@@Itsappuru You can just turn the voice changer on and off while you're speaking with yourself and there you have it, a conversation with your waifu just as how you've dreamt about all this time, which indeed would give a sort of unique/weird, mixed feelings
That said, I kinda wonder what Gura would sound like if the accent he used was totally different from her original accent, like let's say, a British accent (e.g. RP accent or Cockney accent). Would it sound like a posh version of Gura or would it just sound off?
To those not seeing the chart, Scoll down and when you see "(1) Usage with pre-built binaries" in BOLD Text. At the bottom there is a blue hyperlink text that says "Download from Hugging Face". Just before "(2) Usage after setting up the environment such as Docker or Anaconda" in BOLD text You are Welcome 🙂
So I made some discoveries which could help some people. -Chunk: You should increase this if you experience any distortions or voice lag when gaming with this. This adds more graphics processing time, and if you set it longer it wont rush out bad audio. -Extra: This setting gives bonus CPU usage to help iron out the audio. I found that sometimes the changer would translate an F sound to an S sound, but adding a bit "extra" CPU to it (like the 8k setting or higher) fixes the problem. You don't want to max this out unless the only thing you're doing is using your voice, as maxing it out will use all or nearly all of your CPU. -Noise: I recommend using Sup2 option if you have an Air-Conditioner or other background noise. Sup1 didn't work as well and is probably for a different frequency range, so your millage may vary. When you start real time voice changing, you'll see some info in a box with millisecond timers. The thing to watch is the "res" time. If this time starts going up, from around 300ms it starts rising to a thousand then two thousand etc, this means the computer is unable to get the voice processing out in time, its being pushed back in priority you could say. The fix is to increase the Chunk, this will give it more time to process with the remainder of your resources, and switching it you should see the number start decreasing rapidly. If it doesn't just raise it even higher, and also again keep in mind that if you're doing something that is CPU intensive, you need to keep the Extra setting fairly low (like around 8k). I have a powerful computer (10 core i9 10900kf, with a reference 3080ti), and I found that if im going to play a "serious" game like GTAV or StarCitizen etc, its best to have the Chunk as high as 192, or 256, with the Extra set to 8192. If you're just on discord, or playing some very light game, you can crank the Extra up, and reduce the Chunk to maintain high quality audio but process it considerably faster. Hope this helps someone!! Good luck o/
The not actually instant is an important distinction though, it's hard to hold a conversation with that delay and you end up talking over and interrupting the other person a lot by accident. It also doesn't seem to handle laughter or whispering well at all. Still very cool tech though, for sure, for many uses.
chunk is the buffer size used in processing. The lower, the less latency, but if your cpu/graphics card is not fast enough, then some chunks will be left empty, hence the "choppy" sound. In other words, the lower chunk sizes can work without the sound being choppy, but in order for that to happen you need a monster graphics card.
This is genuinely such an amazingly well put together tutorial video, no side-tracks, no random edits or cuts that confuses a complete newcomer as well as also including disclaimers and showing some potential issues with the program, this earns a sub!
Your videos are really well done, the descriptions and explanations are top tier. This is a game changer for me since I'm a DM at a D&D game that we play on Discord and this will help me make the sessions even more entertaining and will help the players feel like they're interacting with unique characters instead of just me acting different voices. I may also mess around a bit while gaming online. Just a bit.😝
@@miorosa318 you're absolutely right for wanting more male vtubers, but in this moment what's more popular is the "cute anime female" trope, even among the female audience there are male vtubers (like Ryugon Daimonji, but he talks in japaneese) but they're very rare so sorry
@@vonakakkolawell not only that but in Twitter male vtubers are just being berated off being a vtuber with others being all radical feminist towards them
Index is how the AI will match the original accent from the model, a higher index will make your voice have the same accent as the choosen model, but may generate artifacts and potentially take a little longer
With the way this voice changer works it isn't really possible for the original voice to leak through However accidently setting the wrong audio device somewhere definitely will, but still technically not at the fault of this program though
@@CreEpeRneSs125 It's possible I already tried a few of them. The popular ones specially. It's very rare to happen but one time can kill your role play.
