I realize this topic is a rehash of rehashes of rehashes so I do apologize. I've been brought in as a consultant for a major record label's project, the details of which I am unable to provide; please don't hypothesize the details (thanks!).

The music industry has acquiesced to 128kbit, especially for the Windows Media Format. I am neither a fan of the bitrate or the format. I've come into the position to advise what would be an appropriate bitrate for this and future projects.

The basic requirements are a balance between transparency and file size for the average consumer and those with "somewhat" discriminating hearing. It is acknowledged that true audiophiles will continue to utilize "lossless" compression algorithms for portable media as well as formats such as SACD and DVD-Audio. Due to storage expense, average bitrates suchas 256kbps are not even on the table. As insinuated, I am leaning towards VBR (if available). Formats can be any, although there is an (unfortunate) bias to WMA. I might be able to push AAC, but MPC will be more difficult--MP3 in a pinch.

I am leaning towards 160kpbs (CBR/VBR average)--at the moment I am undecided on which format. To jump to 192kpbs (CBR/VBR average) would require significant benefit. To stress a point, file-size is more important to the decision markers than transparency, so I find myself also in the position of consumer advocate.

Insights that are clearly justified are welcome. Help me make the argument by clearly stating your position. I am not a decision-marker in any sense, so my proposals must be compelling. This is your chance to have a voice. Thanks everyone!

I see no reason not to tell them we'd like to see them use --alt-preset standard, rather than some arbitrary bitrate.

It seems to be like Ogg Vorbis is the format you're looking for. One of the best formats at lower bitrates and free of patents (and open source) that could scare companies.
My suggestion is Ogg Vorbis with some VBR (-q) setting.
Up to 255 channels and a decent tagging system.

Ogg Vorbis of course. It's free for them to use, and if you deliver the music Q2 (96 kbit average) 98% of people will perceive it as CD-quality. If you deliver at Q5 (160 kbit average), 99,5% of people will perceive it as CD-quality.

Biggest problem I see with ogg is portability. WMA, MP3, AAC/MP4 have been accepted for portable device use and implemented. More comanies are supporting aac/mp4 even as I write this.

What is the primary goal? Are we going to have pre-encoded files on a purchased cd or are we going to pay per file to download? If they are going to be pre-encoded on the cd then I am not sure Ogg would be the best candidate (portable device implementation wise). If we are downloading the files then I think Ogg would be a better candidate in this case.

But, considering all that I have read and the way companies are headed, I think that some form of aac/mp4 would be the way to go overall. There is more support in the portable device department. Most if not all the major "players" are trying to get some form of it going with their products. It is an all around decent codec at the 128kbps size. Decent maybe not transparent. It seems that aac would be an ideal candidate for the reasonably small size and decent sound.

my 2 cents

I would say 160 kbit/sec is right.

This is roughly what the Nero MPEG4 AAC codec produces at the standard variable bitrate setting, and what the highly appraised Musepack (MPC) encoder produces at the standard VBR setting, which even the most advanced listeners have trouble finding problem samples for.
These are the two formats I personally would opt for, with AAC being more accessible for the industry due to support from the MPEG consortium.

MP3 is nice when using VBR (every hard/software supports it), but needs bitrates in the 200k area to reach MPC/AAC's 160k quality imho.

WMA is not popular on this forum, and has a negative reputation in wide ranges of the userbase (my observation). I would only touch that if you offered the content for free..
WMA is no substitue for CD quality, if that is your aim. AAC/MPC/MP3 at the default variable bitrate settings are. (for me.. I'm no audiophile, but very good quality is important for me to enjoy music)

And if that label at the end ultimately wants WMA (because of a deal with MS about which you might not know yet), try to push them to WMA9Pro. It generates higher quality at same bitrate as WMA9Std. See guruboolez test in the Other Codecs forum. Regarding portables, media player does transcode the files anyhow before transferring, so whether Pro or not is a non-issue.

I would advise on 2 options - WMA '98' quality (approx VBR 335) or WMA 128. This allows the user the option of quality when paying for the file or of portability. They can also take into consideration wether they are broadband or dial-up.

