Audacity Forum

Hypothetical situation:
I have an audio file that I suspect is a transcode, ie. Transcoded from 128 to 320 kbps.
I now also have a legit 320 kbps copy.
Here is an example of the suspected transcode (spectrum mode)

Here is an example of a legit file (spectrum mode)


Would I be correct to assume that the more red or bright yellow/white there is in the top part of the image the more likely it is to be a transcode or is it just due to the content of the file?

Note: The two images are not of the same file

ShannenName wrote:Would I be correct to assume that the more red or bright yellow/white there is in the top part of the image the more likely it is to be a transcode

No, that is not a safe assumption. There is very little spectral difference between a 128kbps encoded file and a 320kbps encoded file. The difference due to the content of different recordings will be much more significant than the small difference in spectrum due to the encoding.

You may notice in the "Frequency Analysis window" (Analyze menu > Plot Spectrum) that a transcoded 128>320kbps file has a frequency cut-off at around 16kHz, but this is not "proof" that the file has been transcoded. There could be other reasons why frequencies above 16kHz are absent. Similarly, the presence of frequencies above 16kHz does not prove that the file has not been transcoded, though it would suggest that this is probably the case.

Just because a file is encoded in a high quality format (320kbps) does not necessarily mean that it will be high quality. 320kbps is capable of high quality audio, but it is still possible to make a badly encoded file, even at 320kbps. For example, MP3 encoding will often boost the peak level a little which can result in clipping distortion - it is therefore best to allow a little headroom before encoding. A 320kbps MP3 that has clipping distortion will almost certainly sound worse than a 128kbps encoding that does not have clipping distortion.

stevethefiddle wrote:No, that is not a safe assumption. There is very little spectral difference between a 128kbps encoded file and a 320kbps encoded file. The difference due to the content of different recordings will be much more significant than the small difference in spectrum due to the encoding.

Thank you this is what I suspected but a friend of mine seemed to have a different idea.
Didn't know about the 16khz cutoff though but now that I think about it, it seems logical as lower encodings sometimes cutoff frequencies that we can't hear or can't hear as much.

You can test this for yourself easily enough if you are interested. This is easiest to do in Audacity 1.3.12 as this later version allows you to set the MP3 compression amount in the Export dialogue (the "Options" button) which is a lot easier than digging into Audacity Preferences to change the compression amount.

Simply import a short WAV file (something ripped from CD would be ideal) and export it several times as MP3 with different bit rates. Label them song_title_128.mp3, song_title_256.mp3 ..... then import them back into Audacity and have a look.

If you want to try Audacity 1.3.12 (I recommend it - it has a lot of improvements over the old 1.2.6 version) you don't need to uninstall 1.2.6 first. Both versions can live side by side on your computer, just run one at a time. Limitation: Audacity 1.2.6 can not open projects that have been saved by Audacity 1.3.x

stevethefiddle wrote:You can test this for yourself easily enough if you are interested. This is easiest to do in Audacity 1.3.12 as this later version allows you to set the MP3 compression amount in the Export dialogue (the "Options" button) which is a lot easier than digging into Audacity Preferences to change the compression amount.

Simply import a short WAV file (something ripped from CD would be ideal) and export it several times as MP3 with different bit rates. Label them song_title_128.mp3, song_title_256.mp3 ..... then import them back into Audacity and have a look.

If you want to try Audacity 1.3.12 (I recommend it - it has a lot of improvements over the old 1.2.6 version) you don't need to uninstall 1.2.6 first. Both versions can live side by side on your computer, just run one at a time. Limitation: Audacity 1.2.6 can not open projects that have been saved by Audacity 1.3.x

Will upgrade to 1.3.12 as soon as possible.

I used Mercyful Fate's "Evil" from the album "Melissa". It's perhaps not the best test subject as the range is limited to compared to other songs but it is the only album I have in WAV format that I can confirm is legit right now and it will do for all intents of this experiment.
Here are my results:

128 kbps:
Spectrum View

Analyse->Plot Spectrum View


160 kbps:
Spectrum View

Analyse->Plot Spectrum View


192kbps:
Spectrum View

Analyse->Plot Spectrum View


320 kbps:
Spectrum View

Analyse->Plot Spectrum View


Uncompressed WAV:
Spectrum View

Analyse->Plot Spectrum View


128kbps transcoded to 320kbps:
Spectrum View

Analyse->Plot Spectrum View


128kbps transcoded to WAV:
Spectrum View

Analyse->Plot Spectrum View


Conclusion:
As you can see the 128 file is the same as the 128 transcoded to 320 and the 128 transcoded to wav therefore they sound EXACTLY the same.
From 128 to the legit WAV you can see a decrease in the drop off of sound (in plot spectrum view) after the 16KHz therefore resulting a better sounding (subjective) music file.
In spectrum mode there is little difference between bitrates and therefore using spectrum mode to detect transcodes in NOT an effective or reliable way to detect transcodes.
Also WAV files are approximately 10x the size of 320kbps files but not offer much better quality in sound.

Additional thoughts:
I think there should be a plugin to compare or analyse files for suspected transcodes because when I was downloading 320kbps songs from the Internet (don't jump on my back. I want them on my computer and my cd drive is broken. I own legit copies of the cd's.) there was a lot of "high-quality" files that sounded like shit.


Not related but does 1.3.12 support importing files that have non-English characters in the file name because I tried importing an MP3 that had Russian characters in its file name and it come up with an error?

In Audacity 1.3.12 if you go to "Edit menu > Preferences > Spectrograms" you can adjust the settings for the spectrum view.
Increase the "Window Size" for higher resolution (at the expense of slower rendering).
If you increase the "Maximum Frequency (Hz)" you will be able to see the high frequency drop off for low bit rate MP3s.

ShannenName wrote:I think there should be a plugin to compare or analyse files for suspected transcodes because when I was downloading 320kbps songs from the Internet ... there was a lot of "high-quality" files that sounded like [email protected]%!.

Do you really need a plug-in to tell you something that your ears have already told you? At the end of the day we're more concerned with what they sound like than the specification.

stevethefiddle wrote:In Audacity 1.3.12 if you go to "Edit menu > Preferences > Spectrograms" you can adjust the settings for the spectrum view.
Increase the "Window Size" for higher resolution (at the expense of slower rendering).
If you increase the "Maximum Frequency (Hz)" you will be able to see the high frequency drop off for low bit rate MP3s.

Thanks

ShannenName wrote:I think there should be a plugin to compare or analyse files for suspected transcodes because when I was downloading 320kbps songs from the Internet ... there was a lot of "high-quality" files that sounded like [email protected]%!.

Do you really need a plug-in to tell you something that your ears have already told you? At the end of the day we're more concerned with what they sound like than the specification.

True but someone might need it.