Clone Any Voice - RVC Google Colab Free Train & Inference ( Full Tutorial )
Вставка
- Опубліковано 23 лип 2024
- In this video you will learn how to clone any voice for free using RVC (retrieval based voice conversion) on Google Colab for Free.
Voice cloning and ai voice cloning is a very trending topic nowadays. In this tutorial you will learn how to make ai cover and voice cloning using a free software called RVC.
I made this rvc ai tutorial using google colab free, I made the code to make it easy for you to train rvc custom voice models and also backup the model files to google drive to be able to resume the training ant time you want.
✅ Links used in the tutorial 👇 :
- bahaaai.bahaamahmoud.com/clon...
Fly Me To The Moon:
• Fly Me To The Moon (20...
Related Tutorials:
Clone Any Singer's Voice with AI: Ultimate Voice Cloning Tutorial
• Clone Any Singer's Voi...
Separate Vocals and Instruments from Songs! FREE - Demucs Google Colab
• Free AI Tools to Remov...
✅ This channel "Bahaa AI" specializes in publishing tutorial for artificial intelligence tools to create videos, photos and music using a mobile phone.
Subscribe to watch easy, simple and free tutorials about AI tools.
Brother, you are the king of this business. I love you, I send you greetings and love from Turkey.
You're welcome 🤗
Bro❤ , Its work perfectly . love you bro. Please dont remove that ever. Voice clone sites get lot of money to clone voice but this is awesome. plz bro dont delete this ❤
Thanks for your comment 🌷, I will do my best to keep it online and working 👍
I worked on this all day yesterday and half of the night, I even followed the video step by step as I was cloning and none of them worked and I tried it with 3 different crystal clear voices. I have no idea what's going wrong.
Thank you very much! You helped me a lot!
You welcome 🌷
This only works the first time for each account. I do not understand why. From the second time onwards it becomes very slow. Anyway, finally something that actually works. Really amazing. Of all the video tutorials I've tried so far, this is the only one that really works! Never mind, I'll have to spend time creating different Google accounts.
Hi, thanks for your comment 🌷
Yes you may need to use different Gmails on Google colab, as Google colab sets DAILY limits on usage for free tire users.
After one day or so, you can use the email you used before.
Omg. Epic man. Its work. Greetings from Germany
Thanks 🌷
where is the video for training on higher epochs? don't see it on your videos page
This is amazing! Thankyou so much!!! You have a new subscriber :)
Thanks 🌷
Please kindly make a video for how to resume training model bro. Please kindly do this.
did it work for you?
@@onlineguy55 now it work. Thanks alot
when I upload my voice as a wav he says MessageError: RangeError: Maximum call stack size exceeded. can you help how much is maximum to upload ?
Thank you sir, how can we use this trained model in python for own projects like voice assistant like alexa.
Good work ❤
Thanks 🌷
Thanks brother...Any updates on the quality of the vocal production?
Very clear explanation, thank you very much! i do all the steps and i have cloned with succes. I only want to ask if it's possible to raise the sample rate of the cloned voice (actually 16bit 32KHz) because in the higher range the result seem a bit distorted (the SS overall). Anyway thank you agin for the support!
Thanks for you comment 🌷, I will see if it is possible to make the sample rate editable.
Yes bro, So many people want your video face swap, google colab, because last colab was not working 🙂
FaceFusion Google Colab Updated : ua-cam.com/video/lOQnRUWEU_M/v-deo.html
You are the man!!!
Thanks 🌷
Everything worked up until the inference part, it says model name not found even though i wrote (and rewrote) the exact same model name :( not sure what's wrong
Thank you very much bro
🌷
How can I resume the train to improve quality?
amazing.. do you have a version of this realtime? like playing online games using this?
When I run again why it was error?
Hey Bahaa, what does this mean, "/usr/local/lib/python3.10/dist-packages/torch/autograd/graph.py:744: UserWarning: Grad strides do not match bucket view strides. This may indicate grad was not created according to the gradient layout contract, or that the param's strides changed since DDP was constructed. This is not an error, but may impair performance?"
excellent tutorial
Thanks 🌷
can you make video tutorial of how to continue training the model
Hi, thanks for your comment 🌷
A tutorial about resuming the training on higher epochs will be published soon 👍
Which program do you use to connect the instrumental sounds and the cloned singing voice back together btw?
