❓ THIS VIDEO IS OUTDATED! CHECK THIS ONE: ruclips.net/video/0qAQh6hAj3k/видео.html - Watch this video if you're encountering any issue or error or have any question: ruclips.net/video/LPt-Hyv_Ot0/видео.html ⚠DISCLAIMER: I believe all content used falls under the remits of Fair Use, but if any content owners would like to dispute this I will not hesitate to immediately remove said content. It is not my intent to in any way infringe on their content ownership. This video is meant for educational purposes only and doesn't promote or encourage any illegal activities. You can learn how an AI cover is done by watching this video. No models have been shared or created by the video uploader. If you believe your rights have been violated through this video, feel free to take it down. The video uploader didn't make any kind of monetary revenue through any AI cover. Thus, no specific artist identity has been harmed by the video uploader. The uploader doesn't control how these instructions are being used. The video uploader is not in control of any link shared (which includes Google Colab and x-minus.pro). The video uploader is not affiliated in any way to any of the links shared, neither did he make them. The uploader is not a U.S. citizen, therefore a nonresident of the United States. This content is created for teaching and enlightening the audience about the high level that AI has reached, and is covered in a parodical way. Models: rentry.co/nh2arh 0:29, I meant “description” 😢 If you'd like to know more and get more advanced, you can also follow this guide: docs.google.com/document/d/13ebnzmeEBc6uzYCMt-QVFQk-whVrK4zw8k7_Lw3Bv_A ⚠That Colab has more features that may result annoying and slow to some. Also, keep in mind that it will upload your model on a giant spreadsheet available online by default. You can tick "private_model" to avoid that.
@@deadenderexe4496Training a 8/10 minutes dataset with 250 epochs took me around 1 hour, so I’d say maybe 400 with a 15 minutes dataset? I’m not sure though
I used this other programme which was a ballache and didn't even have any appropriate tutorials but this here, this is a blessing, and you've included the link to the colab as well which is amazing! Thank you so much for sharing this with everyone here! ❤
thanks for tutorial here is a bit more detailed info: 1:14 Upload vocals that you want to use instead of the vocals of the song - make sure there is only one person voice 4:22 Upload the song welcome
@@sofiagangoso7586 longer are better, but you need at least 10-50 minutes of talking/singing etc... you can get better results, if the person can sing or read the words from the song you want to do a cover, thanks for comment
Oh man... I have tried countless tutorials, and each time I would quit because of the many errors. I can't believe this actually worked flawlessly, thank you!
I just want to say that I am so grateful for this tutorial! I am not very into programing, but I really wanted to hear some songs sang by my favorite character and it turned out really good! You made my day (probably even whole week), thank you so much for helping people like me! I hope you have a wonderful day~! :3
Thank you so much for this!! I made a voice model based on about 4 minutes of interview audio but the outcome is super clear and close to his real voice!
If you're at the part where you're at "download model" and you keep getting "there's been an error" here's how I fixed it: for me, the problem was that i was using 7zip. By default, it makes archives .7z files instead of .zip files which the program doesn't recognize. It's not good enough to just rename the file from .7z to .zip though. Make another archive and be sure to save the archive as a .zip from the beginning instead of renaming the .7z file.
@@19gianguyen74 I don't fully understand your question. You have to provide a high quality source speech recording for AI training. Longer and better audio quality = better model. After training a voice model, you have to upload a recording of someone else. Then you convert that voice with your custom voice model. That's all
I hated coding and other videos didn't help much, learning from scratch is amazing! You made it so simple and though I hate coding since I needed to decode to have my pc run again after locking myself out of the password.... I'm just so grateful to be able to learn this! Thank you so much! I am afraid of errors but I'll give it some time since this method is actually way safer
First of all, thank you for everything!! This video and the others you posted helped me out for sure. And lastly, idk why so many people are complaining. My AI, which I trained for only 200 epochs and by using the previous version from the last video, is doing just fine! I love making covers with my model! This is the one thing that's brought me joy since my dog died a week ago, so thank you for that! I also really thought I messed up with my model cause the acapellas I used for my dataset weren't necessarily the cleanest and had a little bit of background vocals from another singer, but it still worked so I'm really impressed!!
i just wanted to thank you so much for this tutorial! i'm having so much fun making ais of my own and its given me a past time activity to burn away the rest of my summer break haha. amazingly easy tutorial that can be done by anyone. thank you so much again
The only problem in this tutorial is to explain what audio comes first. At the start do we upload the voice that will be later converted into song, or are we uploading a pitch map for a voice to be later applied on it. Basically which audio does what.
few weeks late but i srsly admire your time and effort into making these videos!! im unsure if i’ll ever get the hang of it but i see your hard work and dedication and want to let you know it’s highly appreciated. i hope you have an amazing day :) ❤
Tutorialnya lengkap.. dak nyesel nonton sampai habis, dijelasin dari A sampai Z fungsi codingnya, gak kayak yang lain.., Tetap semangat bang.. kontennya Jozz banggets!..❤❤😂
This is so interesting and I've loved trying it out, thank you so much! My main question is what you all generally do to figure out the optimal number of samples (files or minutes) vs. the number of epochs without going through the whole process over and over? And how do you prep your audio samples before plugging into the program?
