I was looking through the comments on cvs for the libmp3lame section, and I noticed the comments on the 'preparation' for nspsytune2. Any major changes in this version? Will the alt-presets have to be retested etc.?
If nspsytune2 will ever happen then sure, all the presets using nspsytune must be checked.
However, Naoki being a very busy man atm makes me believe that this will not happen in the near future.
I'm not really certain of the details of nspsytune2, but I imagine it probably incorporates much of the things Naoki has talked about recently on IRC and other places. A lot of it has to do with better tonality estimation.
One thing I think which is important to point out, especially if there is interest in keeping the alt-presets in synch, is that future developments in the LAME psymodels must be verified on a wide scale by the community. This is absolutely crucial for there to be real progress. If this doesn't happen, I could easily see the effort to move forward fragmented once again due to a lack of collaboration and unity, and especially verification of enhancements. I can't really speak for Naoki, I don't know his intentions with this work, but I'm hoping he agrees...
Now, nspsytune2 should work, maybe...
www.mitiok.cjb.net (http://www.mitiok.cjb.net)
so I guess it's being developed in LAME alpha now???
(edit)
BTW on my system this version when trying to use nspsytune2 just dumps with "abnormal program termination"
Originally posted by maciey
BTW on my system this version when trying to use nspsytune2 just dumps with "abnormal program termination"
You need another program to create a preprocessed out.psy first:
[15:27] <Shibatch> pass 1 produces a file out.psy from wav file,
[15:28] <Shibatch> pass 1 binary is a new program which does some preprocessing of psycho acoustics
[15:30] <Shibatch> making nspsytune2 to be full 1 pass processing requires quite large modification of lame
[15:31] <Shibatch> So, after people conclude that nspsytune2 is better than nspsytune, I'll do that work
As i'm not sure yet if Naoki wants to have public testing in this early stage of nspsytune2, i can't provide that extra program right now. As soon as i'll get his permission, me or others can put the file and some rudimental instructions online.
Mitiok's Feb 16, 2002 alpha doesn't produces the same output as Naoki's special lame nspsytune2 compile from the same date. Just for your information.
ff123
so I guess that means
"Get your hands off nspsytune2 and leave it to THE developers to finish"
sounds OK to me - I just hope Dibrom puts his hands on it and sees if it can improve anything
Originally posted by ff123
Mitiok's Feb 16, 2002 alpha doesn't produces the same output as Naoki's special lame nspsytune2 compile from the same date. Just for your information.
Yeah, something is wrong with Mitiok's compile still.
I have briefly tested (with Naoki's compile) nspsytune2 combined with both --alt-preset standard and --r3mix, so far haven't heard improvement, infact quite the opposite and the size is bigger. Of course can be that it needs totally new tunings, so I really can't say yet much.
[edit]Just tested Sophia2.wav it's better with APS&NSpsytune2 than just with APS, bitrate is somewhat higher though.
Hi,
Please wait a moment before the official release of nspsytune2.
Current version of nspsytune2 seems that frequency resolution is not enough.
Here is the latest version.
One major bug has been found and removed. This is alpha version and not the official release.
I believe pass1 doesn't include patented code.
http://www.milky.ne.jp/~galaxy/20020221.zip (http://www.milky.ne.jp/~galaxy/20020221.zip)
Suggested options are :
nspass1 --base -2 --tmn 18,12,12,5,0,0,0,0 in.wav
lame -q0 -V --lowpass 16 --nspsytune2 out.psy in.wav out.mp3