RVC Tutorial: How to make your own AI voice model

แชร์
ฝัง
  • เผยแพร่เมื่อ 22 ต.ค. 2024

ความคิดเห็น • 111

  • @averayuki
    @averayuki  11 หลายเดือนก่อน +33

    ❗UPDATE (11/24/2023)❗
    Unfortunately, RVC training notebooks are no longer available in Google Colab (unless you pay for compute units = subscribe to Colab Pro), so if you keep getting disconnected / errors, then there is no fix for this in free use of Colab ❌😞.
    ________________________________________________________________________________
    However, if you still want to create AI voice models and have NVIDIA GPU ✔, you can try local training ^^ (I won't record a new tutorial on how to do it, because I'm ultra busy right now but I'll write a small guide here 🤓)
    I use the original project. Link:
    🔹 github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI
    but honestly, it's difficult to install, so I recommend you this easier .zip version:
    🔹 github.com/Mangio621/Mangio-RVC-Fork/releases/tag/v23.7.0
    ✅How to install:______________________________________________________________
    1. Under "Assets" download the: INSTALL_Mangio-RVC-v23.7.0_INFER_TRAIN.bat
    2. Open it and wait for installation to complete. It will create a .zip package
    3. Extract the installed .zip package
    4. Done
    ✅1-min guide of ''How to use it'':_______________________________________________
    1. To run RVC interface simply click on "go-web.bat"
    2. It will open a WebUI, which will allow you to interact with RVC software in Web Browser (you don't need a internet connection to open/use the software) The steps for training are same as in this video ^^. The entire training process is done using a python script in CMD, so you can check its' progress at any time.
    3. If you are still confused then here is my Step by Step picture Guide:
    🔹pic1-no arrows: www.mediafire.com/file/jvhtk6lz6s2ikjq
    🔹pic2-with arrows: www.mediafire.com/file/0s3a1yaij2rbs35
    (Just please note that this method requires a PC with NVIDIA GPU. For no NVIDIA GPU you can only create song covers but not AI voice models)
    If you have any questions, please feel free to email me ^^
    avera222@gmail. com
    Let the music unite us 🥰🎵🎵
    Have a nice day!
    ✨✨✨

    • @TheChaosix
      @TheChaosix 10 หลายเดือนก่อน

      im having trouble proccesing the data, do you have any idea why it may happen? i put the folder route exactly as you told in the instructions, but it gives me an error from not finding the location.

    • @averayuki
      @averayuki  10 หลายเดือนก่อน

      Your path of the training folder is wrong or contains spaces, there is no other option. You must enter the path correctly. If you don't know how to do it, move your training folder to C drive, name it ''asd'' and paste this path in Step 5 - Your Folder's path:
      "C:\asd"

    • @ohmylawrd211
      @ohmylawrd211 8 หลายเดือนก่อน

      @@averayuki Tried this and it doesn't work either. No spaces or special characters on folder or wav file.

    • @what-hq1gl
      @what-hq1gl 8 หลายเดือนก่อน

      hopefully this works

    • @htgfgrd
      @htgfgrd 8 หลายเดือนก่อน

      @@averayuki having the same problem

  • @oliverispissed
    @oliverispissed ปีที่แล้ว +133

    Never thought I would watch a tutorial using vocaloid voices

    • @ballsndada
      @ballsndada ปีที่แล้ว +3

      i never thought i saw u here 😭

    • @oliverispissed
      @oliverispissed ปีที่แล้ว

      @@ballsndada HII

    • @ballsndada
      @ballsndada ปีที่แล้ว +1

      @@oliverispissed Holey shit ur alive 🤯🤯

    • @oliverispissed
      @oliverispissed ปีที่แล้ว

      @@ballsndada yurr

    • @MYMELODYS_BASMENT
      @MYMELODYS_BASMENT ปีที่แล้ว

      ​@@oliverispissedᥙr ᥱ᥎ᥱrᥡᥕһᥱrᥱ ᥲᥣs᥆ ᥕ᥆ᥕ ᥙr ᥒ᥆𝗍 ძᥱᥲძძ😊

  • @pixelatedrednighthaze
    @pixelatedrednighthaze ปีที่แล้ว +57

    4:44 is literally the perfect example of how most people use Fukase's voice.

  • @pixelatedrednighthaze
    @pixelatedrednighthaze ปีที่แล้ว +19

    4:51 He sounds like he's trying to intimidate people into giving him the right pitch and I love it 👍

  • @skylarlark1987
    @skylarlark1987 ปีที่แล้ว +22

    bruh I LOVE that you used to vocaloids to explain this. Very fun and also informative!

  • @AnemoEra
    @AnemoEra ปีที่แล้ว +25

    I have been in the vocaloid comunity for years. I grew up with it. Hearing "Kaito" laugh for the first time made me a little emotional not going to lie. Is there a way to download premade voice ai's? If so, would you be willing to share your Kaito with me? It's so clean. I understand if not, this is all just new to me.

  • @pixelatedrednighthaze
    @pixelatedrednighthaze ปีที่แล้ว +14

    2:01 Why does Fukase have a French accent 💀

  • @NixxieTrixx
    @NixxieTrixx ปีที่แล้ว +6

    Thank you Cryptonloids for the wonderful tutorial

  • @asleep_or_dead
    @asleep_or_dead 6 หลายเดือนก่อน +1

    I love the fact that you used vocaloid voices

  • @Cartoondude135
    @Cartoondude135 11 หลายเดือนก่อน +2

    PERFECT! Now I can even make AI fanloid voices too!