OMG THANK YOU SO MUCH, THANK YOU FOR HELPING ME, I LOVE THE COVER IT SOUNDS SO ACCURATE, IT SOUNDS WAYYY BETTER THAN MUSICFY OR THESE OTHER BAD AI COVER MAKERS, I LOVE THIS, I AM SO HAPPY THANKS FOR MAKING MY DAY!!!! IT SOUNDS SO CLEAN TOO!!!!! THANK YOU I'M SO GRATEFUL, THANK YOU FOR THE TUTORIAL, BEST TUTORIAL EVER!!!!! =DDD
This is great, but can you create a tutorial on how you can continue training a model you've trained once before? I finished training a model at 200 epochs, and would like to continue training it with more audio samples for more epochs -- is this possible?
Thanks for the video. I just wanted to add something. I noticed that many poeple had the issue that the training works the first time but after that it doesn't work because of some errors. I had the problem too and I fixed it by creating another google account and trimmed the audio file of the model I want to train (I shortened it to 5 minutes) after that it worked fine. I hope this helps you to fix the problem
Please make a video on how to save and continue the training (when the time in google colab is over) and how to continue training the finished model if possible
@@rawonionsbefore saving, I need to stop the first function "Open the public URL that will appear below" because the save function doesn't run unless the first one is stopped?
I have a few questions: 1) How much record time is required for the voice to sound good? 2) Does it needs to be different scales too? Or I can just use isolated voice clips that I've saved of a person/character?
1) Any time you want. Obviously, the more you have, the better it will be. 2) Different scales will help the voice sound even better. Like screaming, whispering, etc...
this is great!! i haven't heard the outcome of my model yet (due to a lack of headphones in a quiet place :/) but I have one question left (sorry if you have answered it already): how do you train the model further, like after a day or two, when you have already the downloaded model, without having to start over? thanks in advance :)
for anyone wondering if something is wrong, each epoch seems to take 45-60 seconds for me, so that was a total of just over 3hrs for 250 epochs. and make sure ur at ur computer around the time it finishes bc if ur not it'll time you out and you have to start all over again (there is a 4 hour limit to using it)
I loved your tutorial! It was easier than the others I found on internet. Do you have any video tutorial of how to train voice models or any intentions on doing a tutorial for that?
the website was updated and for some reason when i train the model, when it epochs to 1 or 2 or 3. everytime it says connection errored out and then my colab stops to make it work and i have to do it all over again everytime and it causes the same error a dozen of times and it's getting annoying.
Holy shit. 2 Songs (around 3 minutes of singing) trained in just less than 2 hours. Output wasn't the best but it did better than what I would expect. Unlike SVC, this version of RVC gave an output less than 10 seconds (A 3 minute song .WAV).
Ive said it before but, this is truly the best and clearest Ai tutorial I’ve seen! Ive had some great results covering acoustic based tracks, but dance and remixed tracks are very hit and miss😢 can i ask if there is a better way to cover a dance track with good results?
Every time I try to process the dataset by pressing the button, it keeps giving me an error code telling me that there is no such file directory Edit: I found the problem…when you name the voice, do not use a space. It confuses the file directory. Just name it all one word
Oh man, I have no clue what I did wrong but it's hilarious! It sounds like the singer is in soo much pain ahahaha. He also tries to sing the instruments, gosh it's bad lmao.
Hmm the main Mangio-RVC-Fork I've been fiddling with seems a bit harder to work with but also seems to have a few extra options, one being the Tensor Graph that lets you see when over training is happening, so you can get a more accurate model with less trial and error. Still, certainly looks faster and easier, less hoops to jump through.
@@Rejekts That's good news, not sure but does this model also use G-Drive to back up progress, in case of connection loss, since that tends to help with not losing all your progress at random instances.
Thanks for this! I was wondering if you had any advice for fine-tuning the product and making it sound cleaner, especially with female-to-male voice conversions?
Hello rawonions! Is there a way to continue the training? Like for example, the google colab said something like it exceeded the limit or something. What does the "Google Drive Backup (to resume training)" actually do?
