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.
Topic: Is this indeed a transcoded WAV? (Read 4666 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Is this indeed a transcoded WAV?

I tried purchasing a track following from Junodownload but when I added the track to my cart I received the following warning:



Thinking that perhaps the label had failed to provide Juno with the WAV version I went to Beatport to look for the same file. I thought I was lucky when they did offer a lossless version, but after downloading it and running it through auCDtect I received the following result:

Code: [Select]
auCDtect: CD records authenticity detector, version 0.8.2
Copyright (c) 2004 Oleg Berngardt. All rights reserved.
Copyright (c) 2004 Alexander Djourik. All rights reserved.
------------------------------------------------------------
Processing file:        [249429_Sunset_On_Ibiza_Likuida_Remix.wav]
Data analysis:          [100%]
------------------------------------------------------------
This track looks like MPEG with probability 95%

So I went back and purchased the 320 MP3 from Juno and made Spectograms of both files:

WAV from Beatport:



MP3 from Junodownload:



They are practically identical, so am I correct in assuming that Beatport charged me a premium for a lossy transcode?  :o

Re: Is this indeed a transcoded WAV?

Reply #1 –
Hard to tell. One of the clearer indicators of MP3 loss on a spectrograph is something I call "temporal smearing", but you can't see that from this zoom level.

If you were to chop 30s of the two from 255 seconds to 285 seconds, I can take a look.

Re: Is this indeed a transcoded WAV?

Reply #2 –
Quote from: Canar on
Hard to tell. One of the clearer indicators of MP3 loss on a spectrograph is something I call "temporal smearing", but you can't see that from this zoom level.

If you were to chop 30s of the two from 255 seconds to 285 seconds, I can take a look.

Okay, I have created two 30s clips from each file and saved them into FLAC:

https://www.dropbox.com/s/6dqv3evray2ubb8/Beatport-WAV.flac?dl=0

https://www.dropbox.com/s/qsrd9oxqwaoc7wv/Juno-MP3.flac?dl=0

Could you perhaps also provide me with some info or a link on how to check for temporal smoothing as I have a few other tracks I would like to check.  :)

Re: Is this indeed a transcoded WAV?

Reply #4 –
Quote from: testyou on
You want to align tracks when you do this, such as in my examples.

Lol thanks, I completely missed seeing the place where I could upload an attachment. :-[ Lastly, and this was purchased from another site, but I was wondering if this spectrogram is a clearer indication of an MP3 that has been transcoded to WAV since there is no data above the 20kHz mark?



I've also captured a 30 sec clip for some to take a look at.

Re: Is this indeed a transcoded WAV?

Reply #5 –
That spectrogram certainly has that lossy look above 16 kHz.

Vendor sites accept whatever the content providers give them. It looks like maybe Juno accepts lossy formats and doesn't sell them as lossless. Others, though, quite reasonably require uploads be only WAV or AIFF. But according to one I talked to (Bandcamp), in the interest of keeping things simple for the uploaders, there is no education along the lines of "don't convert an MP3 to WAV; it won't recover what was lost when the MP3 was created". Consequently you get sites like Bandcamp (and apparently Beatport) selling transcoded files as lossless.

Re: Is this indeed a transcoded WAV?

Reply #6 –
You never know what was quality of samples, used for creating original file. So, even original can look like mp3 beacuse it could be created from mp3 samples or from samples transcoded from mp3.

Re: Is this indeed a transcoded WAV?

Reply #7 –
It's things like this that makes me wish there was more permanent metadata-like info that could be encoded with MP3s and then gleaned from transcodes. Alas with all the plethora of encoders, options, and legacy that type of thing is unlikely to happen, and we're stuck having to guesstimate from spectrograms.

For anyone buying lossless online they should be able to know if they're getting what they regularly pay more for, or if they're merely getting a lossy source in new clothes.

Re: Is this indeed a transcoded WAV?

Reply #8 –
If the only sign of lossiness is nothing above 16 kHz, it could have been recorded at a 32 kHz sampling rate.  For example, there's a mode on DAT recorders that does that.

Re: Is this indeed a transcoded WAV?

Reply #9 –
A sign of lossy encoding is "blockiness" in the upper spectrum, where singnal recedes below the threshold of hearing and is removed in one instant, but is present a moment later. If the recording is lowpassed, then quiet signals should be present up to the end of the cutoff where they are rolled off smoothly.

If some MP3 samples were used by the creator, then there should be other samples, or noise, that still extend across the full range of the spectrum somewhere in the song. For example, although these two cases show blockiness, I wouldn't call a transcode on them.


(full size)

(full size)

Analysis of the L-R (Side) channel can reveal blockiness where there is little stereo separation, even if both stereo channels look good.

Dolby AC-3 is a "deceptive" encoder. At high bitrates, it shows smearing, which might be strong and objectionable to some, but the spectrum is full up to 20 kHz, and with shaped dither added, AuCDTect often reports 100% CDDA conclusion on them. Without a reference it is difficult to detect the "spread". Luckily, AC-3 is not used that often in practice.

Re: Is this indeed a transcoded WAV?

Reply #10 –
Quote from: Xiphan on
Okay, I have created two 30s clips from each file and saved them into FLAC:
You also can see in a realtime spectrum how above 16kHz there are steep trenches coming and going. Once spotted you will have it easier to find the origin. Above 16kHz you can hide the pattern a bit with strong dither for the type of spectral views you use above.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!
 

Re: Is this indeed a transcoded WAV?

Reply #11 –
Yeah, looks like a lossy encode. I'd say damn near 100% certainty, but you simply /can't/ know with certainty like that.



Note the "audio pixel" that I have pointed out in the spectrogram. When things like that exist in isolation, that's a really good indicator of MP3. If you look at the rest of the spectrum, you'll see that it looks like the audio's made of little "audio pixels" like that. All of the "pixels" have similar length.  A proper lossless file would not look like that. Your resolution will end up being discrete when you zoom in, but the look is quite different.

I don't have any good lossless examples on hand here and am quickly whipping off a reply, but if I recall to after my evening (which is rather long), I'll try get you a better example.

If someone else gets what I'm saying here, kindly pitch in. :)

Note: much of the vocabulary I'm using here is neither technical nor anything more than neologisms I use internally to describe visual patterns that correlate with MP3 encoding. Please don't assume they're "proper" in any way.