These Audio Submissions Should Be Rated A!
These Audio Submissions Should Be Rated A!
These Audio Submissions Should Be Rated M!
Stolen from Imase - Night Dance -- Seems to be an english version of the Japanese version.
Stolen from https://www.newgrounds.com/audio/listen/946764
Also stolen by same user:
These Audio Submissions Should Be Rated M!
These Audio Submissions Should Be Rated T!
Need a second opinion on this user and their songs. I'm a little suspicious that they are using AI, but it's hard to tell.
This was the song I was looking at the most:
It just hits those same odd tones that Suno and Udio often hit, which instantly triggered that "AI uncanny valley" that I've grown so accustomed to. However, it's still pretty well composed, and the bitrate isn't 192kbps (max Suno AI output), hence my confusion. Also, the descriptions are lazy (common with AI uploaders), but that doesn't necessarily mean they're uploading AI music. They go by "fxstyxx" on YouTube, and that matches the metadata of the tracks on Newgrounds.
Anyone have any insight to pitch in?
Look at this user taking suggestions for songs; definitely not just copy-pasting the google form responses into Suno AI, right? ...Right...?
Suno:
At 5/29/24 04:05 AM, Questtrek wrote: Need a second opinion on this user and their songs. I'm a little suspicious that they are using AI, but it's hard to tell.
This was the song I was looking at the most:
https://www.newgrounds.com/audio/listen/1324293
It just hits those same odd tones that Suno and Udio often hit, which instantly triggered that "AI uncanny valley" that I've grown so accustomed to. However, it's still pretty well composed, and the bitrate isn't 192kbps (max Suno AI output), hence my confusion. Also, the descriptions are lazy (common with AI uploaders), but that doesn't necessarily mean they're uploading AI music. They go by "fxstyxx" on YouTube, and that matches the metadata of the tracks on Newgrounds.
Anyone have any insight to pitch in?
they could be reencoding the ai tracks into something that isnt 192kbps
"Has the light gone out for you? Because the light's gone out for me..."
At 5/29/24 05:21 PM, wolfatthedoor wrote:At 5/29/24 04:05 AM, Questtrek wrote: Anyone have any insight to pitch in?they could be reencoding the ai tracks into something that isnt 192kbps
This is possible. But even if so the waveform will still LOOK like 192 kbps. You can download it and see. The quality if reencoded will ALSO be 192 kbps.
Unless they are also upscaling it. Which in a lot of cases will STILL sound like crap, and I don't even know of an upscaler. It is possible there are other AI "music" sites which have bitrates higher than 192 as well.
Your source for monthly music producer freebies here // Take My Cymbals // ALL my big sample projects, FREE
I do professional audio critique & commissions. Catch me on YT and X! If you got music, I'll playlist you!
At 5/29/24 06:02 PM, ADR3-N wrote:At 5/29/24 05:21 PM, wolfatthedoor wrote:This is possible. But even if so the waveform will still LOOK like 192 kbps. You can download it and see. The quality if reencoded will ALSO be 192 kbps.At 5/29/24 04:05 AM, Questtrek wrote: Anyone have any insight to pitch in?they could be reencoding the ai tracks into something that isnt 192kbps
Unless they are also upscaling it. Which in a lot of cases will STILL sound like crap, and I don't even know of an upscaler. It is possible there are other AI "music" sites which have bitrates higher than 192 as well.
Okay, I analyzed 7eunnn’s music and put it through a spectrogram. You can look at the max kHz the song reaches to determine its original bitrate, even if someone just threw it in Audacity first and uploaded it as a 320kbps track.
An MP3 file with a bitrate of 192kbps should rarely exceed 19kHz. I’ve uploaded an example spectrogram of a Suno AI generation I downloaded a couple days ago. As you can see, there’s a very clear cutoff at 19kHz. However, I’ve noticed that many Suno AI songs will have a lower peak kHz than 19. I can upload more examples if necessary, but just as a rule of thumb, Suno AI songs shouldn’t exceed 19kHz as a peak frequency.
For comparison, I also ran Creo’s “Sky and Soul” (downloaded as a 320kbps MP3 from Newgrounds), and its cutoff is at 20kHz, nearly perfectly. This is the max kHz of an MP3 encoded at 320kbps, which matches the metadata of the file. (This is included in the google drive link below, I can only put two images in a post.)
Now let's take a look at the spectrogram of “Electric Shock” by 7eunnn. It’s suspicious to say the least. There’s a few random peaks (more like fuzz) jumping up to 20khz, but most of the track follows along the line of 19kHz, which is usually indicative of an MP3 encoded at 192kbps. I’m not sure what causes this “fuzz”, but it could be from upscaling or just an artifact of how it was encoded. Or it could just be how it was mixed. I don't know. Regardless, most of the song's max frequency sits squarely at 19kHz.