I use CapCut or any video editing software, just import the 2 tracks one above the other and export as mp3
Useful
Right below the epochs, I have another adjust that isn't showing on your video, it says, Saving frequency during the training. (Recommended : 20) what is this?
Save frequency: every how much epochs the code will save the progress of training.
It is recommended to be 20
cảm ơn bạn mình đã làm được .Video rất bổ ích
Thanks 🌷
Hi i love your content,its help me a lot to recreate my dad voice (he is pased away). Im waiting the tutorial video to improving vocal model quality... I hope you make it. Thank you
Thanks for your comment 🌷, sorry for your loss.
I will work on a tutorial about resuming the training ASAP 👍
Thank you so much for your tutorials. This is the first video I found that was simple enough for me to create a voice model. Do you have any resources on how to resume training (add epochs) on a finished model using Google Colab?
same with me bro.
Hi, thanks for your comment 🌷
A tutorial about resuming the training on higher epochs will be published soon 👍
Sir colab is not free after some hours at all ! Please create some video tutorials that We can run colab code to other clouds or vm service
Bro I got an error at last cell
"colorama" not found
Bro make training video also
thanks a lot
You welcome 🌷
This is complicated and cumbersome.
This is audio to audio. Do you have text to audio?
Dude, is it possible to use this program on a computer?
My laptop model is HP 250 G8 i7 36gb rem Graphics card intel irisr xe
I really need this program to identify voice and change my own voice.
Your program is good, but I tried it 3 times and Google Clob started asking for a fee.
can you find a solution for me
I need this so much but I don't have money, google clob is expensive
Is there any way to not save the epoch by the 20s? Everytime I train a model, it constantly saving itself every 20 epoch and it slows down the training. Can it be aadjusted to like 50epoch?
For example
50 - Saving
100 - Saving
150 - Saving
Unlike the 20 wherein every 20 epoch progress, it saves the data which slows down the training progress like
20 - Saving
40 - Saving
60 - Saving
80 - Saving
😊
Saving every 20 seconds is to save your progress during the training.
Google colab may disconnect and reconnect for any reason.
@@BahaaAI yes but can it be adjusted to 50 epoch savings? Can you do an option to let users save the file by every 50? Because saving the file to every 20 epoch really slows down the training time 🥹
I updated the code, you can now set the save frequency you want from 10 to 100
@@BahaaAI thank you so much! 💙✨
This is what I get at the very last step: ValueError: rate must be specified when data is a numpy array or list of audio samples.
Make sure that you are connected with GPU t4 in Google colab.
I have an RTX 4060, can I rely on it instead of the free period from Google Lab?
Hi, I can't guarantee the results as I don't use these codes locally.
it worked the first time I tried it yesterday and wanted to use it again today with the steps you explained to do if we want to use it again another day but this time appeared this error at the last step: ValueError: rate must be specified when data is a numpy array or list of audio samples. what should I do? thank you!
same problem bro, if u found a solution tell me pls
Mine says "connecting to: (gpu) back end google computer engine python 3" and never actualy conects
Try another gmail, Make sure you are connected to Google Colab with GPU t4
Can you make a TTS to RVC model
Is it possible to download the model immediately in .pth format to import into another tool?
Hi, I just updated the code and added this feature. please check it now 👍
@@BahaaAI thanks
Well I don't know how you did it, but it sure doesn't work for me. I watched your video over n over and even watched the video while I was trying to clone a voice. I tried 3 different voices. All 3 voices was crystal clear, I went step by step following your video pausing it between times and none of the 3 worked! I done it the simple way, Male to Male, ran all 3 through the enhancer, ran all 3 at 200 epochs, didn't sound even close to the voice, so I done all 3 again at 500 epochs, something. So my question is, what could be going on? I worked on this all day yesterday and half the night for nothing. So what do you think???
Hi, please explain how it doesn't work, what is the issue?
