KT-Tech Sound Compression - Music at 32 Kbit/s

Robert Buccigrossi writes: "KT-Tech, whose wireless video compression was featured in a previous Slashdot story, has released a demo for real-time sound compression at http://www.kttech.com/. Like their video, the sound compression is symmetric and is suitable for wireless real-time communication in software. It sounds better than Windows Media and MP3 at 32 Kbit/s for music and 4 Kbit/s for voice." According to the site, "licensing KT-Tech's sound codec is easy," but I bet it's not as easy as .ogg.

Comments

Limited Use

[commonchaos]

This apears to be a pretty targeted solution "suitable for wireless real-time communication in software" so comparing it to wma/mp3/ogg doesn't really apply. As far as if its better or not, it doens't really matter, mp3 is still the de-facto standard for end-user music encoding, simply because everybody uses it. And a licenced codec will never take over the "market"

A one string violin

[banuaba]

"According to the site, "licensing KT-Tech's sound codec is easy," but I bet it's not as easy as .ogg. "
You know, I like free software as much as the next guy, but I understand and respect the fact that companies have to make money. I fail to see why it was necessary to throw in a dig at this company that is doing neat things just because they want to profit from their invention. Just because its not free doesn't make it bad.

Now go ahead and mod me down.

The next golden egg

[RembrandtX]

I'm not sure if this is gonna be the next golden egg. Mp3 by hook or by crook, is now well entrenched into the market.

I do notice some differences at the lower levels .. KT sounds better than mp3 at 32k .. but who has mp3's under 128k ?

As much as it hurts to say this, having multiple compeating forms is gonna be hard in the digital music world. How many non-geeks have a diamond rio.

If you bought into the mp3 craze for $286.00 (a few years ago) and spent a month making yourself computer literate enough to use the rio for your morning workouts at the gym. What are the odds that you are going to be willing to shell out more $$ .. for a product you have to relearn, just cause it sounds a little better.

I think the mass market [the same folks buying into the m-life hype] is going to be a little less inclined to jump on a band wagon .. especially the mpe-sceners .. who were weened on free music. Anyone who was buying music in the 80's remember how long you waited before buying your first CD? (at $45.00 for 'The Wall' i didn't buy a second for quite some time)

I think this falls under the 10X rule again.
[the 10X rule being that something either has to be 10X cheaper .. or 10X more efficient to make a world-changing difference in an established market.]

Re:Bandwidth isn't the problem

[cryptochrome]

Nonsense. Bandwidth will always be a problem. No matter how much bandwidth you add, no matter how big you make your highways, no matter how much oil you drill, people will always use as much as you make, even if it means wasting it or creating enough traffic to degrade the whole thing. There is no substitute for efficiency. A better license can compensate for inferior technology to only a minor degree.

Re:The Beam in Thy Eyes

[banuaba]

I'm not disagreeing with that statement. But there is no *point* to adding that little tidbit on to the end of the article. All it is is a dig at a company that has done something cool. It's offtopic and petty, IMHO.

That's nothing

[epepke]

I can compress any Britney Spears song down to zero bits without loss of quality.

Mass market? I don't think so...

[martyb]

I can see applications for this beyond just the mass market. My first thought would be for carrying additioanl voice circuits over a T1 line. (Say, for carryting voice traffic between two locations in a large company.)

A T1 line suports 24 circuits, each of which has IIRC 64Kbps (ignoring RBS, etc.) Whatever. Each of these circuits can support one conversation. Using this technology, several more conversations could be carried on one circuit. (Their web site states 8Kbits for high-quality voice; 4Kbits for intelligible voice.) Even using the 8Kbit rate, that means 8 conversations could be carried on one voice circuit.

The result? A single T1 could carry 192 conversations instead of just 24. Or, put another way, get 8 T1's of voice capacity for the price of just one T1. At anywhere from $600-$1000 per T1, that adds up really fast.

Now, how long would it be until the phone company decides to replace POTS circuits with one of these? Dial-up users would find their modems capped at 8Kbits? Blech!

They cheated

[Magila]

They added noise to all the other encodings. Don't believe me? I re-encoded their 8 kbps kts stream to 8.5 kbps rm and even after the recompression it sounds better, listen.

We've been doing compression for decades.

[billstewart]

Telcos have been doing various kinds of voice compression for decades, especially on international circuits; an important feature has been detecting modem and fax tones and turning off compression. We used to use analog techniques, then digital, then fancier digital. There's not much compression used within the United States, but lots of international calling on traditional telcos runs at 32kbps. Many of the new low-price international carriers use Voice-Over-IP technology - you may be getting 8kbps. And in addition to the telephones-on-both-ends carriers, there are the international Internet-to-telephone gateway companies like Net2Phone which go for the consumer market.

But the place you really see voice compression on T1s is between corporate PBXs - if you've got enough traffic between your offices to keep 12 or 24 channels full, it might make sense to run a private line, and until the mid-80s lots of companies did this, but by the time everybody's PBX was smart enough to be good at it, the price of Voice-by-the-minute from long distance telcos was cheap enough that almost everybody ripped that stuff out except for multiple offices in the same city. But compression equipment has become cheap enough and good enough that lots of people are rebuilding those networks that we ripped out in the 80s, especially since IP data networks mean that even if VOIP isn't cost-effective by itself, you can piggyback some voice on a data network for not much extra operating cost, and the equipment cost may pay off pretty quickly.

Companies are more likely to use voice compression on international circuits, because the price of pipes across the ocean is usually atrociously high, but the price per minute for phone calls to much of Asia is also atrociously high, so a dedicated line using compressed voice is still often a good deal. It doesn't usually sound as good as a Real Telephone Call, but lots of Asian telcos don't have the best sound quality either. The other big trend that's appearing in international calls is VOIP over internet connections - the quality is more variable, but the price of a T1 or E1 internet connection in Asia is often similar to the price of a 64kbps or 128kbps frame relay PVC.

Not enough info to determine usability

[billstewart]

The web page doesn't have enough information to tell if the technology is usable. It's got some examples of compressed files, and a demo program, which does say that it's not vaporware and that at least for some kinds of sound samples it provides good compression and probably-pretty-good sound quality for the bitrates it uses. But that doesn't tell me enough to know if it's usable for any real applications. It needs several things:

• Documentation on what kinds of sound compression it can support at what bit rates, and what frequencies it supports. Is this the same algorithm for music as for 3kHz voice? Can you do really good voice with 11kHz stereo inputs, or do its voice compression modes only do a better or worse job of reproducing the same raw 64kbps / 3kHz voice stream?
• CPU horsepower requirements for compression. Is this a job for a DSP, or a fast PC, or a slow PC, or a Palm Pilot, or a wristwatch?
• CPU horsepower requirements for decompression. It's often lower than compression, but not always.
• Latency requirements for Compression and Decompression - is this usable for real-time conversations, or only for canned speech? This is separate from the horsepower requirements, which can be fixed by faster processors - many of the common algorithms require N voice samples to run the compression algorithm over, which is ok if you need 10ms of sound, marginally usable if you need 100ms, and unusable for conversations if you need 1 second of sound samples to get the compression rates (even though it's fine for music playback and other one-way sound applications.)
• If possible, standard voice compression quality scores compared to the popular compression algorithms.
• Information on what kind of licensing is negotiable and what isn't - can I give away free players and only pay/charge for compression tools, or do I have to charge every listener money for the decompression client? This makes a huge difference for web applications - it's much easier to get a web page publisher to pay for a tool with better compression than to get their readers to pay - that's Kiss Of Death mode.