG-4 Advanced Simple Profile. But that's just the video part. Now there are people storing it together with AC3 or MP3 sound in an AVI container, there's the OGM crowd that can additionally use Ogg Vorbis audio and put this in an OGM container where we still don't know if it will be thrown overboard by the xiph.org people, then there's Matroska where the specs are firm but the implementations aren't complete yet. And to make this even better, there's a couple of subtitle formats and no definition how you should put a container on a disc. There's the standard ISO format, then there's XCD, but you could also use UDF. Bottom line, what ends up on your CDs or DVDs is pretty chaotic. The industry has failed to see the importance of creating such a standard and DXNs certification program seems to be the first real effort in that area. Of course it would be preferable to have a standard backed by the MPEG4IF but if that doesn't exist what can you do?

BTW, there are some reasons why MP4 and AAC haven't taken off. One is licensing, another is the lack of tools and yet another is that MP4 isn't a format designed for editing. But history has shown us that people creating their own videos need a lot of editing facilities so they won't use a format that's basically a dead end and needs to be demuxed first when you want to perform any editing. I also believe that in the end the AAC only clause has to be dropped because at least when it comes to standalones it makes no sense. Any DVD playing standalone must be capable of handling AC3, and most devices also do MP3, plus there is a lot of existing content that does not use AAC.

[Darrius "Junto" Thompson]

Originally posted by Doom9
according to Junto I thought GMC and QPel was already being tested so the extra effort would be trivial.

Yes the testing could be done if the features could be implemented. Still these 2 features would not be trivial to implement and there are a few issues to try and solve. There seems to be 3 versions of Qpel floating around. Which one do you implement? The first MP4 DSP's may have based their code around the first versions of the Momusys reference code, then there are those who used their interpretation of the MP4 spec for Qpel until then again there was a change were Qpel was modified in the Spec. So there are multiple Qpel solutions that exist and that will still have to be solved.

Darrius

[shlezman]

@junto

I thought you certify Divx codec, so why do you care so much about what xvid or any other codec implements, that's the job of MP4IF, isnt it ?

[Jordan "L0g05" Greenhall]

M4IF doesn't do "certification" they do interoperability testing. They also don't set MPEG-4 profiles- the MPEG committee does that. M4IF is simply an association of companies in (or interested in) the MPEG-4 arena. They could certainly choose to create a Certification process - but, so far, they haven't.

Doom9's issues about DVD players with "XviD" all over them - that don't actually play XviD - is precisely the problem. Our DivX Certification program is a reaction to the market - not an attempt to create anything.

The sad fact is that there really aren't any standards of quality or performance around MPEG-4 right now. A DVD player can claim "MPEG-4" if they can support QCIF Simple Profile at 50 Kbps. Worse - because of the way that the market works, there will be many DVD players that have XviD on them - very few of which will even be able to play 15% of XviD content.

Because we control the DivX name - and have some stake in ensuring that people get a good experience when they buy something that has the DivX name on it - we wanted to make darn sure that DVD players that say "DivX" on them are at least able to perform up to some guaranteed level of quality and interoperability.

The DivX Certification requirements include a lot of elements that aren't even talked about at all within MPEG-4. We have bitrate requirements, visual quality requirements, A/V synch requirements, etc., The goal is simply to Certify that the players are good.

THis is actually why and how XviD fits into DivX Certification as well. Since a DivX Certified player will be high performance - it will defacto handle XviD at least reasonably well. Since there are differences between DivX and XviD - and since we aren't physically testing against XviD - there are no guarantees here.

W/r/t OSD Certification. This can certainly be valuable. But it won't be easy. Try convincing DVD and other CE makers to change their devices to suit *your* demands. It aint easy.

[Jordan "L0g05" Greenhall]

Hey Doom - here is some more background on DivX Certification and the issue of "Super Certification".

The existing DVD players (e.g., the Elta) are completely incapable of supporting Qpel at a hardware level. This player is made on a $18 chip. If one wanted to hold-out for full ASP then the economics call for a $600 (or more) player today (like the first generation MP3 jukeboxes) and getting into the $300 price range with the next generation of chips. I can't see any reason why DivX Certified players couldn't get into the <$100 range in 2004.

For XviD support, the issue is really complicated. The problem is consistency and forward-compatibility. The CE guys won't include something in their device unless they know the device will play content for years to come. That means that we had to lock-in our tools and features to a certain set and have to keep those constant for the lifetime of DivX 5. Multiple B-frames is a good example. We didn't have them when we handed the "gold" DivX decoder to the chip guys. Adding them now would require a substantial re-engineering of the physical hardware. There is no way that the chip guys will do that.

Because XviD is OSS, there is no way to create any consistency in what XviD content will look like. As a consequence, you will get two things:

1. Companies that claim XviD but really can't play much XviD content

2. Companies that have premium brands and want to deliver a good consumer experience - who won't claim XviD compatibility.

Our decision to not include Qpel and GMC was based on four factors:

1. Requiring Qpel and GMC would increase the cost of the chips (according to the chip guys)
2. Requiring Qpel and GMC would push-back availability until after 2003.
3. There is little DivX content encoded using these features (exact numbers are very hard to get - but we estimate less than 5% of new content encoded and less than 1% of all content).
4. The features don't add much value.

Admittedly, this was a very hard choice. We included Qpel and GMC in the codec in the first place because we thought they were worth having. But compared to the silicion trade-off . . .

Note. You have to distinguish between the DivX 1-point GMC and a full GMC implementation. Full GMC is a bear on the chipset side and I don't know of any chip currently or planned for the next 18 months that will be able to support it.

Some more info:

1. Typical Chip design timeframe - 12 - 18 months
2. Typical time to integrate a new chipset design into a hardware design - 6 months
3. When does a hardware design need to be made before it will be adopted for OEM product-line - July.

So, if you don't have a chip available and ready by January of a given year, you won't get it into OEM products that year. There are very specific (and very tight) windows in this cycle and if you miss a window . . .

So the choice was "DivX Certifed" (without GMC and QPel) this year at <$200 or "DivX Certified" (with GMC and Qpel) *next* year at >$200.

J