Quick update for this video: If this discord is full, check out this video to fix it: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-e8CaftH9olk.html In the download table, I directed you to download the gpucuda version of the software if you have a dedicated graphics card. However, when I went to Hugging Face, I accidentally downloaded the directMLcuda. You should use the gpucuda version (if you have a dedicated graphics card)! Sorry for the confusion!
o hmm well thanks also is it my mic problem that the ai voice is not quite saying the words right its just "mumble mumble" and other things i dont understand it saying. or smthng shud i use bluetooth headphones cuz thats what i use also is it okay if the file form http thing has no bat at the end
This is honestly amazing, the ai is spot on and works perfectly for me. your tutorial is also perfect as well and was very easy to understand. i would love to see a video about how to get this to work on discord since i have had trouble doing the same thing with another voice changer before.
@@vollyentertainmentinc.djvo8265 im not even joking it does work, i use gura voice and its very spot on, this may be because of my high voice but whatever
AI acceleration definitely is way better in the 20 series and above, I have experimented with AI models on my GTX 1070 for my desktop and it is ungodly slow, however when I do the same task on my 3070 TI even on my laptop it's near instantaneous so it probably will work way better.
I love this! I also love how the very day I find out this thing exists is the day you put this video out. It was destiny for me to troll my friends as an anime girl! I am having an odd problem however, there appears to be a large amount of static in my right headphone while the app is open, and it crashes when I try to change settings. If you have any insights as to why that might be the case, do tell! Edit: It was just my audio driver that was out of date by two years lmao
I’m sorry that you are having those issues! I have not experienced those so I’m not sure what would help. My first thought is to make sure you installed the right version for your computer, and that the correct version of PyTorch installed. I would try making an issue report on their GitHub and they can help you further! Here’s the link: github.com/w-okada/voice-changer/issues
Does it only happen with that headset? Try an alternative if you can to figure out if it's just some wierd driver thing, or if somehow the program is causing the problem. If it's just the headset, maybe just uninstall and reinstall your drivers~
@@kl0nos I updated my audio Drivers and it worked, but it's still a but wonky sometimes, I also unplug and replug my headphones and that tends to fix it
@@rulercamelot3687 oh i fixed it too!, turns out i was using the directMl instead of using the onnxcpu one :), instaling the other one pretty much fixed it (i have a nvidea graphics card)
I want this to make a funnny podcast with some characters to have a good time. Also I want to use this to modify a voice in a movie to match the character. This is gonna make it real! thanks!!!
my best guess is that this is using some kind of FFT or equivalent audio processing before the AI processes it, which means that 1) the "CHUNK" setting would be equivalent to a "buffer size" (how many samples in the buffer currently being processed). this will add delay but increase performance for lower-end components. 2) the "EXTRA" setting would be equivalent to the FFT resolution, meaning how many frequencies it can process. this will increase the quality of your voice coming in, and thus the quality of the voice coming out.
1 - Damn, hearing the voice of Ai Hoshino once again give me mixed emotions. I miss her so much 😭😭😭 2 - We need a voice model for Dmitry Yazov IMMEDIATELY
those who want to know, the virtual character, Tsukuyomi - chan, the l0li character with white hair he used in beginning for demonstration. She's just a character for free material that can be used for commercial purposes, such as for this voice changer.
Thank you for providing an informational video, much appretiated. Congrats on 1000 subs, well deserved. I know you will keep growing if you keep putting out great content like this, the algorithm has catched you so now just surf for a while. Wish you all the best
Thank you so much for making this video! I've been trying to figure out how to do a voice changer for months now and I never could figure it out until I saw this video. I got a voice model of Abby Trott made for me and it sounds great! c:
As incredibly but and high quality as this is, i find it horribly terrifying that there's also real voices, an in, youtubers' voices, in this... And I'm pretty sure some of them don't know their voices have been taken.