AAC is another option of course, with the same theory. Offer 2 alternatives. The paying user can then decide which they prefer.

No label is going to support OGG or MPC, so your comments are likely wasted in that direction.

I would LOVE to see MP3 APS encoded with LAME become the default for all paying music sites but with no DRM, that is unlikely to happen.

Other than its bloated file size, WMA VBR 98 is really not that bad. It sounds good on my hi-quality heaphones. Its still not as good as MP3 APS and at a 150% of the file size. WMA 128 just sucks. I will never pay for a download in this format/bit rate. Most of this forum would not pay for any WMA, but I bet a lot of 'average' listeners could not tell the difference b/t WMA VBR 98 and the CD. Neither could they tell it from MP3 APS.

Unfortunately your original requirements, less than 192, means no codec is going to cut it as a replacement for the CD sound. AAC 192 seems like the best alternative.

...and this conclussion is based on?

This is what happens when a certain giant has a pocketbook the size of Texas.


Personally, I'd except AAC. This is going to be very widespread shortly. When another (relatively) large company adopts this as a standard for one of the most recognized portable players, others are sure to follow. It's kind of the keeping up with the Jones' syndrome.

As a sidenote MPC, for me, would be even better. There are tagging standards already well established for this format. An observation, AAC seems to have the disadvantage of being at a stage where all companies involved are pushing for their own variation. E.G: Apple vs. Nero. Are they cross compatible?

No corporate support, no portable support, MPC has licensing issues.

Heck, I think both Codecs rock! I would love to see either gain widespread use, especially MPC!!! Its just a fact that they will not.

Hardware support shouldn't be an issue, if a "major record label" was to start using OGG Vorbis hardware support would follow. I also would say, go for Vorbis;
- No patent issues
- Very flexible format
- It's at 1.0, so it's not buggy
- Open source, so further development can be done by anyone willing
- User base & support is growing vastly
- Decent quality level, comparable to the popular mp3

MP3, AAC and OGG are "transparent" at around 192kbps (MPC between 160 - 192 kbps). I don't think this is too high a bitrate considering the following:

1. most people download one song at a time, so even if a file is 5-7 megs, it doesn't take that long;

2. if you're using a hard drive or CD/MP3 portable, bitrate is not an issue;

3. if informed and educated, most people would favor quality over quantity (transparency over file size concerns).

When I first checked out AAC, I searched the web for information, and on Usenet there where many who talked about how AAC would replace MP3 soon. This was messages dated in 1998! Since then nothing has happened at all.

The big problem is getting the average user to accept it. With VHS->DVD it was easy, as DVD held so much better quality and functionality. Getting an average user to accept "AAC is better than MP3" is much harder. One thing that can do it is multichannel audio, I believe. If a user can get surround sound AAC that sounds really good in his new home theater, it might just do it.

Apple didn't have it as a format for its iPod then. Now it does. Watch...

This is excatly what happened to MP3 encoders also. You had various FhG encoders, LAME, BLADE, XING, ect. And yes they are cross compatible.

Huh?

Those MP3 encoders are called MP3 encoders because they output MP3 files! Why shouldn't they be "cross-compatible", whatever that's supposed to mean? Quicktime and Nero have MPEG-4 AAC encoders outputting MPEG-4 AAC files! Both MP3 and MPEG-4 AAC are standards. I don't get what you're trying to tell.

As a consumer who has purchased well over 1,000 CDs: FLAC.

I don't care about your storage costs, which is a silly excuse given $1/GB cost at retail prices. I don't care about your bandwidth costs either, which will be reasonable on a per-song basis.

If I'm paying up to $1/song I demand true CD quality.

True, also I shouldn't say nothing has happened. It's just that thing has gone very slow. I hope it turns around now. Anyway, I shouldn't have strayed from the topic.

Regarding the size, I think broadband is spreading, and getting faster. That is happening here in Sweden anyway, where there is quite fierce competition. Storage is also increasing, bigger harddrives and bigger flash memories. I think a good format like AAC/MP4 with around 160KBit is a good choice.

