Hi folks, some notes about goings on. After opportunity to think aloud, I have the pieces of the VQ system more or less in place. I need to finish the codebook generation core (this is not libvorbis code, but an external utility). I then need to write the encode/ decode/packing code that makes use of the codebook. This future code is already specced. Tony: I haven't worked on the xmms code you sent. I'll actually start on that now before continuing on the VQ code as most of the VQ questions are settled (if not coded). Mike Whitson has jumped back into the command line player. The new command line player (I expect) will be built around vorbisfile.a. Our ISP has finally fixed the routing error that had xiph.org partitioned from BBNplanet the second half of the week. Hopefully everyone got all the mail that backed up in that period :-) No meaningful commits this week. Monty --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/
Hi folks, First a heads-up; I'm in the midst of getting all the VQ code integrated with libvorbis, as well as setting down the final header format. The code in CVS is currently half-there and as such doesn't build. It'll be fixed soon (that's what I'm doing right now). I'm finishing the VQ code integration (the codebook module passes its self tests; that encompasses packing, unpacking, and using codebooks for encoding and decoding. The code for generating codebooks was finished last week). All the VQ infrastructure is complete and will be integrated in the next day or so, then I just need to add the code that *uses* the VQ :-) That's partially in place now; the LSP coefficients will be the first VQ-ized output. All that's left after that is residue. Monty --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/
As of my latest commit, the libraries (libvorbis and vorbisfile) and example programs (encoder_example, decoder_example and chaining_example) all build and run cleanly. I have to check the xmms plugin next, but any modifications will be minor. There's work left to do for the first cut (floor0.c isn't *quite* done and res0.c is still incomplete), but we're close. As of today, the VQ engine is in place and functioning properly, although until I finish res0.c, it's only being used for the spectral envelopes. Fell free to grab the latest from CVS and go to town. Additional changes will affect only library internals, nothing externally visible to an application. Also, those of you with r/w CVS access: From here on out, mainline CVS must stay buildable and runnable. If you want to do any work that requires committing intermediate (nonfunctional) versions, please do so on a branch. Monty --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/
It's been too long since I've checked in a comprehensive update. Things are still short of first cut; my initial tuning turned out to be quite a ways off. However, correcting the tuning will take a little time and I've been promising bitstreams :-) The just-completed checkins feature a complete Vorbis engine that encodes real bitstreams; we're officially past the 'hardwired demo' stage. Because I don't have real encoder settings quite ready, the checked in code features a working but still conservative [roughly] 68-70kbps per channel mode. I know; the bitrate is so so and the sound quality still needs minor improvement. Yes, it wastes most of the VQ engine. Yes, we'll get substantially better than that soon. In other somewhat bad news, the XMMS plugin is broken and Peter and I are working on that now. encoder/decoder_example are the only current means of testing. The good news is that I have my first proof I got the format right; I need to substantially rework the quantization data (and how the encoder uses that data) for improved tuning, but the decoder will not have to change. We pass our first test of format flexibility! So, we're almost at the real thing (well, we have the Real Thing now, but I'll hold off a bit until the Real Thing is as impressive as it should be!) Greg... I'm dying to see that residual quantizer you were speaking of. Care to introduce it to me/us? Monty --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/
So, beta launch happens today. Of course things are behind schedule, don't be silly. That means mostly that not all the features we wanted to have ready at noon today are here quite yet. Specifically High bitrate encoding is still a few days off. "Aw shucks". The beta is at http://www.vorbis.com/ Also, the preecho work I did in the past two days dramatically improved preecho... at the cost of non-short-block quality. Oops. It's not a show-stopper (and it's not dramatic), but it does mean there's a little instability in the high end again. I'll be correcting that as soon as possible of course :-) So the Beta launch is mostly for the benefit of getting the industry's attention. Some of the really good quality improvement work I've done recently won't be on the site until end of week most likely. Everything else is solid. Have fun folks and I will be seeing you again on Thursday when I get back from the mp3 summit. Monty --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/