I think the chunk is what I consider the buffer. The machine will process chunk by chunk. Unless you have a powerful machine, you might want to put higher value there. Or it will sound choppy because your machine can't keep up with the buffer size.
If using a dedicated AMD GPU, use the ONNX(DirectML) version. This also gives benefit to AMD over Nvidia due to ROCm having far more efficient machine learning cores than Nvidia's now-ancient CUDA cores.
I love the fact that that chart IS NO WHERE TO BE FOUND, it also doesn't help that every tutorial is saying "click here" whenever i look away. So now ima just do my own thing and mess around with japanease software until i get a VIRUS.
I'm just on the internet looking for some more like.. voice changers in the sense that my voice echoes and all that but I had to watch this and the cuts between the anime girl voice and your voice is making me laugh so hard.
dude, I was looking for an ai voice changer for other reasons a few months ago I couldn't find anything. I was about to bite the bullet and just do all the audio production myself, but knowing this tool exists is fantastic. I just want a voice changer that makes me sound like a very specific text to speech voice (an oldie) I hope it's possible to do this with this program. I'll probably have to create my own model because no one cares to do what I want to do. If you could change your voice, why the hell would you choose text to speech of all things? I'll tell you, I feel like it.
@@medievalwebman2917 Mine is the rtx 4050 laptop gpu with 6gb vram. But my voice changer have issues such as being choppy, excess pitch, robotic voice leaks. Does the microphone you use matter to make this work properly?
@@Railey_Pg3d mine has only 5 gb vram in gpu but it works just fine. also i made the processing resources all goes to the voice changer also turning off other background processes that interferes with voice changer. maybe you haven't tweak the program yet to use the gpu's core instead of your laptop's core. also i'm using a desktop.
The more ai tools developed the more terrified I become. It's incredibly cool but in the same time also incredibly scary and really makes you think about our future.
99% of the so-called "AI" is just expert systems, i.e. large databases and fast calculations. The only scaring thing is people misusing computers, and it's in been full throttle for a good while now.
@@theAIsearch for some reason when the file it says http but without the bat next to it but it still says batch file. and when i open it, it immediately closes itself
Literally took me more than an hour to figure it out. While it took long enough to download. It wasn't even opening and ngl the time wasn't even worth it. I expected it to work 100% looking at the comments but it annoyed me so much I don't recommend this in 2024.
It's actually a little scary to be honest. The AI voice changer can be used to scam people into the victim thinking it is one of the family members to get their credit card code or something. Not with anime girl voice, of course, that would be random. Thankfully it is mostly used for fun.
I haven't seen anyone talk about this but. A pc tweak that is supposed to help with stuttering in games ended up making the thing run better on my laptop. Acer nitro 5 with an I5 and gtx 1650. I'm able to run the voice changer with 256ms chunk and the lowest extra at 4096 all on f0det of rmvpe. All I did was disable high precision event timer, System timer, and composite bus enumerator in device manager. Before this i couldn't run the program on my desktop at all with nothing running even at 1 second delay it would cut out a lot. so i was gonna delete it but never did. I came across the tweak because I kept getting hitches on fortnite, tried lowering my mouse polling rate to 250 which helped but wasn't enough. Then I came across the device manager tweak mentioned above without realizing that it would allow me to run the voice changer with a 1/4th second delay and fortnite at 120 fps with no issues at all. again on a gtx 1650 i5 LAPTOP. I'm on a desktop now 12400f, 16gb 3200mhz ram, and rtx 3050 oc. and the program is able to run 150ms 4096 extra. and run fortnite 240fps no issues(only with the tweak). But I don't actually use it on my desktop anymore. I use my laptop(with the tweak) to run the program and then use a male to male aux cable from my laptop to my desktop using the line in. So now I can run the program and play any games with no performance hit. I can't stress enough how helpful it is. turn on high precision event timer and system timer it will cut at half second multiple times, turn them off IMMEDIATLEY runs absolutely no issues at a quarter second with 4096 extra, this worked well on rmvpe. it only runs well like this on rmvpe not sure why, any other f0det will cut out no matter what even with the tweaks i tried all of them but only rmvpe works, but imo rmvpe just sounds the best anyways so it's win win. another weird quirk with this is that it runs better with the lowest extra with these tweaks but the more extra the more it will cut out not sure why
while sounding like a cute girl would be amazing, i would pretty much use it so my outter voice sounds closer to the way it sounds for me either that or sound like a Biblically accurate Angel
he didn't mention this in the video but if you're struggling. If you have a nvidia gpu you download the cuda file and if you have an amd gpu you download the DirectML file
After wasting time buying and returning a $300 GoXLR and a $500 RodeCaster Duo for their voice changers I can't believe that this was there all along for free. Thank you, I found your video just in time. The last day to initiate a return for the RodeCaster Duo is today. Just subbed man. Keep up the good work.
