YOUR Voice in an AI Cover! FREE with the Replay Update

the past few months we've covered lots of different ways to clone your voice and then use that cloned voice to do your own AI cover song switching your voice out for whatever the original singer's voice was in the past we've been using a couple of tools at least to get that accomplished one tool to do the actual voice cloning and model building and another tool to do the conversion one of the tools that I featured here a lot is a free tool called replay and I'm featuring it again now because recently they added the ability to train a model within the program so you don't have to use another instance of RVC or any other model building interface to be able to do this just one program all for free if you're not familiar with replay you can download it here and it is available on various platforms Mac and windows and Linux there are some known issues and common complaints when it comes to installing replay and one of the main ones is that it seems to stall or break as I said in my last video the fact of the matter is it just takes a long time to download and extract all the various models that are required so depending on the speed of your computer and internet connection it could take a while and it does sometimes look like it's hanging up and if you're updating replay like we're about to do here you're going to get faced with the same issue because they make you redownload the models every time you update the program I'm not sure why that is unless they're actually changing the models if they're not changing the models I'd appreciate not having to download them every single time so today we're going to use replay to clone my voice and then switch out a song that we create using udio who has also just recently done an update to extend their first generation of song out to 2 minutes which leads to a lot more coherency and a lot of save time so we'll be using that to create the song in my case I think I'm due for an update on this particular machine so when I run replay yep there it is new update available would you like to download it now I click on yes when you run the installer or the updator you're going to get this message and I get a lot of complaints about this or at least a lot of questions this is something that I've seen throughout the course of my history online when independent developers create software that doesn't go through whatever Microsoft deal they need to go through to get certified or whatever can you tell I know what I'm talking about here's the thing I've never had a problem with these things if I ever see this pop up on my screen I want to make sure that the software that I'm installing has been around a while and we've had time to see complaints if there is actually something malicious going on in this case it's perfectly safe so you just click on more info and you click run anyway and go through this process you'll need to close your running instance of Replay for this to work obviously that part of the installation isn't the part that takes the time it's when you run replay and it starts to download the models AI server update blah blah blah download now you'll see it's fetching 74 different files and of those 74 files many of them are compressed and they have to decompress so they have to extract all those files and then go through the installation process so you can be sitting here for some time and you'll think it's got to be broken it's got to be broken it's not broken just be patient we are now at the decompressing final zip and this might take a while and it does in the meantime let's go go to udio and write our song here on the front page of udio they are telling us about their new 2-minute model and other features like being able to set a seed so that you can duplicate your results over and over with just making changes so if you are a subscriber to udio and haven't looked at this yet or are considering it you might go check this out we won't get into it now because this really isn't an in-depth audio demonstration I really just want to create a song that it's going to be easy for me to switch my voice out for the purposes of this demonstration so I want to create something with a voice that's somewhat similar to mine I'm not going to do a female voice voice and then try and change it out I'll do a male voice one of the things I've never done on any of these demos is to do a wrap because the model that I'm going to be creating is not going to be based on my singing voice it's going to be based on my speaking voice and so maybe wrapping would be a more appropriate approach for this but technically it's really no different than switching out for a singing voice so if you're more interested in switching out a singing voice you can do that too I've never actually prompted for a rap so how about a rap song about people give me such a hard time about the topics I come up with cuz it's always about kittens or getting in trouble with a relative or I don't know what but what am I going to do write a love song a rap song about love of cats we can click real quick on the advanced features and see where all this prompt Strength lyrics strength seed all these things you can choose quality over speed here and this clip start feature becomes particularly helpful when you're doing the extension feature which again we're not doing that today we're going to click on Create and see what kind of rap I get about the love of cats now hopefully udio did some Wizardry to my prompt