Holy smokes... Adobe is clearly marketing their Thingy in the wrong way. What it did to your music was kind of awesome! I'd take the weird vocoded vocals it put out, clean them up, and merge them back into the original music track.
I'll be honest when I originally played with the software I figured it was just doing regular old vocal clean up. Using an AI to try and bring the voice wave form into what would he expected of clean audio. I had absolutely no idea they were essentially extrapolating what you were saying and then generating an AI replica voice performance. That has some really interesting potential.
@@snginther I imagine - assuming you're not inputting or outputting anything lossy - it'd likely hallucinate more and more audio as artifacts are added from the process. It'd likely also add in a bit of distortion and make it maybe voices a bit more robot-ey sounding. I think the robot voices would probably be more and more distinct and more words would be hallucinated.
I love videos where Philip is clearly just doing weird experiments purely to satiate his own curiosity, my favourites. Also, interesting to see you in Belfast! Hope you enjoyed yourself
It's not for noise eliminating only, it have other cool features, and also, mainly it is not a vst plugin, but a standalone program It is better to put files in program and after put out in daw, than using vst plugins
RX9 is a godsent for fixing terrible audio, I was tasked to fix an audio clip for some kind of indie movie that had the worst audio I had ever heard, really loud crickets in the background, 3 people speaking all at different distances from the mic with one being really loud, one being really quiet and one not pronouncing words correctly, it took me hours of fiddling around with the program but it came out almost perfect, in the end there was just one line that sounded a bit weird. This program is truly a miracle.
Honestly, some of iZotope's tools are like black magic. I don't mean to devalue their engineers' and programmers' work by saying so. I can't even begin to imagine what kinds of dsp are happening in the background to make those plugins work.
It is amazing how the AI enhanced recordings sound exactly like I would imagine based on looking at these AI generated images. It's an almost perfect audio recreation of those blurry face images.
7:30 That sounds like the AI is triyng to reproduce a frog sound rather than brown noise lmao. Also it seems to turn instrumental music into electronic pretty nicely.
Your kitty has those "sad eyes" that my male cat also has :D they look absolutely cute but so sad when trying to get food from my plate. Happy holidays, philip, the master of clicks.
Wow the first music (before Atomic Amnesia) was actually so cool with the AI voice on, if you worked to use a voice or robotic voice to had on top kinda like how the AI did it would sound awesome !
Your old video on voice/audio was probably the best guide out there for that thing and helps me a lot with my videos. Glad to see a newer video on the subject
I absolutely LOVE these sorts of videos, you're a genius Philip!! and if that was your dad in the video - you have one cool fucking dad, and I'm sure you owe a lot of your love for gaming to him. Also the closed captions on this video were at times hilarious xD Keep it up man!! much love!!
Ah, what a nice coincidence. I just spent the last few days working on a looper/delay for stereo audio, and it had the issue of eventually stacking up pass after pass of background noise when it's set to keep looping indefinitely (i.e. with unity feedback). It wasn't loud enough to notice on its own, but it certainly made a difference after some 20+ repetititons. I knew a noise removing algorithm would be possible to implement, though not knowing how to make one myself, I started digging and found one buried in the tutorials for the programming environment I'm using. Lo and behold, it indeed can sample the background noise during a "silent" moment and boom, the level meters show a blissful 0 dB on the incoming signal. Amazing. Big +1 to the effect of these processing methods being hard to figure out. Ears fatigue easily, and voices sound way different when played back than when hearing oneself speak. AI/ML-based methods only make it even more confusing and harder to predict. And since they're based on statistical averages and whatnot, they at most will give an approximation of how something would have seemed if it were more like the input samples. Also this all highlights that combined approaches will make the most sense.. the Adobe method seems to clearly need a more basic noise removal algorithm before it passes through its convolution matrices (or whatever the hell it's doing if it's not that).