@@BahaaAI everything runs good but the voices isn't even close to the voices I'm trying to clone. There all male to male voices also.
hello dear,
following error occured when i run last step
FileNotFoundError: [Errno 2] No such file or directory:
Bro. Go to show code and add this import os , and try again
@@EpicFacts-D em sorry but what is os? Operating system?
Merhaba. pth dosyasını indirmek istiyorum ama model adını girdiğimde “Dosya bulunamadı. Model adının doğru olduğundan emin olun” uyarısı çıkıyor. Model adı kısmına ne yazmam gerek. Ses eğitimi sırasında yazdığım model adını yazıyorum olmuyor.
The training process must have been completed to be able to download the .pth file.
Or if you have backed up the model on Google drive, you can load it again and download the .pth file using the code.
@@BahaaAI Size dosyayı mail olarak atayım siz pth dosyasını çıksrıp gönderir misiniz rica etsem
Contact me : splendor.magic@gmail.com
It's not working properly, it's showing error 7-8 out of 10 time's, please make a video about an alternative way to cloning voice
Thank you for the video. If I use large data set say 1 hour length audio, will it help the quality Sir?
Hi, it will not make any difference, the code needs a maximum of 10 minutes of dataset.
The most important thing is that the dataset must be a clean human voice.
@@BahaaAI Thanks for reply....To achieve clean audio for singing, is it more appropriate to train with singing or speaking vocals? What do you suggest Sir?
It is preferable to use both in one audio file. That will improve the model a lot.
@@BahaaAI Thanks Sir. Please publish a video on how to improving the model further. Fine tuning procedure. Thank you.
Thanks u,u helpe me a lot,I have subscribed your channel.I want know how to continue to train RVC models.could you tell me how to do that?😊😊
Hi, thanks for your comment 🌷
A tutorial about resuming the training on higher epochs will be published soon 👍
This is amazing! How many times are free users able to use this each day?
Hi, The daily usage limit for each g-mail depends on the session time (2 hours daily). but you can use several g-mails.
Please make a video on how to convert normal video to anime for free
Hello.. Thanks a lot for the wonderful tutorial Bahaa.... How can we download multiple files for the same target? I mean I may want "n" number of files for some analysis... doing manually one after another is tedious right?
Try this code for batch inference :
colab.research.google.com/drive/1vrpcTgU4gBuTS3LioIhrl2dHIVVMv-20?usp=sharing
@@BahaaAI Thanks so much for such kind help!! Appreciate your effort and time :)
run time auto discenect why this error
If this happens after running the inference step, that because maybe you are using a long audio file. If that so, run the step of downloading the file.
You don't mention if this tutorial is for Windows or Mac OSX
Bro. Video very late. please regularly upload video. Thanks
Sorry for being late. I will do my best 🌷
4:55 waiting for the next video
Is there any way for me to train a model that I have already finished?
Hi, I will work on a video tutorial about resuming the training soon.
can it also produce singing model
Yes 👍
Does this method work with Arabic language as well?
Yes 👍
the upload service takes forever fix that shid use something else
I want to start a UA-cam channel with help of ai so which content will be best? 😅
The best content is what you love to talk about.
@@BahaaAI yes , that would be best 🙌
its only work in english ? or all languageS?
It works for any language 👍
Bro. Your face swap colab was error coming... Please upload latest version.... Requesting as a subscriber 🙏
Ok, I will work on it.
@@BahaaAI Thanks Bro
@@BahaaAI bro still waiting for your, face swap google colab video 🥲
MOdel name not found.
Make sure you wrote the name correctly
@@BahaaAI I've used the exact same name time and again, yet it fails everytime.
I liked it, but I have a question, to use this trained voice to speak text, and not to speak other audio, what do I need to do with this trained voice to make it a model for ex: applio video: exemplo applio: ua-cam.com/video/N0UrOr6LBkw/v-deo.html
Will work on that 👍
My friend said it was free in the video, but after watching it a few times, she started asking for a fee.
Google colab has daily limits on free usage. You can use several G-mails to bypass that.
Bro i want to use Vegeta voiceover for my UA-cam script, can you make a video in this, if possible please please 🥺 🙏 ❤ much needed
Does it work in any language? Like filipino indonesian etc
Yes 👍