I have already read that you say that to retrain the model, you have to load it from Google Colab, however, there are two options where it appears to load it, but then what's next in Gradio? Can you explain in more detail how to retrain a model? Btw, many thanks for the video
Hi. Nice tutorial. It is very helpful and gives good recommendations for the settings to use for ourselves. However I am a but confused. The first set of files you upload in the video, is this from the voice you want to do the cover, or for the song you want to be covered? It's not very clear. Thank you for your help.
@@napoleonbobassraczke844 I connected this morning and it was like this! Before that worked well I hadn't received an error! Even the Mangio-crepe mode has disappeared how is this possible?
This tutorial was super clear and easy, I followed all of the steps using my IPad but after I download the model it doesn’t show up in the “1. Choose your model” box, how can I fix that?
Except for the last 3-4 hours, there was no one who wrote this problem, but in the last 3-4 hours, everyone started to write this, don't worry, you're okay, google will fix it in 2 days at most
@@elsewiir8848knk buldum taktığını hani step 1 install everything and connect your drive yazıyor ona bir kere basıyorsun ya normalde bir kere bitince bi daha bas bitince geliyor
I write to express my most profound thanks to you for providing immense value that could possibly rival the worth of the legendary Lost City of Gold (but without the inconvenience of having to deal with ancient booby traps, of course).
I keep running into an issue where I get this message, "FileNotFoundError: [Errno 2] No such file or directory: '/content/-EVC-/logs/(NAME)/preprocess.log'" how can I fix it?
Finished everything, downloaded index and model. Put in zip file and put in a gogle drive. Once i go to gradio-download model and paste the gdrive link(there is no restricted access in link so it’s accessible).. and download it says there was an error. Its annoying knowing how long i waited for the process😢. What can i do
I have done this before, when I try to do it again, it doesn't show the same screen as 0:43. there are many different sections one says 'extract features' 'train index' but it doesnt say 'open the public url that will appear below' what do I do:(
warning to anyone else with really bad internet like me, if it seems like something isn't working, just wait longer. it takes a while. it takes hours for me to train a model. it's been about an hour and a half and this one is at epoch 97 out of 275
The link has changed for me, when I go to the train part,the pitch extraction doesn't exist it's something else and crepe isn't there either,what should I do🥲🥲🥲🥲
Except for the last 3-4 hours, there was no one who wrote this problem, but in the last 3-4 hours, everyone started to write this, don't worry, you're okay, google will fix it in 2 days at most
@@ri1970Except for the last 3-4 hours, there was no one who wrote this problem, but in the last 3-4 hours, everyone started to write this, don't worry, you're okay, google will fix it in 2 days at most
@AndersSundstedtExcept for the last 3-4 hours, there was no one who wrote this problem, but in the last 3-4 hours, everyone started to write this, don't worry, you're okay, google will fix it in 2 days at most
[Errno 2] No such file or directory: '/content/RVC' /content The tensorboard extension is already loaded. To reload it, use: %reload_ext tensorboard Reusing TensorBoard on port 8888 (pid 720), started 0:04:00 ago. (Use '!kill 720' to kill it.) python3: can't open file '/content/demo.py': [Errno 2] No such file or directory
I can't believe I left this video in my 'watch later' for too long and now it doesn't work 😐 bloody hell. I hope there is an updated version at some point
If we want to further train a model we finished, do we upload it via the "Download model" tab and then just go to the "train" tab and do the full process over? How do we load an old model to start with? (Sorry I'm new to this!)
If one of the first two steps takes forever, wait until the next day and try in the morning. It means that there's no GPU available at the moment. (train model/index can take a while, but the first two steps shouldn't take longer than two minutes)
@@klaurcschwackerberg1880 yea, ive learned that recently. Went to a discord server to see whats going on and saw the news. The weird thing is that i was able to get it working the day i posted this comment, i just kept retrying and eventually it worked. Ive decided to try to make a another one the next day but obviously it didnt work
![Blox Fruits Dragon Rework Update [Full Stream]](http://i.ytimg.com/vi/EqDAp8udhm0/mqdefault.jpg)
❓ THIS VIDEO IS OUTDATED! CHECK THIS ONE: ruclips.net/video/0qAQh6hAj3k/видео.html
- Watch this video if you're encountering any issue or error or have any question: ruclips.net/video/LPt-Hyv_Ot0/видео.html
⚠DISCLAIMER: I believe all content used falls under the remits of Fair Use, but if any content owners would like to dispute this I will not hesitate to immediately remove said content. It is not my intent to in any way infringe on their content ownership. This video is meant for educational purposes only and doesn't promote or encourage any illegal activities. You can learn how an AI cover is done by watching this video. No models have been shared or created by the video uploader. If you believe your rights have been violated through this video, feel free to take it down. The video uploader didn't make any kind of monetary revenue through any AI cover. Thus, no specific artist identity has been harmed by the video uploader. The uploader doesn't control how these instructions are being used. The video uploader is not in control of any link shared (which includes Google Colab and x-minus.pro). The video uploader is not affiliated in any way to any of the links shared, neither did he make them. The uploader is not a U.S. citizen, therefore a nonresident of the United States. This content is created for teaching and enlightening the audience about the high level that AI has reached, and is covered in a parodical way.