The other songs are a mixed bag. Some are more questionable than others. “Pulse” has an extremely low bitrate for some reason. “Random” has a line at 17kHz with fuzz up to 20kHz. “Incognito” has much of the song at 19kHz, but there’s notable sections at 20kHz. Like I said, a mixed bag. Here’s all the spectrogram results, also in a higher resolution:
https://drive.google.com/drive/folders/1yToxxK8w0rngyXZL0geLSSLw1bw7gNjL?usp=sharing
You suggested that this could be another AI Music generator and I think that this could be Udio AI (it's another common AI music creator). I’m not very familiar with Udio, but I know it exports at a higher bitrate. I also am not experienced enough to simply listen to the track by ear and “know” what bitrate it is. I only know how to recognize artifacts in AI, and dig through metadata and spectrographic evidence if necessary. And all the evidence I have (at least for electric shock) is either common with AI uploaders, or is suspicious. First of all, the album art on Electric Shock is misaligned and looks mildly AI-generated. Second, it hits all those notes that AI does; those high awkward, garbled notes. Third, the spectrogram is a little suspicious too, since most of the song maxes out at 19kHz, showing that it might have originally been 192kbps, the max output of Suno AI. Also, 7eunnn seems desperate to get scouted and whitelisted for Geometry Dash as seen by the tags, never a great sign. Obviously, I could be missing something critical (I’m not a musician, composer, or expert in spectrograms), so please use your best judgment.
@G2961 Any insight on if this is Udio? You’re more familiar with it than I am.
Thanks,
Quest
At 5/29/24 08:25 PM, Questtrek wrote: A LOT of great information!
I'm gonna look over at that drive straight away and check out the user in question.
One thing I wanted to say before I do is that it's possible to train yourself to hear the difference in 192 kbps music and other bitrates by ear simply by taking 2 of your favorite songs, and then listen to them at various bitrates! There's an easy way to do this, by simply searching them up on a YT downloader or something and selecting to download in various formats, so you can hear how the compression and different frequencies react to the change in bitrate!
You can download 4 different bitrates, ofc including 192 and 320 kbps, but the lower and higher ones will give your ear an idea what it's listening for!
EDIT: You were on point. ALL those tracks pinged my spidey senses.
Your source for monthly music producer freebies here // Take My Cymbals // ALL my big sample projects, FREE
I do professional audio critique & commissions. Catch me on YT and X! If you got music, I'll playlist you!
At 5/29/24 10:15 PM, ADR3-N wrote:At 5/29/24 08:25 PM, Questtrek wrote: A LOT of great information!I'm gonna look over at that drive straight away and check out the user in question.
One thing I wanted to say before I do is that it's possible to train yourself to hear the difference in 192 kbps music and other bitrates by ear simply by taking 2 of your favorite songs, and then listen to them at various bitrates! There's an easy way to do this, by simply searching them up on a YT downloader or something and selecting to download in various formats, so you can hear how the compression and different frequencies react to the change in bitrate!
You can download 4 different bitrates, ofc including 192 and 320 kbps, but the lower and higher ones will give your ear an idea what it's listening for!
EDIT: You were on point. ALL those tracks pinged my spidey senses.
so we should prob call in someone to nuke all the tracks in question, i suppose?
"Has the light gone out for you? Because the light's gone out for me..."
At 5/29/24 10:37 PM, wolfatthedoor wrote: so we should prob call in someone to nuke all the tracks in question, i suppose?
Don't worry, already nuked. You can never be ENTIRELY sure with these, but based on the muddy quality of well, the whole track and the low bit rate, plus the million other tracks I've heard out of Suno, I wiped 'em. The descriptions claimed 3 different DAWs in the tags, which was totally unnecessary for the genre of dnb, and the tags screamed GD spam.
But in the case we ever are wrong, things can always be un-nuked. Bear that in mind. If you think a song fits the criteria, err on the safe side and report it, just with the caveat you aren't 100% sure.
Your source for monthly music producer freebies here // Take My Cymbals // ALL my big sample projects, FREE
I do professional audio critique & commissions. Catch me on YT and X! If you got music, I'll playlist you!
At 5/29/24 08:25 PM, Questtrek wrote:At 5/29/24 06:02 PM, ADR3-N wrote:Okay, I analyzed 7eunnn’s music and put it through a spectrogram. You can look at the max kHz the song reaches to determine its original bitrate, even if someone just threw it in Audacity first and uploaded it as a 320kbps track.At 5/29/24 05:21 PM, wolfatthedoor wrote:This is possible. But even if so the waveform will still LOOK like 192 kbps. You can download it and see. The quality if reencoded will ALSO be 192 kbps.At 5/29/24 04:05 AM, Questtrek wrote: Anyone have any insight to pitch in?they could be reencoding the ai tracks into something that isnt 192kbps
Unless they are also upscaling it. Which in a lot of cases will STILL sound like crap, and I don't even know of an upscaler. It is possible there are other AI "music" sites which have bitrates higher than 192 as well.
