Recent Posts

Listening Tests / Re: A killer sample for Opus 256 VBR

Last post by .halverhahn - Today at 11:20

The sample is not full-scale clipped, but it has some quite similar square-wave-character samples in a row.
I'm just guessing this creates some click sound that accentuate the drumbeat.

Lets see what happens when listening to a 20kHz lowpassed version of the original.

MPC / Origins of MPC 1.95z67 source code on Github ?

Last post by m00 - Today at 11:10

Hi, just wondering if anyone knows the origins of this source code on Github ?
https://github.com/ralph-irving/musepack
MPC codec 1.95z67

I've had a look through, and it seems to contain a more up to date version of some things, and very much work in progress in places, but it seems like an evolution of the source on the Musepack site for the 1.16 encoder..

The main things I'm curious about are the results of the Psychoacoustic modifications, to having 1 analysis windows per sub-frame, withthe familiar 256 sized transient analysis, but now with ultra-short 64 sized transient analysis as well, and also curous about the idea behind the noise compensation in the quantiser, the 2D one in particular, though in the current code it was disabled, so I wonder..

Does anyone know where this came from ? Is it some FK later code ?
Just a very curious bunny

Listening Tests / Re: A killer sample for Opus 256 VBR

Last post by Teddy_the_barber - Today at 11:07

Quote from: .halverhahn on Today at 10:27

I assume the cause for the difference are the clipped drumbeats in the original, the Opus version is missing them (see attached screenshot).

Interesting. The difference is too high in frequency though. I don't think this would be audible. The waveform has no clipping neither in the flac nor in the opus file.

Support - (fb2k) / Re: Bugged playback time reported by foobar

Last post by Case - Today at 11:07

Why are you using old 1.6.x version? I know Peter found null output related bugs in 2023 and 2024, which I assume are fixed in 1.6.18. Definitely fixed in 2.x series. I think it's highly likely your problem is related to null output.

Listening Tests / Re: A killer sample for Opus 256 VBR

Last post by Teddy_the_barber - Today at 11:03

Quote from: .halverhahn on Today at 10:27

I've got quite similar results on the first run. Listening window was from 2.5 sec to 12.8 sec.

Total: 6/8
p-value: 0.1445 (14.45%)

Spoiler (click to show/hide)

Code: [Select]

foo_abx 2.2.1 report
foobar2000 v2.24.1
2025-03-20 11:03:41

File A: Falling In Flames.flac
SHA1: 38252adad2987cb728ca6d5aac5bd718282ccdae
File B: Falling In Flames_opus256.opus
SHA1: 4328288c7ef522107e3a8215b0dae5752516df83

Output:
Default : Primary Sound Driver
Crossfading: NO

11:03:41 : Test started.
11:06:10 : Test restarted.
11:06:10 : 01/01
11:07:33 : Test restarted.
11:07:33 : 01/02
11:08:39 : Test restarted.
11:08:39 : 02/03
11:09:46 : Test restarted.
11:09:46 : 02/04
11:10:14 : Test restarted.
11:10:14 : 03/05
11:10:46 : Test restarted.
11:10:46 : 04/06
11:11:27 : Test restarted.
11:11:27 : 05/07
11:11:38 : Test restarted.
11:11:38 : 06/08
11:11:38 : Test finished.

 ---------- 
Total: 6/8
p-value: 0.1445 (14.45%)

 -- signature -- 
89bde793bfcfe37642363a4a2e3b27155d03c7ed

The thing is, drums on this track aren't even that pronounced, they're somewhat in the back. The accent in the mix is made on guitar riffs which are a lot louder and overwhelm my hearing quite fast so I can't reliably pass more trials.

Support - (fb2k) / Bugged playback time reported by foobar

Last post by regor - Today at 10:36

This is an error I have had sometimes but had not been able to reproduce it until now (and make captures)

Noticed it because some components or scripts were crashing when the float value was expected to be lower than a safe integer.

The console (%playback_time_seconds%):
- For JSP3: [11:27:24] JSP3 3,1638862116397E+134
- For SMP: [11:20:17] 3.1638862116397002e+134

The status bar (%playback_time%):
- CUI: 3150551671 Wk ...

Playlist: (%playback_time_seconds%):
- CUI: -9223372036854775808

Since the bugged value is provided to CUI and all components and display elements, it is not only a SMP error. I have experienced this at v1.6.17 (win 11)

The file is paused (?) but fb is playing. I am not sure how I reach this point, has happened to me sometimes when playing a file but pausing it too fast (I think).

I don't use any kind of components for loops or things like that.

Listening Tests / Re: A killer sample for Opus 256 VBR

Last post by .halverhahn - Today at 10:27

I've got quite similar results on the first run. Listening window was from 2.5 sec to 12.8 sec.