Models: rentry.co/nh2arh
0:29, I meant “description” 😢
If you'd like to know more and get more advanced, you can also follow this guide: docs.google.com/document/d/13ebnzmeEBc6uzYCMt-QVFQk-whVrK4zw8k7_Lw3Bv_A
⚠That Colab has more features that may result annoying and slow to some. Also, keep in mind that it will upload your model on a giant spreadsheet available online by default. You can tick "private_model" to avoid that.
What is possibly the highest epoch I could use without the google colab timing out
@@deadenderexe4496Training a 8/10 minutes dataset with 250 epochs took me around 1 hour, so I’d say maybe 400 with a 15 minutes dataset? I’m not sure though
@@rawonionsI just went with 300 but I think my dataset is around 15
@@rawonionswould you have to pay for colab pro if you wanted to go over that?
@@hauntedwebmonsterYes, but you can just switch account ;)
I watched like 5 other tutorials and this was the most straight-forward and understandable one, thanks!
Is everything free ?
@@AmanKhan-bj7im of course
liar
cursed pfp
its time to make my cat sings never gonna give you up
waiting
HELP-
Lol
I wanna see it lmao
BRUH XD
I used this other programme which was a ballache and didn't even have any appropriate tutorials but this here, this is a blessing, and you've included the link to the colab as well which is amazing! Thank you so much for sharing this with everyone here! ❤
thanks for tutorial
here is a bit more detailed info:
1:14 Upload vocals that you want to use instead of the vocals of the song - make sure there is only one person voice
4:22 Upload the song
welcome
Hello! Do you have any idea how long the audios should be?
And should the person speak or can they shout, sing, laugh, etc to get better results??
@@sofiagangoso7586 longer are better, but you need at least 10-50 minutes of talking/singing etc... you can get better results, if the person can sing or read the words from the song you want to do a cover, thanks for comment
@@Match0Mancan u tell me how can I convert a female voice to male voice?
@@n34r to be honest dont have a clue try to find some tutorial or use some voice changer
Thanks man I was a little confused on 1:14
Oh man... I have tried countless tutorials, and each time I would quit because of the many errors. I can't believe this actually worked flawlessly, thank you!
I just want to say that I am so grateful for this tutorial! I am not very into programing, but I really wanted to hear some songs sang by my favorite character and it turned out really good! You made my day (probably even whole week), thank you so much for helping people like me! I hope you have a wonderful day~! :3
Lucky you. All I got are a bunch of unexplained errors and none of them want to just shut up and rock for me
@@nickess9538 Which ones?
@@rawonionsidk y for me it does not do the index path for me can u help me pls ?
@@rawonions i want to add my voice to song that i want so which record do i add first ? my voice or song?
@@fryldzme too i want make ai cover with my own voice, can you please make a tutorial for it 😢
Your channel is literally the best when it comes to RVC learning. Thank you so much for your dedication
Thank you so much for this!! I made a voice model based on about 4 minutes of interview audio but the outcome is super clear and close to his real voice!
may i ask how many voice audio did you used to make it clear?
@@Kratos.... 8, each about 30 seconds long
@@sharonshxm thanks! just tested it yeah its a clear output
If you're at the part where you're at "download model" and you keep getting "there's been an error" here's how I fixed it:
for me, the problem was that i was using 7zip. By default, it makes archives .7z files instead of .zip files which the program doesn't recognize. It's not good enough to just rename the file from .7z to .zip though. Make another archive and be sure to save the archive as a .zip from the beginning instead of renaming the .7z file.
Thank you! it works!
And if even when archived in zip format it still gives an error "There's been an error."
@@VeronikaMescudi yeah i got the same problem here, anyone knows if it's just a website general error?
if anyone STILL encounters the error, check if you named your model the same as the zip file itself. otherwise it may not work.
@@beenibroh still didnt work lol
For anyone having random filename errors: your voice model name cannot contain any spaces
Thank you!
May I ask what's in the audio file we want to convert!? Help pls i bad at this😢
legend
@@19gianguyen74 I don't fully understand your question.
You have to provide a high quality source speech recording for AI training. Longer and better audio quality = better model.
After training a voice model, you have to upload a recording of someone else. Then you convert that voice with your custom voice model. That's all
i love you
I hated coding and other videos didn't help much, learning from scratch is amazing! You made it so simple and though I hate coding since I needed to decode to have my pc run again after locking myself out of the password.... I'm just so grateful to be able to learn this! Thank you so much! I am afraid of errors but I'll give it some time since this method is actually way safer
coding ?