An MP3 file with a bitrate of 192kbps should rarely exceed 19kHz. I’ve uploaded an example spectrogram of a Suno AI generation I downloaded a couple days ago. As you can see, there’s a very clear cutoff at 19kHz. However, I’ve noticed that many Suno AI songs will have a lower peak kHz than 19. I can upload more examples if necessary, but just as a rule of thumb, Suno AI songs shouldn’t exceed 19kHz as a peak frequency.
For comparison, I also ran Creo’s “Sky and Soul” (downloaded as a 320kbps MP3 from Newgrounds), and its cutoff is at 20kHz, nearly perfectly. This is the max kHz of an MP3 encoded at 320kbps, which matches the metadata of the file. (This is included in the google drive link below, I can only put two images in a post.)
Now let's take a look at the spectrogram of “Electric Shock” by 7eunnn. It’s suspicious to say the least. There’s a few random peaks (more like fuzz) jumping up to 20khz, but most of the track follows along the line of 19kHz, which is usually indicative of an MP3 encoded at 192kbps. I’m not sure what causes this “fuzz”, but it could be from upscaling or just an artifact of how it was encoded. Or it could just be how it was mixed. I don't know. Regardless, most of the song's max frequency sits squarely at 19kHz.
The other songs are a mixed bag. Some are more questionable than others. “Pulse” has an extremely low bitrate for some reason. “Random” has a line at 17kHz with fuzz up to 20kHz. “Incognito” has much of the song at 19kHz, but there’s notable sections at 20kHz. Like I said, a mixed bag. Here’s all the spectrogram results, also in a higher resolution:
https://drive.google.com/drive/folders/1yToxxK8w0rngyXZL0geLSSLw1bw7gNjL?usp=sharing
You suggested that this could be another AI Music generator and I think that this could be Udio AI (it's another common AI music creator). I’m not very familiar with Udio, but I know it exports at a higher bitrate. I also am not experienced enough to simply listen to the track by ear and “know” what bitrate it is. I only know how to recognize artifacts in AI, and dig through metadata and spectrographic evidence if necessary. And all the evidence I have (at least for electric shock) is either common with AI uploaders, or is suspicious. First of all, the album art on Electric Shock is misaligned and looks mildly AI-generated. Second, it hits all those notes that AI does; those high awkward, garbled notes. Third, the spectrogram is a little suspicious too, since most of the song maxes out at 19kHz, showing that it might have originally been 192kbps, the max output of Suno AI. Also, 7eunnn seems desperate to get scouted and whitelisted for Geometry Dash as seen by the tags, never a great sign. Obviously, I could be missing something critical (I’m not a musician, composer, or expert in spectrograms), so please use your best judgment.
@G2961 Any insight on if this is Udio? You’re more familiar with it than I am.
Thanks,
Quest
Song was deleted now, lol
This Audio Submission Should Be Rated A!
These Audio Submissions Should Be Rated M!
A few more AI Songs:
Full account of AI songs. Pretty sure it’s Suno AI v2, as it has a 128kbps bitrate, and has less noise. Also the date they were uploaded aligns with the timing of Suno v2’s lifespan. Their first uploaded song admits to being made with Suno AI anyway.
I’m confused with this one. It appears to be a fantasy band created by the user about Club Penguin, which is fine, but I’m suspicious that it’s all AI. The bitrate is 128kbps, so I can’t immediately determine if it’s Suno v3 via the bitrate (and a spectrogram would be of little use). I have other suspicions though. First of all, the lyrics. They’re so specific to Club Penguin that I feel that it was written by an AI, which is a feature in Suno. I've heard lyrics like this before on other AI songs. “The Island Of Harmony” is a good example of these hyperspecific lyrics. Second, the user doesn’t ever credit or explain who or where the vocals come from. “The vocalist” sounds a little different on some songs. Please take a look at this user.
Thieves:
At 6/2/24 02:56 PM, MatthewMcChanIsBack wrote: https://www.newgrounds.com/audio/listen/443694
I don't think this song, "Villa - Young Rida" contains any material that would need to make it M. I listened to it, and I never heard any strong language or adult subject material. There's a brief line about a cop being higher than the rapper, but that's about it. It should be T at most, but E probably is fine too. Let me know if I missed something.
At 6/2/24 02:56 PM, MatthewMcChanIsBack wrote: https://www.newgrounds.com/audio/listen/435282
This Audio Submission Should Be Rated T!
The song is "shoota- villa (intro 4 mixtpe)".
Don't know why you're flagging it as needing to be M rated, the only thing in this song is a single use of the word "shit" at 0:19. Are you listening to the songs you report, or are you just flagging it because it's rap? If I missed something, please elaborate why this song should be M.