Total: 6/8
p-value: 0.1445 (14.45%)

Spoiler (click to show/hide)

Code: [Select]

foo_abx 2.2.1 report
foobar2000 v2.24.1
2025-03-20 11:03:41

File A: Falling In Flames.flac
SHA1: 38252adad2987cb728ca6d5aac5bd718282ccdae
File B: Falling In Flames_opus256.opus
SHA1: 4328288c7ef522107e3a8215b0dae5752516df83

Output:
Default : Primary Sound Driver
Crossfading: NO

11:03:41 : Test started.
11:06:10 : Test restarted.
11:06:10 : 01/01
11:07:33 : Test restarted.
11:07:33 : 01/02
11:08:39 : Test restarted.
11:08:39 : 02/03
11:09:46 : Test restarted.
11:09:46 : 02/04
11:10:14 : Test restarted.
11:10:14 : 03/05
11:10:46 : Test restarted.
11:10:46 : 04/06
11:11:27 : Test restarted.
11:11:27 : 05/07
11:11:38 : Test restarted.
11:11:38 : 06/08
11:11:38 : Test finished.

 ---------- 
Total: 6/8
p-value: 0.1445 (14.45%)

 -- signature -- 
89bde793bfcfe37642363a4a2e3b27155d03c7ed

I assume the cause for the difference are the clipped drumbeats in the original, the Opus version is missing them (see attached screenshot).

Speech Codecs / Using speech codecs to store Hi-Fi, stereo music like on CDs

Last post by MihaiPopa12346 - Today at 10:06

What I remember last days is:

Quote

If you make the Frankenstein of dividing a piece of music in 6 bands (each band has 4 kHz of bandwidth), divide in left and right channels, and encode each using TWELP at 3.6 kbps, you can store music with TWELP at 43.2 kbps, either 44.1 or 48 kHz and stereo!

So, that's what I said. Using speech codecs to store high-fidelity, stereo sound! Challenge is to write a tool to do that.
Maybe not the best idea but will be better than MP3 at the same bitrate.
Here is a table showing this in a more simple form:

Codec	Bitrate for each band and channel	Total bitrate for full-band, stereo music (44.1 or 48 kHz)	Approximate equivalent to more common codecs
TWELP	3.6 kbps	43.2 kbps	Opus or HE-AACv2 at 40/44 kbps
Codec2	3.2 kbps	38.4 kbps	MP3 at 40 kbps or Opus/HE-AACv2 at 36/40 kbps
Voxware MetaVoice	3 kbps	36 kbps	Opus or HE-AACv2 at 36 kbps
MetaVoice, TWELP, MELPe or Codec2	2.4 kbps	28.8 kbps	Opus or HE-AACv2 at 28 kbps
TWELP or Codec2	1.6 kbps	19.2 kbps	Opus, xHE-AAC or HE-AACv2 at 20 kbps
Codec2	1.4 kbps	16.8 kbps	Opus, xHE-AAC or HE-AACv2 at 16 kbps
Codec2	1.3 kbps	15.6 kbps	Opus, xHE-AAC or HE-AACv2 at 16 kbps
Codec2, MELPe or TWELP	1.2 kbps	14.4 kbps	Opus, xHE-AAC or HE-AACv2 at 12/14 kbps
Codec2 or TWELP	700 bps (0.7 kbps)	8.4 kbps	Opus or xHE-AAC in mono at 8 kbps
TWELP or MELPe	600 bps (0.6 kbps)	7.2 kbps	Opus or xHE-AAC in mono at 6/8 kbps
TWELP	480 bps (0.48 kbps)	5.76 kbps	Opus or xHE-AAC in mono at 6 kbps
TWELP	400 bps (0.4 kbps)	4.8 kbps	Close to xHE-AAC in mono at 6 kbps. Opus at 5 kbps
TWELP	300 bps (0.3 kbps)	3.6 kbps	Opus at 3/4 kbps

Support - (fb2k) / Re: Performance issue on 64bit: ASIO driver 2.2.4 high CPU usage

Last post by Case - Today at 09:52

Happy to hear you have finally upgraded from a 32-bit OS.

Out of curiosity, what kind of CPU usage do you get when not using ASIO?
The CPU use of entire foobar2000 process should be pretty close to 0% and the CPU usage of the ASIO host executables should be much smaller. I see it being at less than 0.01% for the majority of time. All it does is communicate with ASIO drivers. The 32-bit and 64-bit ASIO hosts are otherwise identical but they are just compiled for different architectures. One loads 32-bit driver, other one loads 64-bit driver. Difference in foobar2000's perspective is that one runs asiohost32.exe, the other one runs asiohost64.exe. There should be zero difference in performance.

I don't know if it helps anything, but you could use Process Explorer to open the process using 100% CPU and then check the Threads tab to see if it shows anything interesting. You'll instantly see the most taxing threads on top. For example here the majority of processing time is spent in ASIO driver's dll. Both it and asiohost.exe show <0.01% usage, but driver uses twice the cycles compared to foobar's code.

Validated News / Re: [OPEN SOURCE] corticod (neural multimedia cocec)

Last post by genuine - Today at 09:36

They should provide a list of the music used in education. And they must explain how they obtained them. But they can't do that.
These systems operate by exploiting countless works of countless artists without paying anything. You may support such works, but I cannot respect it.