Well this isn't really coding, but it does involve file management and some other technical aspects.
First of all, thank you for everything!! This video and the others you posted helped me out for sure. And lastly, idk why so many people are complaining. My AI, which I trained for only 200 epochs and by using the previous version from the last video, is doing just fine! I love making covers with my model! This is the one thing that's brought me joy since my dog died a week ago, so thank you for that! I also really thought I messed up with my model cause the acapellas I used for my dataset weren't necessarily the cleanest and had a little bit of background vocals from another singer, but it still worked so I'm really impressed!!
I am sorry for your loss, man. I wish the best to you and your family! I am 100% sure your dog loved you and hopes to meet you in heaven!
i just wanted to thank you so much for this tutorial! i'm having so much fun making ais of my own and its given me a past time activity to burn away the rest of my summer break haha. amazingly easy tutorial that can be done by anyone. thank you so much again
the result is shit
Falcon can, falcon, punch !!!
The only problem in this tutorial is to explain what audio comes first. At the start do we upload the voice that will be later converted into song, or are we uploading a pitch map for a voice to be later applied on it. Basically which audio does what.
Dataset audio = Voice you want to be cloned
Final audio = Voice that the Cloned voice (model) should replace
@@rawonions my audio came out distorted and music sounds very noisy... anyway to fix dat?
@@lajimola7335 same here :(
Yeahh i fucked up the entire process now i have to start over because someone in the comments explained what audio needs to go where
Thank you so incredibly much. I just made my first AI cover. I couldn't have done it without you
few weeks late but i srsly admire your time and effort into making these videos!! im unsure if i’ll ever get the hang of it but i see your hard work and dedication and want to let you know it’s highly appreciated. i hope you have an amazing day :) ❤
I have to say you are a legend with your videos on this topic ! Liked and subscribed !
I got the previous RVC version you used in your other video to work for me yesterday! I think imma stick to that one. But this is much appreciated! ☺️
You can also try this one, it should be easier, better and even faster :)
@@rawonions I guess I could give it a shot! Thank you! Your videos have been most definitely helpful! ☺️❤️✌️
@@keshaslays2229 Thanks!
Tutorialnya lengkap.. dak nyesel nonton sampai habis, dijelasin dari A sampai Z fungsi codingnya, gak kayak yang lain.., Tetap semangat bang.. kontennya Jozz banggets!..❤❤😂
Dude that was simple, concise and efficient. Thanks a bunch!!
Thank you so much! This is one of the most clear and easy to follow tutorials I've seen in a while!!
he or she add 4 records in the first part of the video. But which record is it ? please help me :/
Your voice, if you want to do your model :)
@@rawonions can you use .wav files??
This is honestly so much more easier! Thank you!
In the current google colab, which cell do I first run? It doesn’t have the “Open the public URL” Cell anymore.
Same!
Thank you so much! i watched like 20 other tutorials but they didnt work but your Tutorial was the best and it worked! :)
Looks like they updated the Collab and... its not good. Loved this tutorial so much. The training is no longer usable with this tutorial
This is so interesting and I've loved trying it out, thank you so much! My main question is what you all generally do to figure out the optimal number of samples (files or minutes) vs. the number of epochs without going through the whole process over and over? And how do you prep your audio samples before plugging into the program?
There's a graph on Tensorboard that shows you the point of overtraining.
You can find an explanation on the guide in the pinned comment.
OMG THANK YOU SO MUCH, THANK YOU FOR HELPING ME, I LOVE THE COVER IT SOUNDS SO ACCURATE, IT SOUNDS WAYYY BETTER THAN MUSICFY OR THESE OTHER BAD AI COVER MAKERS, I LOVE THIS, I AM SO HAPPY THANKS FOR MAKING MY DAY!!!! IT SOUNDS SO CLEAN TOO!!!!! THANK YOU I'M SO GRATEFUL, THANK YOU FOR THE TUTORIAL, BEST TUTORIAL EVER!!!!! =DDD
This is great, but can you create a tutorial on how you can continue training a model you've trained once before? I finished training a model at 200 epochs, and would like to continue training it with more audio samples for more epochs -- is this possible?
Yes!
Save it through the Colab page, and when you need to train it again, run the GUI then Load the mode from the Colab page.
@@rawonions Can you give a more detailed explanation? I don't really understand
@@Criscuitt docs.google.com/document/d/13ebnzmeEBc6uzYCMt-QVFQk-whVrK4zw8k7_Lw3Bv_A/edit#heading=h.qcwjng5plv85
OMG I LOVE YOU SOOOO MUCH IT WORKED... it ain't simple but If you follow every step of this wondeful tutorial, YOU GET THERE!!!!