"and i can then use those 3 seconds of footage BEFORE i bum myself as reference to show the computer the sound of silence. but this silent isnt silent." - 2kliksphilip 2022
Kudos for the HLL footage at 3:16 :) And really sensible cautious approach to de-noising. You don't neet any confirmation of your choices from random people on YT, but you can have it for free haha. On my own, I often listen to the removed noise extract when I de-noise, so I more easily can judge if I'm removing any useful audio information.
Holy shit you've GOTTA start using that Adobe tool in your music! Whatever it did, it had me grooving out in my seat. There's SO much musical potential in there to be explored!
Their technology has improved, but they've started charging. Would love to see an update video on the state of play with alternatives. All the best love your work.
I love all the videos you’ve been making about AI and life and whatever else you happen to be into. I’ll always look forward to your uploads Philip :-)
Hey philip, cool fact! If you press ctrl + r you can quickly use noise reduction, instead of having to scroll back through all again. I'm a fellow Noise Reduction enthusiast, so I understand the struggle.
This is exactly how I envisioned noise reduction in the future. Voice typing has been around for a long time. So AI should just analyze its voice and voice the recognized words. It remains only to make the option to create profiles so that you can record the voice of a friend and now it will work as a filter, replacing any voice with the voice of your friend.
The ways you used it are so cool! The music conversion segment made me giggle at how crazy it sounds. You can even hear remnatns of the music with the isolated voices! Very cool thank you!
Wowsers, thank you for the subtitles! I do not need them but I really like them. Also, push-to-talk is the best way to use microphones...and I am a minor saying that! Though, I am not able to use it much because I use Discord online as I hate downloading too many things. Honestly, I prefer the end of your last video slightly echoey as it makes it seem more...real? But also, that A.I "enhancement" of the end of your last video Really changes how you sound. I actually do know what brown noise is! It also reminded me of GLaDOS!
This Adobe thingy could be great for paranormal videos. Imagine recording an empty scary house then putting it through this highly sophisticated AI to hear the ghost communicate with you
So I'll be honest here, when I bought the keyboard that I use to this day, I thought mx switches was the backlighting of the keyboard... and opted for blue. Discord's Krisp is the only reason I can still use this keyboard, and with a relatively cheap mic I don’t hear much difference in sound quality. (If anything my problem with Krisp is that it outputs my voice only when I am speaking somewhat loudly and clearly, which is annoying occasionally but usually for the best)
DM Dokuro (known for his Terraria Calamity Mod soundtrack, absolute masterpiece) has been getting into other genres of music recently and in one of his most recent songs, "Keep Walking", he's been making use of an AI generated voice from Synthesizer V and it sounds absolutely incredible! Honestly can't tell the difference between a real voice and this. ruclips.net/video/enPUGIzyB98/видео.html I love the mic transition in the beginning by the way haha, the little things like that are just
Tiktok's voice filters are very similar to this, but they sound better since they are not trying to emulate u, but rather be there own sound. they also create voices from noise just like in the music examples
It's interesting that you found a way to "isolate" those generated voices. I remember there's some videos popping up about using Adobe Podcast in famous songs with the same results as the ones you're exposing here. Regarding AI tools for audio, I'm not sure if you're going to experiment with other tools in the market (Clarity VX by Waves, DeRoom by Accentize, most of Acon Digital tools, etc.) and find their pros and cons.
That dolby thing would be great for electronic music to add lyrics by having it try first! It sounded amazing, tbh and might be able to help your own creativity.
I’d love to see you run a single track of one of your songs through the adobe thing or make a song specifically for it. It sounds like an auto tuned voice and it’s kind of cool. Would at least make for a unique song.
I was recommended a video a week or so ago where someone put an aphex twin song through this ai and everyone agreed it just amplified the aphex twin feel of the song.