because that's pretty boring prompt in terms of musical style so normally it fleshes out a little bit I didn't click manual mode so I did give it that creative freedom come on well I can tell you it is not generating its longer songs as fast as sunno did we did a video few days ago about Sun's new update to 4 minutes and it was generating songs like this and I've been waiting oh here we go sorry okay let's check out the first one feline raps City that's not a r that's not a rap not a rap see the type of C make me Chas it no cash flow but the my heart independent no need for a playmate fix my tough days with a soft gaze walks like a queen in every scene got the CLS out but the love King walks like a queen in everything got the CLS out the love okay works for me we don't need to belabor this I'm going to download this just need the audio rooftop Perfections and now let's go check out what's going on with replay we appear to be all done and installed so the first thing I like to do when I get back into replay is make sure that the server is running because sometimes depending on which computer I'm using it doesn't always run automatically and I have to do it manually but as I scroll down here if it says select a song that means you're ready to go but if it says server not running well that means the server is not running and all you have to do is go up under app and click on where it will say start server instead of stop so we need to train our model so click on train a model the model is going to be named Bob narration you'll see why in a moment the training data I have already put in a file in a directory and it is called Bob Doyle narration it is 17 minutes and 23 seconds of me reading in a high quality Studio here's what it sounds like this video was created to give you a high level stand to establish final costs or CEO who or nurse manager not giving up while they're tired all right so you get it I'm doing my announcer stuff all right so that's what it's going to basically sound like and so when we play the model back that's the voice that's going to be wrapping that should be interesting or very boring all I need to do is drag this file right here now you can have multiple files or you can actually record straight from here and if the files have music on them for example you're wanting to train a voice from recorded songs and get the singer's voice under settings you would make sure that this is not clicked pre- stemmed which means that the vocals have not been removed in my case it is Premed because it's just me reading I'm going to leave everything else here it has detected my Nvidia card so it's going to use Cuda or the faster way to do this I'm not recording anything right now but if I were I could just choose whatever microphone input I have and I just click create model that is super simple all the other ways I've showed you involves going into different settings in an RVC interface and changing this that and the other thing to make it happen this is one click now it will take just a bit of time the downside to this method is there's no way to see what's going on that I know of I've looked under here it says view logs but they're not real time they're just files and I don't want to do it that way because we chose Auto Epoch which is basically a segment of the training we don't even know what the total number of epo is going to be so going over here and looking at this limited amount of data that they're giving us that it is Trading Epoch number four doesn't really tell us how much longer we can expect it to be generally speaking these things can go anywhere between 50 epochs to 200 epochs and if I don't like the quality of this then I would redo it again and I would turn this off and choose a higher number but a higher number of epoch doesn't necessarily mean higher quality so it's a little bit of trial and error still so we will come back when it is done let's see what the log says no jobs running that means it should be done now I can go create a new AI cover I'm going to drop that song on here roof top Perfection I am going to go to settings and turn off pre stemmed because I actually want it to do the separation for me leave their relative pitch the same choose Bob narration as the model see if there's anything else I might need to change I don't think so let's click on create song so now it's doing the track separation now it's loading the model to do the conversion it all happens very fast changing voice we're going to have it in seconds here we go see the type of cat made me Chas it z c slow but the firsts when my heart at independent no need for flate fix my t with a s okay I can hear that that's my voice but I don't know if you can or not because the inflection the dialect is still in there it's just the vocal quality walks like a queen in every scene got out but would love Walks Like A queen just to see let's have it convert something that's a little bit more musical maybe let's see if I've got one of these older ones that I did that I can remix this is a 1905 song so I think we're probably okay chose Bob narration separating the track which is Give My Regards to Broadway and let's see if my voice is a little more recognizable with this [Music] one did you ever [Music] see it's hard to tell because it's so high now I'm going to adjust the relative pitch down and octave old New York it is a little low it's out of the range and that's why we get that artifacting let's just isolate the voice and not hide it with music and let's convert