Uploading blocked audio recordings to VKontakte
I have long been a reader of Habr, but have not written a single article yet. To write this article I pushed this post: Lock audio records Vkontakte. Restore the playlist .
I liked the methods used by the author to bypass the VKontakte filter, but as a person who has been involved in sound processing and writing music for a long time, I can say that this is not the best decision to take in this situation, and here's why:
1. The quality of audio recording deteriorates, with multiple conversions and deliberate degradation of quality.
2. Cutting out a piece of work changes the music.
I suggest a slightly different approach to solving this problem. Who is interested in the topic - please under the cat. I will tell you more about several methods by which you can bypass the filter, including the method that allows you to download blocked music without changing its content (i.e., completely preserving the sound).
All tests were carried out in the Audacity program, the song performed by rapper Pitbull and singer Marc Anthony, the name Rain Over Me was chosen for analysis. This song is suitable for the test for several reasons: it is “banned” on VKontakte, its spectrum contains all frequencies in full, it has both “quiet” and “loud” parts, which will allow you to experiment with dynamics. To begin with, make sure that it is impossible to listen to the audio recording, because it has been deleted at the request of the copyright holder:
')
As you can see, that's right, the original audio recording has been removed, and if we try to download the same one, we will receive a warning about copyright infringement:
Downloading the same audio recording, but under a different name, was also unsuccessful. Tags in both posts (original and version with a modified name) - were absent.
Now let's try to download the same audio recording, slightly changing it in some way.
In my opinion, this is the easiest method. Consider several uses for EQ:
Here I immediately bring the result of what I came to, in fact, I unreasonably began to try to “feed” VKontakte audio with a slightly changed tonal balance, so that it was almost impossible to hear the difference (for example, I used a low-pass filter and high-pass filter with very soft settings - the difference in sound was insignificant). VKontakte rejected entries with such changes, but, interestingly, accepted, processed like this:
This means that the recognition is only in the midrange, high and low - are not affected. The first option is equalization of the entire track: it is necessary to cut out a small area of ​​the mid-frequency range. I tried to apply EQ only on a small section of the track - it did not work.
In the comments to the article, the user Gorthauer87 suggested “to slightly mix up some noise that the ear could not see”. The thought is correct, in fact it was the very first one that came to my mind. In the recording industry, dithering has been used for a long time, which is nothing more than adding noise, although noise is now in the stage of noise shaping - i.e. the frequencies most audible to the human ear are removed from it.
Adding white noise in a quantity sufficient to clearly hear it did not help to load the audio recording. This method does not suit us, because if the noise can be heard, then we can assume that the audio recording is corrupted and is not subject to listening = (
After processing the entire track with a compressor with fairly “soft” settings, the audio recording was successfully loaded. I give a screenshot of the settings of the compressor:
Hearing the difference with the original unprepared listener is problematic, and if you listen with headphones or not very high-quality acoustics, it is impossible.
But this method surprised me with its simplicity, while it turned out to be the highest-quality of all the ones considered earlier. You just need to add a small piece of an arbitrary signal to the end of the track (already after the music ended). As a result, the useful signal does not change at all, but the VKontakte site does not try to block such tracks. Specifically, in this example, I added a sine wave of 440 hertz to the end of the product, the duration of the noise is about 1 minute.
If this article is interesting for readers, then I can take a closer look at each of the ways to download music (I guess the “corona” method will soon be “covered”), depending on the material (classic, rock, pop, electronics). Each of the options will work better on one material and worse on another, that is, somewhere more noticeable to a person, and somewhere practically without changing the source code.
The article was not as complete as I would like, for the reason that my DAW refused to work and began to give an error (so I did all the work in Audacity). I assume that there are several other ways that work:
1. Work with stereo base (its expansion / contraction)
2. Processing of the entire track by multiband exciter
3. Multiband compression of the mid-frequency range, to a much lesser extent than discussed in the article.
4. Adding a very weak reverb signal, indistinguishable by the human ear.
5. Expansion , both single and multiband.
If there is an opportunity to work on a powerful DAW, then I can consider these cases as well.
I am not a supporter of copyright infringement, as I myself create music, but I believe that the copyright situation has now passed all permissible limits. Prohibitions and restrictions are only temporary solutions that do not eliminate piracy, it is necessary to solve the problem completely, stimulating alternative means spreading music. In fact, who of those people who suffered from the deletion of the VKontakte audio recordings will go to the store and buy a CD? Or buy music online?
* A sample of the audio was taken from youtube.com, where it was posted in the public domain by the copyright holder. Audio recordings were used exclusively for research purposes.