The robotic voices in the music was pretty sick, ngl What I would like to see (or rather, hear) is the same sample run through the same (or multiple different) voice enhancers, to find out what kind of recursive nonsense that could generate
The density of random shit that Mr. Kliks puts into his video with no warning is high. Mentioning the 3 seconds of silence before he bums himself and then moving on like it was a normal day in the office was glorious.
![[#2024MAMA] G-DRAGON - 무제(Untitled, 2014)+POWER+HOME SWEET HOME+뱅뱅뱅+FANTASTIC BABY | Mnet 241123 방송](http://i.ytimg.com/vi/Ox29z5Nf1Uk/mqdefault.jpg)
Holy smokes... Adobe is clearly marketing their Thingy in the wrong way. What it did to your music was kind of awesome!
I'd take the weird vocoded vocals it put out, clean them up, and merge them back into the original music track.
yeah gives a unique outcome
i was hoping he was gonna do that, missed opportunity. he even got the cleanest vocals he could but still didnt add it back in
I'll be honest when I originally played with the software I figured it was just doing regular old vocal clean up. Using an AI to try and bring the voice wave form into what would he expected of clean audio. I had absolutely no idea they were essentially extrapolating what you were saying and then generating an AI replica voice performance. That has some really interesting potential.
I wonder what would happen if you cleaned up the voices, and re-ran them though, over and over...
@@snginther I imagine - assuming you're not inputting or outputting anything lossy - it'd likely hallucinate more and more audio as artifacts are added from the process. It'd likely also add in a bit of distortion and make it maybe voices a bit more robot-ey sounding. I think the robot voices would probably be more and more distinct and more words would be hallucinated.
Using the AI in music actually sounded somewhat decent. I'd love to see more examples of it later on.
Maybe future of writing new songs. But they would be weird
Me too
Philip AI Album PLZ
my thoughts as well!
He should remix a whole album like this
8:20 - I think this robo voice could even ADD something to the songs. Insane room for creativity here!
Exactly!! It's great!
just use a vocaloid
i recommend checking out bibio - sugarette, sounds exactly like the voice at 8:00 😁
It's GLaDOS
Friday night funkin vibes
Dude the music was actually so fire, using that kind of sample in different genres would actually be such an awesome use of it
Tride! never expected you to be in a kilksphilip video! nice to see you
Alcoholic portal 2 music
Zammuto - It Can Feel So Good fills the same itch those moments gave me
krato
Pogo makes similar music to what you are suggesting.
I’m genuinely surprised how good it sounded with your music. Definitely a load of nonsense but surprisingly badass.
this is one of the best 2kliksphilip videos in a while. makes me think of Phil's upscaling craze a few years ago
Have we left that yet? I think AI and upscaling and everything (2kp) so far has and will be connected.
And just to top it all off, I went ahead and upscaled the image.
that was when he was going insane due to lockdown
agreed
I love videos where Philip is clearly just doing weird experiments purely to satiate his own curiosity, my favourites. Also, interesting to see you in Belfast! Hope you enjoyed yourself
^^^
IT CAN AUTO GENERATE VOCAL FOR BEATS, that's unexpected feature
DaVinci Resolve also has a voice „Enhance“ effect like the adobe thingy. I‘ve found it very helpful in certain situations, you should give it a try!
I didn't know that, I should try it out sometime.
iZotope RX is a (really expensive) VST made for eliminating noise that's extremely powerful. It's used by mixing engineers to restore old records.
It's not for noise eliminating only, it have other cool features, and also, mainly it is not a vst plugin, but a standalone program
It is better to put files in program and after put out in daw, than using vst plugins
@@woodywillchange Oh yeah, forgot it comes standalone too. I've only ever used iZotope's stuff as plugins.
RX9 is a godsent for fixing terrible audio, I was tasked to fix an audio clip for some kind of indie movie that had the worst audio I had ever heard, really loud crickets in the background, 3 people speaking all at different distances from the mic with one being really loud, one being really quiet and one not pronouncing words correctly, it took me hours of fiddling around with the program but it came out almost perfect, in the end there was just one line that sounded a bit weird. This program is truly a miracle.