Thanks for the video. I just wanted to add something. I noticed that many poeple had the issue that the training works the first time but after that it doesn't work because of some errors. I had the problem too and I fixed it by creating another google account and trimmed the audio file of the model I want to train (I shortened it to 5 minutes) after that it worked fine. I hope this helps you to fix the problem
RVC is now banned from Collab ! since a few days !
This is REALLY helpful. Thanks a lot! It worked perfectly fine at the second attempt, first one was laggy. I'm quite impressed.. I love ai ^_^
Please make a video on how to save and continue the training (when the time in google colab is over) and how to continue training the finished model if possible
100% this, I've asked this same question all over the place, and no one will answer me.
@@fitzcannon maybe its not possible, ive asked too, perhaps make from scratch but this time train to 500?
It is possible. Choose "Save" in the Colab page, and when you're ready to continue training, choose "Load" and add more epochs.
I've answered many times, choose "Save" in the Colab page, and when you're ready to continue training, choose "Load" and add more epochs.
@@rawonionsbefore saving, I need to stop the first function "Open the public URL that will appear below" because the save function doesn't run unless the first one is stopped?
I have a few questions:
1) How much record time is required for the voice to sound good?
2) Does it needs to be different scales too? Or I can just use isolated voice clips that I've saved of a person/character?
1) Usually around 5 minutes?
2) It doesnt matter what scales!!! you could have someone talking in complete monotone and itll still work.
1) Any time you want. Obviously, the more you have, the better it will be.
2) Different scales will help the voice sound even better. Like screaming, whispering, etc...
@@rawonions Amazing!
@@rawonions my voice is not good and convert it to Ai it's like mine, how to improve it
Thank you a lot for this tutorial, this is the only google colab link that worked for me.
I was wondering if you could upload a video of how to save a model and retrain later. Thank you. Great video for sure!
this is great!! i haven't heard the outcome of my model yet (due to a lack of headphones in a quiet place :/) but I have one question left (sorry if you have answered it already): how do you train the model further, like after a day or two, when you have already the downloaded model, without having to start over? thanks in advance :)
I'm wondering the same thing
love the simplicity of this! thank you so much :)
for anyone wondering if something is wrong, each epoch seems to take 45-60 seconds for me, so that was a total of just over 3hrs for 250 epochs. and make sure ur at ur computer around the time it finishes bc if ur not it'll time you out and you have to start all over again (there is a 4 hour limit to using it)
sks alot of dks when it does that
@@RedAvery1 huh? sorry i dont know much about this
@@RedAvery1😂
I was clueless until this video helped me, so thank you sir or lady!!!
what's happening? I can't train models anymore, the connection stops every 10 minutes. :(
You really are the best Ai RUclipsr
I loved your tutorial! It was easier than the others I found on internet. Do you have any video tutorial of how to train voice models or any intentions on doing a tutorial for that?
? This one
the website was updated and for some reason when i train the model, when it epochs to 1 or 2 or 3. everytime it says connection errored out and then my colab stops to make it work and i have to do it all over again everytime and it causes the same error a dozen of times and it's getting annoying.
google banned RVC
I'm back to thank you. I am so so so satisfied with what I made using this platform.
The only pity is the limited usage time of GPU lol
Holy shit. 2 Songs (around 3 minutes of singing) trained in just less than 2 hours. Output wasn't the best but it did better than what I would expect.
Unlike SVC, this version of RVC gave an output less than 10 seconds (A 3 minute song .WAV).
Yeah! A new inference method has been added, so it is now even faster than mangio-crepe :)
Ive said it before but, this is truly the best and clearest Ai tutorial I’ve seen! Ive had some great results covering acoustic based tracks, but dance and remixed tracks are very hit and miss😢 can i ask if there is a better way to cover a dance track with good results?
when are you making a new video with the updated version??
Every time I try to process the dataset by pressing the button, it keeps giving me an error code telling me that there is no such file directory
Edit: I found the problem…when you name the voice, do not use a space. It confuses the file directory. Just name it all one word
ty
Thx
MY GOD TY SO MUCH IDK WHY IT WOULDNT WORK!!!!!
Oh man, I have no clue what I did wrong but it's hilarious! It sounds like the singer is in soo much pain ahahaha. He also tries to sing the instruments, gosh it's bad lmao.
I did the same thing with mine and it sounds like he being tortured 💀
@@TwistPods It's so fuckin bad ahaha I think I got too much reverb in the audio dataset, no clue.
you need to isolate the vocals
If he tries to sing the instruments, it means you didn't split the Acapella part from the Instrumental part as told in the video.
@@rawonions Yeah i thought that was for the samples 💀
dude, thank you so much
this is the easiest and most logical way
Hmm the main Mangio-RVC-Fork I've been fiddling with seems a bit harder to work with but also seems to have a few extra options, one being the Tensor Graph that lets you see when over training is happening, so you can get a more accurate model with less trial and error. Still, certainly looks faster and easier, less hoops to jump through.