As an "audiophile" I do not want to use lossless for an average download - not very portable and a waste of bits for typical / background listening. I use lossless for archiving and backup, but not for portability. I can always break out my original CD if I want "lossless." Forget SACD and DVD-A - benefits are for record companies, not the customer (me). Love the potential for the 24/192 quality etc., but the encryption, watermarking, etc. make SACD and DVD-A downgrades versus CD for me - and at a higher price? Doesn't pass the "stupid" test. The market numbers will back me up - these are loser formats. They would have been very interesting in 1992, but in 2002 they are irrelevant.

Regarding compressed distribution - why only one answer? This is typical of the music industry. One format makes sense when you are distributing physical media like CD to keep the shipping container sizes and display areas reasonable, but online it doesn't make sense at all. The music industry needs to learn how to segment their market to maximize customer satisfaction - and this will also maximize their revenue at the same time. This is Economics 101 - if you're consulting them then you should know how to explain it.

My recommendation is this, two different formats for two different audiences:
1) 160kbps MP3/OGG/WMA or whatever for "average customers"
2) 256kHz MP3/MPC/AAC/OGG or similar for "audiophiles"

I personally would love to see a third format - lossless for the content I really love, but being practical here I think two quality levels is reasonable. If you are "trying it out" or don't really like a track very much, then get the average quality. If you really like the track or are very discerning about sound quality get the audiophile level. Of course the audiophile level should come with cover art and lyrics too and other extras and the price would be reasonably higher, like 10%. Yes its a higher quality encode, but I am spending my bandwidth to download and store it, so make it a reasonable premium and throw in a little more content to make it worth everyone's while.

For goodness sake you must convince them to at least use VBR on all encodes. Remember that VBR=constant quality and CBR=variable quality. Muic is dynamic, the bitrate should be too.

Why would you be paying for tracks that you already own?

And are you saying you would pay for 256kbps for other songs with no option to get the lossless version?

I will buy the ORIGINAL CD of a track I really care about.

I would pay a dollar (although I would like to pay less ) for a decent quality encode of a track from an album I am either not interested in buying or "sampling".

Lossless tracks for sale...it is just not gonna happen. Not for a while, at least.
If downloaded music is not for you...well, it's not for you, then.

The best response yet on this thread. This is what I was trying to say earlier, but DigitalMan better articulated. I would only add WMA VBR 98 to option #2. Like it or not, WMA is the preferred distribution format for the record labels. Besides its bloated size, WMA VBR 98 is as good as any 256 format to all but the 1% of the most highly trained ears.

DigitalMan even had the good sense to suggest OGG. In this forum, subscribing to any format other than OGG is a recipe to get flamed!

I very much like this direction. Two streams, with ~160kbpz AAC and high-quality WMA (please try to push for WMA9Pro, it really does make a difference, especially when you're aiming to please the die-hard listeners) would apply to nearly everyone. the 160kbps stream would be for people not interested in large downloads, and the higher-quality stream could be provided to the people who felt it was important. AAC, now supported by Quicktime, would play on any up-to-date Mac with OS X, and WMA will play on any modern PC. Killing four birds with one stone Also, as a third option you could offer a 64kbps file for people on a dial-up connection. MP3Pro and OGG have proven to be the best codecs at that bitrate range.

And why not AAC 160 kbps and stay with AAC for higher bitrate ? AAC CBR allow 512 kbps encodings.
Why WMA9PRO ? This codec IS NOT compatible with old decoder (try with winamp for exemple) and current portable player.
I can't understand this choice (or may I say dream ?)

I think he is trying to say that high bitrates are not an option because of the expensive space it will take to host them, not the time it will take to download them. Thus, 2 streams would seem unreasonable given these criteria. As for the quality issue, I would have to lean towards AAC at around 160-192kbps, encoded with the best quality encoder available (at this point it's the one included with QuickTime 6.3, I forget exactly which one it is, Dolby or FhG Professional??) using ABR or VBR. AAC seems to be the future of portables, and already has achieved very good quality with the latest encoders. Ogg at -q5 would work perfectly as well, especially sense it is free to use , but it lacks support for portables (if this is an issue.) MPC is the best quality of the bunch, and -q5 averages ~180kbps~ for all files. I have yet to hear an MPC file at -q5 that can be distinguished from the original without listening real hard. MP3 and WMA should be used as last resorts.