Honestly, some of iZotope's tools are like black magic. I don't mean to devalue their engineers' and programmers' work by saying so. I can't even begin to imagine what kinds of dsp are happening in the background to make those plugins work.
Atomic Amnesia with the Adobe thingy sounded entirely like a Friday Night Funkin song with the "vocals". I'm fascinated
It is amazing how the AI enhanced recordings sound exactly like I would imagine based on looking at these AI generated images. It's an almost perfect audio recreation of those blurry face images.
Philip you got to make a full remix of one of your songs using this. The results are oddly amazing.
Absolutely
7:30 That sounds like the AI is triyng to reproduce a frog sound rather than brown noise lmao. Also it seems to turn instrumental music into electronic pretty nicely.
Your kitty has those "sad eyes" that my male cat also has :D they look absolutely cute but so sad when trying to get food from my plate. Happy holidays, philip, the master of clicks.
Wow the first music (before Atomic Amnesia) was actually so cool with the AI voice on, if you worked to use a voice or robotic voice to had on top kinda like how the AI did it would sound awesome !
Your old video on voice/audio was probably the best guide out there for that thing and helps me a lot with my videos. Glad to see a newer video on the subject
I absolutely LOVE these sorts of videos, you're a genius Philip!! and if that was your dad in the video - you have one cool fucking dad, and I'm sure you owe a lot of your love for gaming to him. Also the closed captions on this video were at times hilarious xD Keep it up man!! much love!!
You seriously just created a new genre. I promise this will be huge really soon.
Ah, what a nice coincidence. I just spent the last few days working on a looper/delay for stereo audio, and it had the issue of eventually stacking up pass after pass of background noise when it's set to keep looping indefinitely (i.e. with unity feedback). It wasn't loud enough to notice on its own, but it certainly made a difference after some 20+ repetititons.
I knew a noise removing algorithm would be possible to implement, though not knowing how to make one myself, I started digging and found one buried in the tutorials for the programming environment I'm using. Lo and behold, it indeed can sample the background noise during a "silent" moment and boom, the level meters show a blissful 0 dB on the incoming signal. Amazing.
Big +1 to the effect of these processing methods being hard to figure out. Ears fatigue easily, and voices sound way different when played back than when hearing oneself speak. AI/ML-based methods only make it even more confusing and harder to predict. And since they're based on statistical averages and whatnot, they at most will give an approximation of how something would have seemed if it were more like the input samples. Also this all highlights that combined approaches will make the most sense.. the Adobe method seems to clearly need a more basic noise removal algorithm before it passes through its convolution matrices (or whatever the hell it's doing if it's not that).
"and i can then use those 3 seconds of footage BEFORE i bum myself as reference to show the computer the sound of silence. but this silent isnt silent." - 2kliksphilip 2022
And that's kids how a new genre of music was created.
Don't worry, we've all bumed ourselfs before Phillip
damn what a find to apply the brown noise filter and it worked so nicely. this could also be used for isolating vocals from tracks to make acapellas
Hahaha the music segment was hilarious! Merry Christmas everyone
So we definitely need more of these music examples. They sounded amazing
Oh my god that final dehumidifier enhancement was amazing.
Kudos for the HLL footage at 3:16 :) And really sensible cautious approach to de-noising. You don't neet any confirmation of your choices from random people on YT, but you can have it for free haha. On my own, I often listen to the removed noise extract when I de-noise, so I more easily can judge if I'm removing any useful audio information.
The best part about PTT is that beginners don’t press it until they finish half their first word.