If you want to use the tensor graph just check the tensorboard option! I added it today.
@@Rejekts That's good news, not sure but does this model also use G-Drive to back up progress, in case of connection loss, since that tends to help with not losing all your progress at random instances.
@@VB-92 It does not. But it's kind of better this way in my opinion.
@@rawonions i want to add my voice to song that i want so which record do i add first ? my voice or song?
Great video, I just have a question, how could I keep training an already trained voice model?
Uhh I don't know omuri fan
@@Josh72769 Uhh why do you reply this roadblocks fan
@@heartrockingcity same as you, we don't know.
Thanks for this!
I was wondering if you had any advice for fine-tuning the product and making it sound cleaner, especially with female-to-male voice conversions?
If it's a popular song, there are probably some covers of the opposite gender, maybe even acapella
A lot of people also adjust the pitches from what I've seen!
i get the whole time error connection timed out any fix?
Hello rawonions! Is there a way to continue the training? Like for example, the google colab said something like it exceeded the limit or something. What does the "Google Drive Backup (to resume training)" actually do?
easiest tutorial ive ever seen thank you
I have already read that you say that to retrain the model, you have to load it from Google Colab, however, there are two options where it appears to load it, but then what's next in Gradio? Can you explain in more detail how to retrain a model? Btw, many thanks for the video
easy gui is not working anymore!! any alternatives, that let you select batch size etc???
Try on hugging face!
There are some RVC repositories for ai audio.
@@KazutoBlogch i only found one and it doesnt let me train. Inference only!
Once i've trained a Model can i improve the existent one with other amount of datas?
Hi. Nice tutorial. It is very helpful and gives good recommendations for the settings to use for ourselves. However I am a but confused. The first set of files you upload in the video, is this from the voice you want to do the cover, or for the song you want to be covered? It's not very clear. Thank you for your help.
Its for the voice you want to do the cover, you're uploading the samples to train it
Very grateful for your simple and detailed tutorial!
Why are the menus red? And that I can't use my models anymore? It worked very well before
the red color symbolizes the user's face after receiving an error for the 50th time in a row
@@napoleonbobassraczke844 I connected this morning and it was like this! Before that worked well I hadn't received an error! Even the Mangio-crepe mode has disappeared how is this possible?
GOOGLE BANNED RVC
This tutorial was super clear and easy, I followed all of the steps using my IPad but after I download the model it doesn’t show up in the “1. Choose your model” box, how can I fix that?
I'm having the same exact issue
I need this for a project I have with my friends. Thanks so much!
The results are amazing, truly!! Can I train an already existing model more? And how? In case I want to improve it.
Help me please
When i try to resume the training and i uplod my model name it say
File not found
i'm also having this error. i think it has to do with gradio itself, but i don't know how to fix it.
your model name should not contain any space
Now when I click the link, it shows me this other software and there is no "open the public url that will appear bellow" thing at all
Mine keeps saying it cant find the Hubert base, i went to go download it from the link it sai to downlaod from, what do i do next?
Except for the last 3-4 hours, there was no one who wrote this problem, but in the last 3-4 hours, everyone started to write this, don't worry, you're okay, google will fix it in 2 days at most
@@fatihterzi2425 ne zamana düzelir bu durum aynı hatayı alıyorum
@@elsewiir8848knk buldum taktığını hani step 1 install everything and connect your drive yazıyor ona bir kere basıyorsun ya normalde bir kere bitince bi daha bas bitince geliyor
I write to express my most profound thanks to you for providing immense value that could possibly rival the worth of the legendary Lost City of Gold (but without the inconvenience of having to deal with ancient booby traps, of course).
I keep running into an issue where I get this message, "FileNotFoundError: [Errno 2] No such file or directory: '/content/-EVC-/logs/(NAME)/preprocess.log'" how can I fix it?
When are you getting it?
@@rawonions when I press “process dataset”
Finished everything, downloaded index and model. Put in zip file and put in a gogle drive. Once i go to gradio-download model and paste the gdrive link(there is no restricted access in link so it’s accessible).. and download it says there was an error. Its annoying knowing how long i waited for the process😢. What can i do
Are you on MAC?
@@rawonions yes
@@triplea9911 Read the pinned comment.
I followed all the steps until I realized that the 'Download Model' button doesn't appear.
I have done this before, when I try to do it again, it doesn't show the same screen as 0:43. there are many different sections one says 'extract features' 'train index' but it doesnt say 'open the public url that will appear below' what do I do:(
warning to anyone else with really bad internet like me, if it seems like something isn't working, just wait longer. it takes a while. it takes hours for me to train a model. it's been about an hour and a half and this one is at epoch 97 out of 275
Welp, I know what im about to spend hours doing and im so excited to
hey at 2:22 It says that "extracting is shut down because hubert_base.pt doesn't exist, what do I have to do ?