In short:

• AAC for compatibility with future portable devices and very good quality ~160kbps.
• OGG for free licensing and great quality sound at around 160kbps, although no portables are supported (yet..)
• MPC for the highest quality possible using today’s lossy codecs.
• MP3 and WMA should be last resorts.

OK...

1. Explain VBR to them - matching the quality to the music.
2. Sell them AAC at around 128-192kbps VBR
3. Tell them to make sure that what's going into the encoder is of consistent high quality, mastered to sound good (not just loud).

From what you've said, that seems to be the kind of level they're looking for.


FWIW as a paying customer, I'd be happy to buy content if the price and quality is right. Note that the sound quality isn't just down to the encoder. There are some terrible sounding CDs out there, and that's not because CD is a bad quality format.


If quality is an issue, mention this: if everthing you release via this service is very well mastered, then the impression of quality via the service will be quite high, even if you used 128kbps mp3. If the quality of the tracks is variable before encoding, then even 256kbps AAC isn't going to rescue it.

I'd go for ~160kbps AAC, with great care put into what goes into the encoder!

A higher quality version is of no consequence to 99% of the public. If anything, for downloads, you probably need a much lower quality version available for the modem users out there - say 96kbps. Depends on your demographic.


Cheers,
David.

I agree with 2BDecided here:
AAC has already shown in the 128kbps listening test, that it is quite a capable format (well, no one ever seriously doubted that) and that current implementations are high quality.
The winner of the test (Quicktime 6.3 AAC based on Dolby's implementation) currently has the downsize of not featuring a VBR mode, which according to Guru's tests done with Psytel and Nero, dramatically increases the quality while keeping filesize small.

Nero's AAC codec would be a good choice, since it has an excellent VBR implementation and also is going to include HE-AAC encoding soon, which could be used for providing a low bandwith version of the track.

Vorbis would be another valid option, especially if AAC's licensing is too expensive and/or access to the source is needed for implementation in the 'system'. It provides good quality using the VBR mode at quality 5 and really shines at lower bitrates too, although listening tests have proven that Nero's AAC codec beats it at both bitrates.

Personally I'd advertise Vorbis, since I still believe in the future of this codec and Xiph's plans to provide a complete collection of open media technologies. Support from a record lable could be exactly what it needed to gain new people for the further development (which is semi-dead as of now, since Theora needs all resources Xiph has available).

dev0

I didn't mean to imply I would be paying for tracks I already own, rather that for new content ("average download") I think lossless is not necessary. Given the incredibly poor state of mastering in the industry ATM I'm not sure lossless is that relevant for most tracks.

I would pay for 256kbps for other songs with no option for lossless. Only about 10-20% of my content I would say falls into the "can't live without it" catagory which I would want lossless for. Everything else would be very well served by a well done 256kbps version for me.

For my top 10-20% favorite content I would even be willing to pay a little more for lossless, although of course not for content I already own. I suppose I anticipate that some day there will be new content I actually like that is mastered reasonably well...

On a different subject, the speculation that it would be "too expensive" to host two versions of the file is interesting. Storage is relatively cheap and you only need to store it in one (or a few mirrored) locations. Heck, the old catalog doesn't even need to be maintained/updated - encode it once and serve it for 30 years. If customers are willing to pay extra for the quality, then it very likely might not be too expensive and might be cheap and profitable to boot. For that matter why not store it losslessly on the server and render it in whatever format the customer wants/is willing to pay for? That way you don't need to care about reencoding the library when/if customers demand a different format and you can easily vary the quality levels sold.

