Talk:Speex

Untitled comment

Although not a formal part of the Ogg multimedia project... this doesn't seem to be true any more; Xiph lists Speex in their page about the Ogg project. TMC1221 04:55, Jan 29, 2004 (UTC)

Like FLAC, Speex can be put into an Ogg bit stream, but is usually NOT. FLAC, Speex, and Ogg are all Xiph.org projects, but FLAC and Speex are not sub-projects of the Ogg project like Vorbis is.

(I am the author of Speex) No, Speex is almost *always* in an Ogg stream (when it's not used in VoIP). Speex is a Xiph.org project in the same way as Vorbis now and Vorbis is *not* a sub-project of the Ogg. Last thing, the content of this page is mainly taken from Speex manual and should be marked as such (i.e. give credit where credit is due).

Thanks for pointing that out. I've marked the page as a copyright violation.

It's not a copyright violation, but it is a licence violation. I've added the credit and copyright information, in a similar form to that which we ask for (Wikipedia:Copyrights#Users.27_rights_and_obligations). That look OK to you 130.. ? -- sannse (talk) 10:29, 1 Nov 2004 (UTC)

FYI, a license violation is a copyright violation. "You're copying my work, do you have a license?" "Yes, I have the GNU FDL." "You aren't following the terms of the GNU FDL. Do you have another license?" "No." -> copyright violation. Graue 00:39, 20 May 2005 (UTC)

IP misinformation

The current version of the article states: "Designing for VoIP instead of cell phone use means that Speex must be robust to lost packets, but not to corrupted ones since IP ensures that packets either arrive unaltered or don't arrive." This isn't true, as the IP checksum only covers the IP header, and data part of an IP packet *CAN* arrive with errors in it. I suspect this was meant to be UDP rather than IP, but I'm not 100% sure. I'm changing it for now, but could someone please confirm that Speex uses UDP? Thanks. - James Foster 05:25, 15 September 2005 (UTC)

Speex doesn't know or care about UDP, but most VoIP apps employ UDP, so it's a fair change to make.

Quality Scale

The quality scale says it goes from 0 to 10; but which is HIGHER quality. For instance, in some encoders, "0" is the highest quality, in others, "10" is the highest. What is it for Speex?

According to, 0 is lowest (2.15 kbps) and 10 is highest (24.6 kbps). –Mysid 08:04, 21 November 2005 (UTC)

Complete article

• I want to expand this article to describe the algorithm used in Speex (in more details than generic CELP). Is that the right place do to it? What are the guidelines as to how much details is appropriate? Jmvalin 13:25, 20 March 2006 (UTC)

• • Keep in mind the official technical and stylistic recommendations for article size. If you're going to use advanced concepts to describe the algorithm, be sure to provide a reference (either a Wikipedia article or an external link) so that a novice can garner a basic understanding. You might also think about following the summary style method and write a basic summary of the Speex algorithm then link to another article (e.g., Speex algorithm) where you go into more depth. Again, be sure to utilize links to other Wikipedia articles. I'm somewhat of a novice, so I can provide feedback. Thanks for Speex, by the way. —ShaneCavanaugh 22:05, 20 March 2006 (UTC)

• I think the article in its current version is a mess. I see lots of people editing minor bits here and there, but what's needed a a near-complete rewrite. A while ago, someone basically copied random parts of the Speex manual, which is why it doesn't really look like an encyclopaedic entry. Jmvalin (talk) 23:54, 10 March 2008 (UTC)

"claims to be unencumbered"

Does anyone else claim that it is encumbered by patents? If not, scratch the "claims to be". — Omegatron 00:08, 29 March 2006 (UTC)

Actually, I think the "claims to be" is appropriate because (unfortunately) you can't ever prove anything with patents (unless you go to court and even then it takes years). Sad states, but "claims" is the best you can use with patents, regardless of whether it's open-source or proprietary. Even "patent" codecs merely *claim* to be covered by patent XYZ Jmvalin 11:54, 31 March 2006 (UTC)

Going to court doesn't prove something is unencumbered by any patents. All it can prove is that something is unencumbered by some specific patent. There could be another patent lurking somewhere that the court case didn't address.Phr (talk) 06:23, 27 May 2006 (UTC)

There is no real way AFAIK to 'prove' something isn't encumbered by patents. Claims to be is definitely the most accurate terminology since there has been no real search or anything AFAIK. Even Dirac doesn't claim it's definitely not encumbered by non-BBC patents, just that BBC has has extensive searches done Nil Einne 15:22, 11 January 2007 (UTC)

Speex was specifically designed to avoid known patents in this field. For example, it avoids ACELP which is more computationally efficient. --Gmaxwell 04:28, 22 March 2007 (UTC)

callReplay

I just realised an application called CallReplay uses speex

Comment on quality

I found this interesting comment from the author of speex: [1]. He discusses quality and comparisons with Ogg Vorbis. That link might be useful as a reference. --Gronky (talk) 17:44, 20 April 2008 (UTC)

when was speex created?

that would be a good addition to the article. --Rajah (talk) 03:22, 23 August 2008 (UTC)