Same
same
SAMMEEE
any fix?
Same, and we are the first ones having it. I guess something is wrong in the server
The link has changed for me, when I go to the train part,the pitch extraction doesn't exist it's something else and crepe isn't there either,what should I do🥲🥲🥲🥲
the colab I got when I cklicked on your link in the description looks different then yours in the video
still training for over 4000 seconds, zenyatta bout to make some heat wit this
so when i try to convert my audio with my model i get an error,though ive done this before and never got an error
I also get the error phrase " Extracting is shut down because hubert_base.pt does not exist". It wasn't like that before ㅠㅠㅠ
Except for the last 3-4 hours, there was no one who wrote this problem, but in the last 3-4 hours, everyone started to write this, don't worry, you're okay, google will fix it in 2 days at most
@@ri1970Except for the last 3-4 hours, there was no one who wrote this problem, but in the last 3-4 hours, everyone started to write this, don't worry, you're okay, google will fix it in 2 days at most
@AndersSundstedtExcept for the last 3-4 hours, there was no one who wrote this problem, but in the last 3-4 hours, everyone started to write this, don't worry, you're okay, google will fix it in 2 days at most
Can you do an updated tutorial? Because train index and train model switched places.
The 9/12 update has broken the GUI it seems. I get constant connection errors when training the model, and the Mangio Crepe option is no longer these.
BEST VIDEO IVE SEEN FOR THIS
why does the the google link for it no longer work? loads something called tensor board and lets me do nothing now
[Errno 2] No such file or directory: '/content/RVC'
/content
The tensorboard extension is already loaded. To reload it, use:
%reload_ext tensorboard
Reusing TensorBoard on port 8888 (pid 720), started 0:04:00 ago. (Use '!kill 720' to kill it.)
python3: can't open file '/content/demo.py': [Errno 2] No such file or directory
This video was really helpful! Thanks!
You should make a video on how to resume training.
You need to make a new tutorial, the Google Colab is now diferent, HELP ME!!!
Calm down! Google Colab is fine.
everytime I open the google collab link there is always "403. That’s an error".Should I wait til for several hours before I open the link again?
on discord I read an information that said google blocks collab for RVC.
RVC is now banned from Collab ! since a few days !
As of 9/15/23, I believe it is working again as it lets me train my vocals without interruptions.
but mine still says "9/12" when i click the link provided
@@j-un8yj Yeah, but for me it works. Try and see.
it doesn't stop but it keeps saying error when i try to convert vocals
YOIIII, awsome video i just have some questions
- does it need to be an mp3 file?
- should my files be 10 second clips?
I can't believe I left this video in my 'watch later' for too long and now it doesn't work 😐 bloody hell. I hope there is an updated version at some point
I will record it today :)
@@rawonions thank you so much! 😊
@@rawonions today💀💀💀💀💀💀💀💀💀💀💀💀💀💀💀💀💀💀
Thank you so much, man. You're a true lifesaver
If we want to further train a model we finished, do we upload it via the "Download model" tab and then just go to the "train" tab and do the full process over? How do we load an old model to start with? (Sorry I'm new to this!)
for those who cant find the "open url" button just play the last code in the site, it will work aswell.
MEANS
If one of the first two steps takes forever, wait until the next day and try in the morning. It means that there's no GPU available at the moment.
(train model/index can take a while, but the first two steps shouldn't take longer than two minutes)
what about the last step, conversion? i ran out of GPU at this stage and selected mangio crepe and it's been loading forever
@@cynikov try again on another day
Im getting red error signs when i hit convert at 4:35
Same
we must wait now for the fix
the runtime keeps disconnecting for me whenever i get to the "train model"
Train for 1 epoch. Save to google drive. Load from google drive. Train for 5. Save. Load. Train for 10....
RVC is now banned from Collab ! since a few days !
@@klaurcschwackerberg1880 yea, ive learned that recently. Went to a discord server to see whats going on and saw the news.
The weird thing is that i was able to get it working the day i posted this comment, i just kept retrying and eventually it worked.
Ive decided to try to make a another one the next day but obviously it didnt work
There's an recent error that said
ModuleNotFoundError: No module named 'gtts'
Can you fix it, raw onions?
same here :(
RVC is now banned from Collab ! since a few days !
TYSM😭😭 I suffered a lot trying
Really good toutorial, can I train the same model again in order to improve quality?
Yes! Follow this video if you wanna know how: ruclips.net/video/LPt-Hyv_Ot0/видео.html
@@rawonions thanks :)