a spoken file because that's really Apples to Apples here I'll clear the this out I'm just going to choose an audio file of my friend Rob and this is what the original sounds like he was 5 foot nothing he was maybe 51 and he was about 100 PB and he was the 98b wrestling champ in the state of Wisconsin so let's just drag it over drop it test it it was 5 foot nothing all right now let's convert it oh I should have clicked on pre- stem because it's trying to do a track separation which it doesn't need to do we'll see how it sounds and if it sounds bad I'll do it again without that here we go this is goofball demo I've still got it set for relative pitch way too low let's put it at zero now normally with Rob I do have to make some adjustment cuz his voice is a little bit higher than mine let's just see what it does here without changing the relative pitch at all was 5 foot nothing he was maybe 51 and he was about 100 lb see that's a little bit high so I'm going to take this I'm going to lower it just maybe five and run it again it was 5 foot nothing he was maybe 5'1 and he was about 100 lbs and he was the 98 lb wrestling champ in the state of Wisconsin so this all right so that's totally me it's just still got his dialect smallest weight class he was state champ in wrestling so what he could do from a physics point of view like he he backed down from no one so we've got a good voice model we just need to find a better sample of a song to use let's go back to udio and do a 40s kuner male solo song about love and loss no kittens after these messages we we'll be right back how we doing it really takes a while really come on it ain't fast boys and girls hello man I am growing old waiting for this this is taking entirely too long but what are you going to do ah here we go all right darling hello goodbye why do I see no lyrics here no play button here how about this one every time we I can't help but yearn for you my heart won't Retreat though our loves very sweet for I can't stop this song we always no no no no no no this one got anything yet to going to you dear that is not a 40s kuner male solo on any level all right you're going to force me to use the name 40s Jazz ballad in the dial of Frank Sinatra about love and loss no kittens create and wait forever still working on it time to cue the Mr Bean JF where he's waiting in the field what's the deal I'm going to run out of film prompt edited oh yeah because you didn't give me a 40s kuner so I had to use the name male vocalist traditional pop oh you better give me what I'm looking for or I'm going right to sunno I should have done anyway I could have been done by now four times over trying to give udio the benefit of the doubt cuz they got the new update and this is how okay we got one I think shadows in the sand nope no play button yet well we're having a moment aren't we udio and I it's okay I'll just wait for blur Of Heartache to finish and a second now yeah I don't know maybe shadows in the sand is ready yes I still recall the way you'd hold my hand it lingers on like shadows in the sand I've let you go but then a [Music] sent okay that's close we you have background singers to contend with let's listen to the second one here blur Of Heartache without you life a BL take my heart I cannot feel it take my soul I won't conceal it take my lips I'll never use them bless if I [Music] I that's what we're using all right so I'm downloading that big bad mama I don't know why I said that okay replay here we come Bob narration is still chosen I drag blur heartache in there make sure that Premed is not selected because it needs to separate it out and we will keep the relative pitch at zero and click create song without you life a [Music] blur take my heart I cannot feel it take my soul I won't conceal it yeah that's me oh bless if I can [Music] oh how ins fired right now I showed you a variety of ways to create that model and then use it inside of Replay for a variety of results your model should match what you're doing mine was a narration model would be much better suited for speaking but it works fine with the right song where the voice is in the right range that's a very important part of this whole model thing it doesn't automatically magically work for everything ideally the model should be trained in the same basic vocal range as the original singer of course you have the ability to change the pitch and make those adjustments and a lot of times it works great but sometimes it doesn't and there are so many things that go into that but if you want to get started with AI voice conversion and only use one tool that's free I highly recommend replay if you can just be patient enough to get through the installation or update process if this is the kind of information you like to learn more about well why not subscribe to the channel it's absolutely free of charge and this the kind of stuff we cover almost all the time if you subscribe now I will not look for you I will not pursue you but if you do not I will look for you I will find you you and I [Music]

Channel: Bob Doyle Media

Views: 19,713

Keywords: ai music, create ai covers, udio ai music, replay update, clone your voice, clone ai voices, create ai cover songs with your voice, create ai cover songs with my voice, easy voice conversion, how to make ai cover songs, how to make an ai cover, how to make a song with ai, ai cover songs, replay ai tutorial, ai voice, voice cloning

Id: 3VTPMilwsHo

Length: 17min 7sec (1027 seconds)

Published: Tue Jun 04 2024