This would work great for remastering the live cutscenes in that old bat-hoovering game you made years ago in TGF
Incredible technology. I'd love to see a video on A.I. isolation methods for music such as demucs and spleeter
Working in audio post-production, this video was wildly amusing and inspiring 😄
i can see you are working on most awaited playlist "audio crime 2"... thanks philip :)
God this is so fantastic, imagine how good this will be a few years down the line
If you ever wondered what you tracks would sound like with a vocal track, now you know it would sound awesome! Good job Adobe!
i'd love to hear this with other languages and see what it does with nasalization, back fricatives, glottal stop and secondary articulation
I see you are a man of linguistics as well
@@irgendwer3610 i didn't know anyone here saw this wow
Playing with it now! Thanks for doing something with AI Audio Enhancement. I know I asked about it a long while ago. I really enjoyed this video.
The AI voice songs at the end kinda reminded me of Fill your Pages by Zsolt Marx
Holy shit you've GOTTA start using that Adobe tool in your music! Whatever it did, it had me grooving out in my seat. There's SO much musical potential in there to be explored!
I am intrigued by putting instrumental songs through the AI. it reminded me of so many songs all at once, almost overwhelmingly so.
I loved how when you did the Seance with the Dehumidifier, the edits made it look like some of the usual Ghost Hunting shows 😂
not gonna lie the sounds the dehumidifier made is pretty close to how all voices sound in cruelty squad
This explains why your videos are always so soothing to watch, the audio quality is always really good
Oooh I love the robotic kind of bad talkbox sound that you get from putting it into the AI voice thingie.
The dehumidifier and cat bit is nightmare fuel, but also a creative tool for ghost stories.
Can't wait for the full release of Atomic Amnesia - Adobe thingy version
Their technology has improved, but they've started charging. Would love to see an update video on the state of play with alternatives. All the best love your work.
dehumidifier spits some facts
this is very much more reverb removal just from how it actually sounds versus actual enhancement, extremely compressed too
I got to this super late but this is fascinating! Probably my favorite video of yours so far
Now all I want is hearing this Adobe thingy decipher Churchill's speeches
I love all the videos you’ve been making about AI and life and whatever else you happen to be into. I’ll always look forward to your uploads Philip :-)
Hey philip, cool fact! If you press ctrl + r you can quickly use noise reduction, instead of having to scroll back through all again. I'm a fellow Noise Reduction enthusiast, so I understand the struggle.
This is exactly how I envisioned noise reduction in the future. Voice typing has been around for a long time. So AI should just analyze its voice and voice the recognized words. It remains only to make the option to create profiles so that you can record the voice of a friend and now it will work as a filter, replacing any voice with the voice of your friend.
The ways you used it are so cool! The music conversion segment made me giggle at how crazy it sounds. You can even hear remnatns of the music with the isolated voices! Very cool thank you!
Yet another video full of info I never needed or will need... But it's so cool and I'm glad I know it now.
I was literally thinking about a technology like this earlier today. Philip is tapping into my thoughts.
No he isn't
@@maxmorelikewantedfortaxeva2366 yes he is
He is living in your walls.
@@Bella1899 NO
@@maxmorelikewantedfortaxeva2366 YES.
The clip of baby Philip playing is super cute.
Wowsers, thank you for the subtitles! I do not need them but I really like them.
Also, push-to-talk is the best way to use microphones...and I am a minor saying that! Though, I am not able to use it much because I use Discord online as I hate downloading too many things.
Honestly, I prefer the end of your last video slightly echoey as it makes it seem more...real? But also, that A.I "enhancement" of the end of your last video Really changes how you sound.
I actually do know what brown noise is!
It also reminded me of GLaDOS!
That first AI song is kind of a bop.
Okay the music thing is really cool, and I could see someone using that as a means to make more really interesting music.
I really really want to try making music with this
Finally we have lyrics for your greatest music masterpieces
Huge missed opportunity to not test this thing out with the hyper-compressed Half-Life 1 voice lines to see if it can somehow ‘decompress’ the audio
i absolutely loved that music stuff lol, youre a genius now i can generate garbled ai dream vocals
So glad this video was recommended on the AWP video, I was wondering why you sounded sooo differently in that video
This Adobe thingy could be great for paranormal videos. Imagine recording an empty scary house then putting it through this highly sophisticated AI to hear the ghost communicate with you
lol those old vhs tapes are so wholesome and fun.
