http://www.uni-jena.de/~pfk/MPP/

- Downloads removed: New version available -

*replaygain: Some bugfixes (mostly file handling)
*replaygain: Report modes which suggest K setting report
*replaygain --listreport C:AUDIO D:AUDIO E:AUDIO
*Thumb profile decoders (mppdec, WinAMP, XMMS plugin):
-Level bug fixed (generated level was 2.5 dB too low for high frequency bands)
-Removed metallic sound from high frequency bands
Both were decoder issues!

Encoder
*New tool uses Perceptional Noise Substitution (PNS) for high frequency bands if this is less or not audible
*This tool is on for --thumb and --telephone.
*It can be set via --pns x, where x should be something between 0 and 0.2 for high quality and 0.5 and 2 for low quality settings.
*Bitstreams encoded with PNS can't be decodes by old decoders, newer decoder can decode it but sounds dull. Decoder from Thursday last week have a level error of 2.5 dB (too less) and a coloring problem. Decoder from today sounds okay.
*--pns 0.25 is useful for --radio profile if 110...130kbps is the target. This degrades sounds less than a reduced TMN oder NMT or a gained ATH.
*But this can't be decoded by Andree's Decoders, so I can't enable this as long as these versions are available. This is not to blame Andree, but to avoid confusion !!!!!
*All profiles are a little bit tuned for my ears.

Telephone:
-A little bit less in quality than the old thumb profile
Thumb:
-TMN increased from 4 dB to 10 dB
-Bandwidth increased from 9.5 kHz to 12.5 kHz
-CVS enabled
-PNS with threshold 1.0
-adaptive ATH enabled
-Tonality estimation for bass enhanced (formerly only in Xtreme+ profile)
Radio:
-ATH gained by 1 dB
-Tonality estimation for bass enhanced (formerly only in Xtreme+ profile)
Standard:
-TMN decreased by 1 dB °)
-NMT increased by 0.5 dB
-Tonality estimation for bass enhanced (formerly only in Xtreme+ profile)
Extreme:
-Tonality estimation for bass slighly modified (reduced for 0...20 Hz)
Insane:
-NMT increased by 1 dB
BrainDead:
-NMT increased by 1 dB

Note that a --tmn 20 actually generates a TNM of 23 dB for Andree's Encoder and a TMN of 26.5 dB for Klemm's Encoder for 3 kHz (23 dB and 29 dB for 120 Hz).
Note that PNS encoding is very early alpha, far from being usable as a normal tool. It was basicly a hack which works too good so I publish it. There is a rich source for improvements starting from false positive encoded (artefacts), false negative encoded (waste bitrate or quality on CBR), stereo imaging problems.

I wonder if the replaygain.exe in http://www.uni-jena.de/~pfk/mpp/bin/mppdec-windows-1.02.zip is actually the 0.93 version? It shows 0.81d, but the date is 05/13.

Also I wonder what this actually means:
*But this can't be decoded by Andree's Decoders, so I can't enable this as long as these versions are available. This is not to blame Andree, but to avoid confusion !!!!!

Can't enable? By default or at all?

Lol Frank.. You seem to be a real prankster sometimes..
Suddenly Perceptual Noise Substitution changes to Petri Net Simulation..
"New tool uses PNS (Petri Net Simulation) for high frequency bands if this is less or not audible "
Yeah right..

What about Planetary Nebula Spectrograph?

Seems to be, all the new features are working.


I think it's referring to this:

Maybe --radio profile would be better with the said pns value but won't be enabled by default before all old decoders are replaced. Anyway, my interpretation may be incorrect but atleast --pns option works as it should.

Hmm.. where is a list of the new features..? Is --smart in a usable state now? The display bug is still there (when you replaygain a second time, it shows "-5.34 dB -> -5.33 dB" without actually changing the gain). Oh, and it could be faster. Okay okay i'll stop now.. :angel2:

Xmms is dying on me when switching between tracks encoded with the 1.02 encoder. Anyone else seeing this? I can try and throw together a debug build and track it down.

I have the same problem with XMMS and new tracks.

--
GCP

If you run replaygain it shows all the switches, the new features I was talking about are the --listreport and --listallreport.

No.

Ah, those ones. Just tried --listallreport on a directory with many MPC albums:

Suggested mode for
title based replay gain: K-17
album based replay gain: K-17
title based clipping prev: K-18
album based clipping prev: K-18

I'm currently using K-14, and i'm already getting a heart attack when i first play a few MPC files and then a non-MP3gained MP3 or a WAV... the MP3/WAV is so goddamn LOUD... well, at least my neighbors know that i'm home. :diabolic:

Solution: Applying Replaygain to all my files, including all MP3s. Oggs, whatever. But i haven't sorted them into single folders yet... and as a music lover (some say audiophile, but that has recently gotten a slightly negative touch :lurk: ), i want to use Album Gain, of course. Ergo, i'd have to sort the MP3s. But that operation would take one or two manyears...

Shld wrk nw.

Frnk Klm, Lnx sr

Frank,

While your coding, could you cook me some dinner?

I'm very excited about the lbr stuff. Keep up the good work!

Ruairi

Looks like it's a segfault in bitstream.c, in Bitstream_read, at this line:

return out & mask[bits];

Looks like bits > the size of the mask array.

Any ideas?

To reduce level of MP3 nearly lossless, there's MP3GAIN.
To reduce level of WAV files losssless, you can currently LPAC.
Monkey's Audio fails.

A 16 bit WAV has the following bits:

[sxxxxxxx|xxxxxxxx]

Reducing the Level by 1:2 (that is 6,0205999... dB) you got:

[ssxxxxxxx|xxxxxxxx|x0000000]

Reducing level by 1:4:

[sssxxxxx|xxxxxxxx|xx000000]

A well designed lossless encoder should compress this to nearly
the same size as the original file (< +0.1% in size).

Monkey's Audio performs really nasty with such files. Often the files are much larger than the original 16 bit WAV files.

How about replaygaining the ones that aren't sorted now with radio gain, and sorting/albumgaining the individual albums when you have the time, or when you feel like listening to one of them as an album? At least that way you won't blow your speakers/ears/neighbours

WinAmp Plugin 0.93a is available:
http://www.uni-jena.de/~pfk/mpp/bin/in_mpc_0.93a_en.zip

The bitrate indicator in Winamp has gone into orbit. It now skips around at about 10Hz updates!

Edit musepack.ini in Plugins directory, change UpdateBitrate=1 to UpdateBitrate=38. That will change the update speed to 2 seconds.

thank you very much JohnV......

Using the same quality settings, I'm getting way more decode overloads with 1.02 than with 1.01j.
With replaygain/clipprotect that means a few more dB of attenuation.

Went back to 1.01j -- any reason to use 1.02 other than the low-bitrate improvements?