These binaries (http://www.din.or.jp/~glit/TheOddStage/Progs/Vorbis/autobuild_e.html) (05/31) claims: OggEnc v0.95 (libvorbis rc4)
Note from rc55: I have had it confirmed by Jack Moffitt that these are indeed NOT the rc4 final release binaries. As and when they are available we will announce them.
I have tested this oggenc binary and found that the new low bitrate options are coming along nicely. Here are a few commandlines and result bitrates.
"-q 0 --downmix" averages around ~40kbps
"-q 0 --resample 22050 --downmix" averages around ~32kbps.
"-q 0 --resample 32000 --downmix" averages around ~36kbps.
So, dialup rates at a reasonable quality are achievable, but not finalised.
RC 4????? wow....:eek:
could anyone (Garf, Monty) please confirm this ?
l also noticed this when I did a CVS sweep yesterday. I haven't checked out the CVS today, yet, but unless there is a change today, then the ogg/vorbis libs haven't really changed since April 8. Yesterday's changes to oggenc were about resampling and downmixing to mono from stereo.
I was really expecting a library change, or at least a change to the version date in the lib. I held up releasing a compile on this basis.
If I find anything before anyone else, I'll post about it.
I haven't seen any CVS activity lately that would indicate that RC4 is very close to release. Monty has been doing some work on a separate branch, but that has yet to be merged to the main branch. And most commits have had comments like "avoid loosing work". Doesn't seem like clost-to-release tuning.
thanks for the information
I think OggDropXPd should implement thoose new options...
same opinion here
OK guys! I'll try and squeeze it in betweeen World Cup games!
don't forget flac and ape input capabilities
again a new version. Does anyone know about the changes ?
Monty said that the noise normalization (a technique to increase especially lowbitrate quality) is now in CVS and that RC4 won't be out for a little while. He's doing initial tunings of new techniques now, and probably soon there will be a short listening test phase in order to further tune RC4.
that's interesting. Please let us know about these testings
Noise Normalization I see interesting? what C source-file is that located in I want to take a look at that. I want to get a sense of how it may work any ideas?
Dont know about source file, but here's a short information about the Vorbis' noise normalization technique:
With RC3 in some situations especially at low quality levels Vorbis is cancelling out noise and the miss of that is audible. This causes noise pumping and swishing. Noise normalization tries to prevent the total cacelling of noise in some situations. This has been a problem especially at lower quality levels.
<xiphmont> the reason Vorbis, right now [in RC3], at low bitrate gets muffled/sparkly is that bands aren't being quantized with any attention to maintaining audio energy. If the quantization level causes everything to quantize to zero, oops. Noise normalization attempts to maintain by-band energy in noisy sections of the spectrum.
Also in RC4 there should be new masking data curves which should increase quality for every quality levels.
Are these new masking curves going to be the same Ehmer curves that have been used all along for masking tones up to 8kHz from 40 - 100dB and then from 8 - 16 kHz will there be any changes from "imperial testing" or will there be a whole new set of masking curves all together?
Monty hasn't decided that yet, but he suspects it's all new curves.
One more quick question, will they be curves specifically design by him? or will there be ones that have been derived from other research documentation? I am just trying to get an idea and a sense of what will be in the forth coming.
Originally posted by HotshotGG
One more quick question, will they be curves specifically design by him? or will there be ones that have been derived from other research documentation? I am just trying to get an idea and a sense of what will be in the forth coming.
<xiphmont> I'm using all the research I have. Including published data. Problem is there is no published *raw* data. All the published data is already highly fudged or simplified.
By popular request , oggdropXPd v0.5beta , has just been uploaded to the usual places and should be available soon.
This version has the new 'downmix stereo to mono' and 'resampling' options implemented.
I did some test encodings with the June 15 binaries from that site. I mainly use electronica music to test, and I'm pretty impressed from the quality change from RC3 to this version, and it did also shave avg. 1.5kbit off the bitrate at -q 0 setting