I liked the methods used by the author to bypass the VKontakte filter, but as a person who has been involved in sound processing and writing music for a long time, I can say that this is not the best decision to take in this situation, and here's why:
1. The quality of audio recording deteriorates, with multiple conversions and deliberate degradation of quality.
2. Cutting out a piece of work changes the music.
I suggest a slightly different approach to solving this problem. Who is interested in the topic - please under the cat. I will tell you more about several methods by which you can bypass the filter, including the method that allows you to download blocked music without changing its content (i.e., completely preserving the sound).
All tests were carried out in the Audacity program, the song performed by rapper Pitbull and singer Marc Anthony, the name Rain Over Me was chosen for analysis. This song is suitable for the test for several reasons: it is “banned” on VKontakte, its spectrum contains all frequencies in full, it has both “quiet” and “loud” parts, which will allow you to experiment with dynamics. To begin with, make sure that it is impossible to listen to the audio recording, because it has been deleted at the request of the copyright holder:
')
As you can see, that's right, the original audio recording has been removed, and if we try to download the same one, we will receive a warning about copyright infringement:
Downloading the same audio recording, but under a different name, was also unsuccessful. Tags in both posts (original and version with a modified name) - were absent.
Now let's try to download the same audio recording, slightly changing it in some way.
Filter traversal methods
1. Changing the tonal balance of the entire work, or part of it, that is, processing by the equalizer:
In my opinion, this is the easiest method. Consider several uses for EQ:
Here I immediately bring the result of what I came to, in fact, I unreasonably began to try to “feed” VKontakte audio with a slightly changed tonal balance, so that it was almost impossible to hear the difference (for example, I used a low-pass filter and high-pass filter with very soft settings - the difference in sound was insignificant). VKontakte rejected entries with such changes, but, interestingly, accepted, processed like this:
This means that the recognition is only in the midrange, high and low - are not affected. The first option is equalization of the entire track: it is necessary to cut out a small area of ​​the mid-frequency range. I tried to apply EQ only on a small section of the track - it did not work.
2. Adding noise
In the comments to the article, the user Gorthauer87 suggested “to slightly mix up some noise that the ear could not see”. The thought is correct, in fact it was the very first one that came to my mind. In the recording industry, dithering has been used for a long time, which is nothing more than adding noise, although noise is now in the stage of noise shaping - i.e. the frequencies most audible to the human ear are removed from it.
Adding white noise in a quantity sufficient to clearly hear it did not help to load the audio recording. This method does not suit us, because if the noise can be heard, then we can assume that the audio recording is corrupted and is not subject to listening = (
3. Work with track dynamics
After processing the entire track with a compressor with fairly “soft” settings, the audio recording was successfully loaded. I give a screenshot of the settings of the compressor:
Hearing the difference with the original unprepared listener is problematic, and if you listen with headphones or not very high-quality acoustics, it is impossible.
"Crown" method
But this method surprised me with its simplicity, while it turned out to be the highest-quality of all the ones considered earlier. You just need to add a small piece of an arbitrary signal to the end of the track (already after the music ended). As a result, the useful signal does not change at all, but the VKontakte site does not try to block such tracks. Specifically, in this example, I added a sine wave of 440 hertz to the end of the product, the duration of the noise is about 1 minute.
If this article is interesting for readers, then I can take a closer look at each of the ways to download music (I guess the “corona” method will soon be “covered”), depending on the material (classic, rock, pop, electronics). Each of the options will work better on one material and worse on another, that is, somewhere more noticeable to a person, and somewhere practically without changing the source code.
The article was not as complete as I would like, for the reason that my DAW refused to work and began to give an error (so I did all the work in Audacity). I assume that there are several other ways that work:
1. Work with stereo base (its expansion / contraction)
2. Processing of the entire track by multiband exciter
3. Multiband compression of the mid-frequency range, to a much lesser extent than discussed in the article.
4. Adding a very weak reverb signal, indistinguishable by the human ear.
5. Expansion , both single and multiband.
If there is an opportunity to work on a powerful DAW, then I can consider these cases as well.
Conclusion
I am not a supporter of copyright infringement, as I myself create music, but I believe that the copyright situation has now passed all permissible limits. Prohibitions and restrictions are only temporary solutions that do not eliminate piracy, it is necessary to solve the problem completely, stimulating alternative means spreading music. In fact, who of those people who suffered from the deletion of the VKontakte audio recordings will go to the store and buy a CD? Or buy music online?
* A sample of the audio was taken from youtube.com, where it was posted in the public domain by the copyright holder. Audio recordings were used exclusively for research purposes.
Source: https://habr.com/ru/post/183554/
All Articles