7:00 is how I imagine English sounds to non-English speakers.
The music bit producing robotic voices was quite nice.
I need to experiment with the music combination with adobe thingy
The music section reminded me of how Pogo remixes the vocals into their work. Eerily familiar to me at least.
It is pretty good at clarifying audio- I can see some accessibility applications.
So I'll be honest here, when I bought the keyboard that I use to this day, I thought mx switches was the backlighting of the keyboard... and opted for blue. Discord's Krisp is the only reason I can still use this keyboard, and with a relatively cheap mic I don’t hear much difference in sound quality. (If anything my problem with Krisp is that it outputs my voice only when I am speaking somewhat loudly and clearly, which is annoying occasionally but usually for the best)
You're pioneering some really interesting ideas with that software.
DM Dokuro (known for his Terraria Calamity Mod soundtrack, absolute masterpiece) has been getting into other genres of music recently and in one of his most recent songs, "Keep Walking", he's been making use of an AI generated voice from Synthesizer V and it sounds absolutely incredible! Honestly can't tell the difference between a real voice and this.
ruclips.net/video/enPUGIzyB98/видео.html
I love the mic transition in the beginning by the way haha, the little things like that are just
1:40 hey.
Tiktok's voice filters are very similar to this, but they sound better since they are not trying to emulate u, but rather be there own sound. they also create voices from noise just like in the music examples
lmao the music at the end, it legit turns your stuff into daft punk tracks
7:53 You better make a bandcamp album of AI interpolated robo-speech electronic music
Because this stuff is BOPPIN'
It's interesting that you found a way to "isolate" those generated voices. I remember there's some videos popping up about using Adobe Podcast in famous songs with the same results as the ones you're exposing here.
Regarding AI tools for audio, I'm not sure if you're going to experiment with other tools in the market (Clarity VX by Waves, DeRoom by Accentize, most of Acon Digital tools, etc.) and find their pros and cons.
That dolby thing would be great for electronic music to add lyrics by having it try first! It sounded amazing, tbh and might be able to help your own creativity.
I’d love to see you run a single track of one of your songs through the adobe thing or make a song specifically for it. It sounds like an auto tuned voice and it’s kind of cool. Would at least make for a unique song.
I was recommended a video a week or so ago where someone put an aphex twin song through this ai and everyone agreed it just amplified the aphex twin feel of the song.
My favorite Englishman. Good video sir, regards from an anglosphere cousin.
what it did with the music actually sounds so cool! new genre by getting AI to make vocals out of synths, that'd be cool.
Please upload the music with adobe thingy on your channel, or someone else's if needed. I actually want to listen to it.
Oh I so remember the old days when I used to watch you playing Skyrim as odgrub
8:00 that, that was strangely not awful
The robotic voices in the music was pretty sick, ngl
What I would like to see (or rather, hear) is the same sample run through the same (or multiple different) voice enhancers, to find out what kind of recursive nonsense that could generate
The AI voice added to your music sounds pretty good wtf, daft punk definitely came to mind.
So glad you did the AI enhancement anyway! It was quite runnish 😂
The density of random shit that Mr. Kliks puts into his video with no warning is high.
Mentioning the 3 seconds of silence before he bums himself and then moving on like it was a normal day in the office was glorious.
The AI making incomprehensible lyrics out of the melody is really great actually.
that VHS footage looks so fun and wholesome! :D
Merry Kliksmas!
oh my god. the music is so incredible....
merry christmas philip!
7:02 sounds like most modern mumble rappers
I also hold down the PTT button whenever I scream or laugh, it's just natural for some reason.
I knew you would find a way of bumming yourself Phil, thanks for the Super Helpful In-Depth Tutorial!