I have done everything In this video, but for some reason the voice doesn't voice. I just keep hearing my own voice and the (VOL:) stays at (0.0000) on the top left corner of the model. Also finding issues with the (FO Det.:) options that give you (harvest and crepe) only give me (crepe_tiny and rmvpe_onnx) so for some reason nothing is changing for my voice.. Please help..
Just bc of this I'm gonna Subscribe to you, I see alot of voice changer and this is second by far good and I well used this okada voice-changer with voicemod desktop controller
does anyone know how his "start_http" turns into a shortcut just from moving it to another file? i copied everything he did and mine still says batch instead of shortcut + it won't open the black bar edit : i've made it into the shorcut but it still won't open update : someone in the comments said they clicked start_https first, it'll start downloading, and then they could finally use start_http. i've followed them and the black bar finally opened and is now downloading, i hope this works it worked and i can finally use it now! i don't see the tune settings though i wonder where it is?? okay clearly i am stupid, i found it from clicking which voice to use it won't pick up my voice no matter what i do
If you guys get ratio increase res, you can change chunk to higher or less. Every models had different Chunk to fit fine with your device, so if you had problem with increase res, try it. I use Ryzen 5 3400g with Radeon Vega 11 and it work fine with lag around (res) 400-1200ms.
@@boozonggy people having different voices is not replacing all humans with AI, but I do agree a future of it is not necessarily a "good" thing. The main argument for it is trans people. I'm not gonna get into the politics of it, but if you support trans people then instead of hoping for a future where everyone uses AI to change their voices you should probably hope for a future where people can freely train their voices to sound more female without being shamed for it. Now, no sane person should hope for a future where everyone uses AI for simple bodily functions. That sounds awful. If you life futurism and "evolved societies" and that sorta stuff, you should be rooting for biological enhancements on humans and not brainchips and AI.
random ass video idea for someone to use: play a Mario Kart or Mario Party game with your friends, but everyone uses this voice changer to sound exactly like the character they picked... i have to yet try this myself, but it would be hilarious oh, and good tutorial btw...
I love how pytorch is needed to run this and it's only mentioned briefly. Can't find anything that actually shows me what I'm doing wrong with pytorch and why I can't get it to work or run the ai voice changer. Wish is was covered better in this tutorial........needs to be covered more in this tutorial.
@@theAIsearch Do I need to run something through visual studio first or something? Because when I try to start the start http file, the prompt screen shows for a moment then it exits out of it? I have the same graphics card as you do and I followed instructions precisely so I'm not sure what is going wrong.
@@notaprincess1 If you created a shortcut, did you copy the file or make a link? Make sure you make a link, not a copy. Otherwise, try reinstalling. I've had some issues with it where I needed to reinstall it, and it worked for me after that!
@@theAIsearch I opened it straight from the folder it came in. I didn't make a link or a copy. Uninstalled and reinstalled. Still not working. It opens the prompt, acts like it's going to work, then immediately closes.
I'm so paranoid about downloading ai stuff now because of voice ai, and how it scams you. but i feel like this is a lot more trust worthy and the channel seems dedicated to this sort of stuff. (pls don't find my location and nuke it)
