Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Recent Posts
1
FLAC / Re: FLAC v1.4.x Performance Tests
Last post by bennetng -
Quote from: Porcus on
(for this specific file!)
Quote from: Porcus on
More sines tested: https://hydrogenaud.io/index.php/topic,122444.0.html
To avoid misunderstanding, the method I mentioned above was undithered, but the file I posted on last page was RPDF dithered.

I've actually forgotten I posted a similar comment in another thread until you mentioned it so I did not intentionally use the same settings in waveform generation. Also, even with a prime number, adding dither can still further clean up the noise floor.

So here is the undithered version.
2
FLAC / Re: FLAC v1.4.x Performance Tests
Last post by Porcus -
Quote from: ktf on
Quote from: Porcus on
@ktf : out of "curiosity", is it possible to force-feed libflac a given predictor vector and test (6771 * 2-12, -1) vs  (27085 * 2-14, -1) vs  (54170 * 2-15, -1) on this file?
Sure, you can hack around in the encoder. It wouldn't make sense for any normal situation, because most of the time, the signal isn't the same throughout a whole file.

As said before, I think trying to optimize for sine waves is a waste of time: they trigger a specific condition in the encoder that isn't triggered on signals that either aren't a sine or change between the start and end of a block.

Yes, - or: those vectors (for this specific file!) would be a test whether this is the correct takeaway from it. Because that "specific condition" should not occur when searching for the best among second-order predictors (coefficient matrix is invertible at order = 2).
If those vectors improve, then you got a more general round-off issue at hand - which may in the larger picture be part of the explanation why -p sometimes does better than one would expect.

But those two other vectors do not improve - or at least are so near-tied that nobody bothers with the difference - then the encoder finds the best 2nd order predictor within the limitations of the format (that is, the bound on resolution), and "your" conclusion is reasonable. I.e., the opportunities for improvement are not worth looking more into, as they are indeed due to this very specific degeneracy (namely, the least-squares minimizer x solves Ax=b but with A singular/non-invertible; for sines that happens when order > 2).
3
Support - (fb2k) / Re: Replaygain for Onkyo TX-NR818 Receiver
Last post by Chibisteven -
Quote from: AMIlls001 on
Thank you for the quick follow up.  Unfortunately I need to connect to an analog input source.  I use Zone 2 out to an amp & speaker selector to play an additional 7 sets of speakers.   The Onkyo TX-NR818 Zone 2 will only work if the source input is an analog connection.   Right now I have an Arylic S10 connected to the receiver analog source (which allows me to wirelessly connect from my PC to the Receiver) and everything was working great accept MediaMonkey was playing many Flac files  at a very low volume and not adjusting for the Replaygaiin tags.  I was hoping that Foobar2000 would be different.  Does the source input have to be HDMI (or digital) for Foobar2000 to recognize the replaygain tags? Thank you

foobar2000 just reads and playbacks files on your computer.  What foobar2000 outputs is what gets sent to your chosen audio endpoint in your computer regardless if it's an analog or a digital connection that it's tied to which is ultimately inputted to your receiver in your case.  If using ReplayGain you may need to turn up the volume to compensate for the drop in overall volume.  Multiple zones are handled by your receiver which an application running on your computer can't see.
4
CD Hardware/Software / Re: playing ripped mp3 files in car shows band as "unknown"
Last post by fooball -
There are at least two ways to convey track information to a player, one being in the form of tag metadata within the MP3 file itself, the other by naming the file to something appropriate.

Then there's the question of where the information comes from.  Some CDs carry "CD text", which the ripper can interrogate and use (typically) to set the track filename appropriately.

When there is no CD text, the user has to enter the information manually, or the ripper can access a third-party enthusiast on-line database to try to identify the CD and download the track information (if the CD can be positively identified, and if somebody has contributed the metadata for it).

All these things can be configured in EAC, but if any of it does not produce the desired results you'll need to edit it in manually.  Tags can be manipulated in FB2K of course, but the best tool (IMO) is MP3Tag.

PS: in the tag world, "band" is known as "artist".
5
General Audio / Re: AI language models can exceed PNG and FLAC in lossless compression, says study
Last post by Porcus -
The point of this study seems to be that when you have a language modelling tool - "foundation models" in their lingo - it represents the "language" it models in a compressed way. And apparently not only the language it was trained on - which was written English, not recordings of spoken English. Supporting the notion that these models learn how to handle language generally, not just one particular language.