The bottom line is that nobody has ever tried offering multiple quality / content levels of music to customers - why not try it and see what the customers think? Assuming that it is too expensive is assuming customers won't pay for it, which may or may not be true. This problem is best solved looking at it from the perspective of the customers who have different levels of quality needs (the 128kbps MP3 crowd and the SACD/DVD-A 24bit/192khz crowd) vs. looking at it from the perspective of the record companies trying to minimize cost. History shows that serving the customer needs will win in the long term every time.

I'd give another vote for Ogg Vorbis too. The main (maybe the only) reason why I, and also a lot of people here, haven't bought an HDD based player yet is that Ogg has still not been implemented.

Sure, AAC looks very promising too.

Maybe someday, some company will dare to care a little...

Just see that poll.

I've read every single one of the replies and I am making note of them. I can't respond in specifics nor say what is decided, but there are excellent and thought provoking points here.

I am able to say that the reason why file size is such an important issue is because 10TB would only hold a fraction of our catalog at 128kbps (and that goes for the other 4 major distributors as well). Simply increasing the bit rate 4KBps drastically reduces storage capacity (of otherwise adequate storage) by the thousands of titles. Further, if you examine the iTunes service for example, Apple is also storing data like album artwork. It adds up quickly to a significant amount of data. Unfortunately, ideologies quickly give way to realities. Like many of you, as someone who can hear artifacts in blind listening tests, it definitely bothers me too.

My thoughts:

Free previews, could be at 64kbps (average), in AAC (Mpeg), Ogg Vorbis (Xiph), Mp3 (Thompson). It is important that they know what "lossy" as in lossy compresion means. It is NOT cd quality, it is more akin to a tape copy. (With mp2 you could even claim it is )

Pay per content should be Lossless if quality is a must. MPC standard at the very least, this format only needs a little research with possible patent issues, but you can't get a better quality/filesize ratio with others. It also seems very light(fast) for encoding and decoding, i'm sure a hardware implementation won't need to be as complex as with the other formats, consider it has more things in common with mp2. Second alternative would be with AAC, now with Apple backing, it seems the perfect candidate (specially if you need to get out of Microsoft chains), but current encoders are in evolving state, which can be good or bad, not sure of their point of view. Something finished NOW, would of course be MPC. Ogg Vorbis is excellent in the patent arena, its encoder is also in an evolving state. A backing from some important group could be just what Xiph needs. In the last we have the widely accepted Mp3 format, you only need to deal with Thompson, and then use lame APS (or lame APS -Y) and be just fine with it and enjoy current hardware support and a highly tuned encoder.

No matter which format is choosen, means for playing will come. People say things like "no hardware support", but content must come FIRST for hardware manufacturers to notice what their posible client demand will be. In 1996, there were no hardware mp3 players, yet people were encoding and decoding mp3 files. This could also means an added business oportunity, if they choose a format currently without hardware support, they could push a player for it Choosing AAC/MP3/etc means paying royalties to someone else, those could be invested in a new player using a royalty free alternative.

For online models, 1$ per song sounds to me unreasonable. Its the same cost of a song per CD, but without CD quality, without printed art, without physical media, no shipping costs, etc... Something in the order of CENTS (say 25c) per song sounds better. I don't know if a 30minute song should apply too, or what to do with classical music? Also think that a low price is the best method to discourage the so called "piracy".

But please, unless there is already a very good deal (and even if there is), avoid Microsoft like the plague. You only need to study their business model history. First, its all free and good looking, then later they restrict things and impose their own rules, then you either belong to them or get out of business. They really believe in the drug dealer method. Just study their past, they are not worth considering. And BTW, their format really sucks in terms of quality, you have already seen posts of people claiming 355kbps file being similar in quality to 160kbps of other formats, that should give a clue... Also support in other platforms is just a grace we can enjoy today, who knows if later they decide against Apple or any non Windows (or even older windows!) platform?. Didn't you see what they just did with Internet Explorer?. No more standalone product, it is now part of the Windows OS, period. Goodbye Apple Yes, they love to do that.

And last: "Copy protected" (corrupted) CDs with 48kbps wma audio files are not the way to earn respect...