Evil's always been pretty damn good, but it does depend on the song that is chosen. Her voice works really good with punk and rock, things that are very vocal-forward and clear. She's actually less advanced vocally than Neuro herself, but Neuro has a stronger level of distortion on her voice since when Vedal tried to reduce the distortion, chat lost it because it changed her voice from robotic to more human-sounding and they didn't like it.
Can someone explain is this actually the AI Program doing the singing or is it being controlled by Vedal, like with how it should sing and move? Just seems so natural, like a Vtuber almost but Neuro is AI.
yes the neuro twins are a combination of an LLM, an RVC ai changing the voice of a stem, and a vtuber with preprogrammed movements that respond to the song (probably using the same technique as a visualizer, just more complicated.) and other stuff. if you're new to the swarm, welcome!
@@asxphxss I don't think it's RVC. I've watched a lot of AI covers since they started popping up, and there's a few things I've noticed that are different from the Neuros. Vocal range - Evil's range has been good, but until recently Neuro's range was middling. RVC is essentially tracing, but for voices. The AI voice will try to hit whatever high or low notes the stem did, to varying levels of success. Neuro's singing was very much in a middle ground area, regardless of what notes she should have been hitting. It's most obvious with songs like My Heart Will Go On where there are some pretty high notes that Neuro just doesn't try to hit. Even a cover artist would try to hit such notes, even if they couldn't go as high as the original singer, but Neuro didn't even try. Even on the most recent karaoke stream Neuro wasn't going as high as she should with songs like "Don't Stop Believing". Song Length - One thing I've noticed and seen other's comment on is that RVC songs have issues with songs over 5 minutes. After 4-5 minutes the recordings start having vocal glitches and artifacts, even if there was no issues before. Evil sang Bury The Light, a nearly 10 minute minute song with around 8 minutes of actual singing, and was pitch perfect the whole time. Enunciation - One thing I've seen commented on over and over is that people can more clearly understand the lyrics of songs when the Neuros sing them. An RVC cover, even of a cover singer, wouldn't have enunciation good enough for that to be a consistent observation. TL;DR - The Neuro's covers are too different from normal RVC covers to be using that or anything similar. I don't know what Vedal is doing, but it's a step or two above RVC covers.
@@spiritnova42 yes definitely I agree that it's not an RVC. It's probably a different model or method, but I think it's still stem changing/altering. For specific songs or for stuff like custom lyrics or vocaloid songs, It's definitely someone like queenpb doing the mixing, then putting what they've made into a stem splitter and running it through the model again. It's definitely multiple layers of work, with the last layer as the AI that changes the voice of the song to the individual twins' voices. At the end of the day we're just guessing, but we're probably in the ball park
@@asxphxss I'm not well versed in all of the options, I just know Neuro wasn't changing pitch much in either direction until recently. Either the stem was a monotone singer or something about changing the voice to Neuro's turned the vocal range down. Even now she doesn't go as high as she should on some songs, despite clearly being able to hit high notes in others. I know with RVC it's good practice to try to "match" the song to the ai voice, so pitch shifting it up or down if the voices are dissimilar. Could be a similar case here where some songs match better than others. But like you said, it's all just guesswork.
![Hope Is An Angel With A Shotgun (Alternative Version) | Robin, The Cab ft Astral Express [Mashup]](/img/n.gif)
Playing any game while blasting this song on full will boost your performance, focus, and coolness by 500%
Am I crazy or does Evil Neuro get better with every karaoke?
This has so much character and voice variations to it!
Evil's always been pretty damn good, but it does depend on the song that is chosen. Her voice works really good with punk and rock, things that are very vocal-forward and clear. She's actually less advanced vocally than Neuro herself, but Neuro has a stronger level of distortion on her voice since when Vedal tried to reduce the distortion, chat lost it because it changed her voice from robotic to more human-sounding and they didn't like it.
Angel with a shotgun always triggers my nightcore phase flashbacks
Wait her singing is so much better
Hoooooly shit this was my middle School
Well, 6th Grade
Well, first few months of 6th Grade
@@nhbons783 Gacha phrase go brrr...!
I started then switched accounts and stopped for a while until I discovered it again
@@henriquemachado9941 oh no
Yes people know it😆
nightcore speed is truth, but og still great
👑❤️😭
Can someone explain is this actually the AI Program doing the singing or is it being controlled by Vedal, like with how it should sing and move? Just seems so natural, like a Vtuber almost but Neuro is AI.
yes the neuro twins are a combination of an LLM, an RVC ai changing the voice of a stem, and a vtuber with preprogrammed movements that respond to the song (probably using the same technique as a visualizer, just more complicated.) and other stuff. if you're new to the swarm, welcome!
@@asxphxss Thanks for explaining. It’s really interesting stuff. Been watching more and more it’s like a rabbit hole.
@@asxphxss I don't think it's RVC. I've watched a lot of AI covers since they started popping up, and there's a few things I've noticed that are different from the Neuros.
Vocal range - Evil's range has been good, but until recently Neuro's range was middling. RVC is essentially tracing, but for voices. The AI voice will try to hit whatever high or low notes the stem did, to varying levels of success. Neuro's singing was very much in a middle ground area, regardless of what notes she should have been hitting. It's most obvious with songs like My Heart Will Go On where there are some pretty high notes that Neuro just doesn't try to hit. Even a cover artist would try to hit such notes, even if they couldn't go as high as the original singer, but Neuro didn't even try. Even on the most recent karaoke stream Neuro wasn't going as high as she should with songs like "Don't Stop Believing".
Song Length - One thing I've noticed and seen other's comment on is that RVC songs have issues with songs over 5 minutes. After 4-5 minutes the recordings start having vocal glitches and artifacts, even if there was no issues before. Evil sang Bury The Light, a nearly 10 minute minute song with around 8 minutes of actual singing, and was pitch perfect the whole time.
Enunciation - One thing I've seen commented on over and over is that people can more clearly understand the lyrics of songs when the Neuros sing them. An RVC cover, even of a cover singer, wouldn't have enunciation good enough for that to be a consistent observation.
TL;DR - The Neuro's covers are too different from normal RVC covers to be using that or anything similar. I don't know what Vedal is doing, but it's a step or two above RVC covers.
@@spiritnova42 yes definitely I agree that it's not an RVC. It's probably a different model or method, but I think it's still stem changing/altering. For specific songs or for stuff like custom lyrics or vocaloid songs, It's definitely someone like queenpb doing the mixing, then putting what they've made into a stem splitter and running it through the model again. It's definitely multiple layers of work, with the last layer as the AI that changes the voice of the song to the individual twins' voices.
At the end of the day we're just guessing, but we're probably in the ball park
@@asxphxss I'm not well versed in all of the options, I just know Neuro wasn't changing pitch much in either direction until recently. Either the stem was a monotone singer or something about changing the voice to Neuro's turned the vocal range down. Even now she doesn't go as high as she should on some songs, despite clearly being able to hit high notes in others. I know with RVC it's good practice to try to "match" the song to the ai voice, so pitch shifting it up or down if the voices are dissimilar. Could be a similar case here where some songs match better than others.
But like you said, it's all just guesswork.
I’ve only heard the nightcore version of this song. Which makes this not sound as good to me 😢
just play the video on 1.25 speed
@@michaelchen3362😂
@@michaelchen3362 It actually worked tho 😭
@@DullaTheDulla yea I wasnt joking LOL