They did not at all try to compress a varied set of music, their test sets was ten hours of speech (mono, sampled at 16k), so the "2 CDs" objection is moot.

I downloaded it - it is is delivered as FLAC in .tar.gz , saving some percentages due to ... padding! @ktf, if the noise was split into similar-length files, then I think it would make up five-ish percentage points, explaining the 107?
The "clean" set (whatever that means), 2620 files was compressed to 142 kbit/s and the "other" set, 2939 files, to 136. They used flac 1.2.1, which doesn't matter much to the results: recompressing with 1.4.3 at -5 and -8p loses or gains 1 kbit/s. OptimFROG at --preset 8 gets them down to 130 resp. 123. So yeah, that is about as low it goes in audio compressors.

Of course their compressor itself - with the parameters - is yuuuuuge:
* The smallest of the Chinchilla models is twice this data set - 2 GB - and saves 5.5 points over FLAC. Breaks even with FLAC if you have 36 GB of speech.
* The biggest is 140 GB. So you need 1.5 TB to compress before it overtakes FLAC.
... and imagine the computing power required.
6
FLAC / Re: FLAC v1.4.x Performance Tests
Last post by ktf -
Quote from: Porcus on
@ktf : out of "curiosity", is it possible to force-feed libflac a given predictor vector and test (6771 * 2-12, -1) vs  (27085 * 2-14, -1) vs  (54170 * 2-15, -1) on this file?
Sure, you can hack around in the encoder. It wouldn't make sense for any normal situation, because most of the time, the signal isn't the same throughout a whole file.

As said before, I think trying to optimize for sine waves is a waste of time: they trigger a specific condition in the encoder that isn't triggered on signals that either aren't a sine or change between the start and end of a block.
7
General Audio / Re: AI language models can exceed PNG and FLAC in lossless compression, says study
Last post by ktf -
I must say the results of this study are very.... odd.

Apparently FLAC compresses the ImageNet data better than PNG, gzip and LZMA when offered in chunks of 2048 byte? That seems very unlikely? I'd say the results of LZMA, gzip, PNG and FLAC are suspiciously similar when considering that they work with completely different methods. Sure, PNG is more or less a general purpose filter with a specific context-aware filter, but how would you even feed non-image data to a PNG compressor? That context-aware filter depends on height and width of the image, which is not applicable for non-image data.

Also, I don't really understand the 'chunking' with the audio data. It seems to me they have chopped up the data in chunks of 2048 bytes and concatenated them. If the data is 16 bit per sample, that means only 1024 samples for each chunk, which really isn't representative for any kind of audio. Similar for pictures:

Quote
We extract contiguous patches of size 32 × 64 from all images, flatten them, convert them to grayscale (so that each byte represents exactly one pixel) to obtain samples of 2048 bytes. We then concatenate 488 821 of these patches, following the original dataset order, to create a dataset of 1 GB.
This doesn't seem in any way representative of any real-world use case? Image and audio data don't fit 1 byte per sample most of the time, so the data seems 'crafted' to me.

Finally, I don't understand where the 107% figure comes from for FLAC when compressing noise. It does much better than that. When I compress noise as an 8bps single channel stream with a blocksize of 2048 (as suggested in the paper) I get only 0.5% overhead, not 7%.
10
Support - (fb2k) / Re: Replaygain for Onkyo TX-NR818 Receiver
Last post by AMIlls001 -
Thank you for the quick follow up.  Unfortunately I need to connect to an analog input source.  I use Zone 2 out to an amp & speaker selector to play an additional 7 sets of speakers.   The Onkyo TX-NR818 Zone 2 will only work if the source input is an analog connection.   Right now I have an Arylic S10 connected to the receiver analog source (which allows me to wirelessly connect from my PC to the Receiver) and everything was working great accept MediaMonkey was playing many Flac files  at a very low volume and not adjusting for the Replaygaiin tags.  I was hoping that Foobar2000 would be different.  Does the source input have to be HDMI (or digital) for Foobar2000 to recognize the replaygain tags? Thank you