  • @NoahtheAIPlayer
    @NoahtheAIPlayer ปีที่แล้ว +1

    Yeah um, If you guys want easy mode, and the exact same thing, I recommend Kits AI

  • @user-zg1wm7fj9t
    @user-zg1wm7fj9t ปีที่แล้ว +3

    Now we need Hakaine Meiko AI model

  • @Kryptonn682
    @Kryptonn682 ปีที่แล้ว +3

    ImportError: cannot import name 'Undefined' from 'pydantic.fields' appears after running step 3 what happened and what do I do?

  • @waguri_sama
    @waguri_sama ปีที่แล้ว +3

    ValueError: at least one array is needed to combine google collab 4 Steps RC (reject) - is it rarely supported Is there a tutorial on how to fix the error?

  • @L31600
    @L31600 ปีที่แล้ว

    i just screamed not expecting luka and kaito talking to me

  • @Razzer1.0
    @Razzer1.0 3 หลายเดือนก่อน

    Finally a tutorial for ediots like me who have 0 programing skils=')

  • @BabyCake0619
    @BabyCake0619 ปีที่แล้ว +10

    Very helpful! Thank you! But I have just 1 question... What are the voice samples supposed to be clips of??
    Every individual syllable? Or random words? Or each word of the song you wanna cover? What kind of vocal samples are we making using Miku?

    • @averayuki
      @averayuki  ปีที่แล้ว +5

      Random words are the best 🖤. And it really doesn't matter if it is singing / talking (though I have used both 🎵💬).
      The most important thing is the quality of the samples. If the quality is low, then even if you have 1000+ or more samples, the voice will still be poor. So focus on quality :)

  • @SebanLuster
    @SebanLuster ปีที่แล้ว +1

    i just screamed at a child with his own voice

  • @Cartoondude135
    @Cartoondude135 2 หลายเดือนก่อน

    Hang in there, Sakura, Tomoyo, Eriol, Takashi, and Syaoran! You'll all be speaking English in no time!

  • @rYugoSP
    @rYugoSP ปีที่แล้ว +1

    ya so i have a problem, when I click on one click training, then I get error and my connection with "4 steps rvc" is lost forever, what i can do?

  • @waltervd_o3o
    @waltervd_o3o ปีที่แล้ว +1

    ok so the data set
    does that have to be any particular words ?
    kinda give it as many different words so it can interpet them right ?

  • @Sevennn_
    @Sevennn_ ปีที่แล้ว +4

    I love the test models plz share soon!!!

  • @matimercad0rs
    @matimercad0rs ปีที่แล้ว +1

    😭😭🙏🏻🙏🏻 thank you

  • @riccardofacella9489
    @riccardofacella9489 ปีที่แล้ว +1

    I have a problem. I followed all the steps, I downloaded RVC-GUI from a zip and then, after extracting all the necessary files, I turned on the program, put the voice I want to modify and the audio voice whose dataset I had uploaded to Colab, I put those settings you gave at the end of the video... but nothing. It says "Please select a model and input audio file". What did I do wrong?

  • @finleyrobinson9426
    @finleyrobinson9426 ปีที่แล้ว +1

    i follow all the steps but everytime i output and save the file in the google drive, the folder dosent actually have any data in it. pls help? what am i doing wrong

  • @CyberAtlasX1
    @CyberAtlasX1 ปีที่แล้ว +3

    Says error when I click one click training.

    • @XhocsCole
      @XhocsCole ปีที่แล้ว

      Same issue here, did you ever figure it out?

  • @IsaacEwenFrost
    @IsaacEwenFrost 10 หลายเดือนก่อน

    2:55 When I run this cell, it says the zip doesn't exist in the path, but it does and I have entered it correctly and followed all steps. This is really annoying because I have been trying to make an RVC model for so long and can never do it.

  • @RuneLightLovely
    @RuneLightLovely 3 หลายเดือนก่อน +1

    But I cannot connect the GPU, actually (because of "cannot connect to gpu backend" error, blah blah blah...)...but *HOW*!?

    • @slaaicovers
      @slaaicovers 3 หลายเดือนก่อน

      did you read the ping comment?

  • @EdwinVargas-x4e
    @EdwinVargas-x4e ปีที่แล้ว +3

    "ValueError: need at least one array to concatenate"
    I got the same error and already checked like 10 times im following along and put same as you, can i just send you the dataset $5 paypal lmk lol

    • @emmamorrison9387
      @emmamorrison9387 ปีที่แล้ว

      In the description there's another link that has the exact same steps if the Value Error appears. The same thing happened to me :)

  • @L1NKV
    @L1NKV ปีที่แล้ว +1

    Question: Is this even safe 2 do? Like can't you get into trouble for using an AI instead of the real voicebank?

  • @Joonpathy
    @Joonpathy 11 หลายเดือนก่อน +1

    Those links didn't show to me. It's showing error. Wat do I do?

  • @waguri_sama
    @waguri_sama ปีที่แล้ว

    Thank you

  • @RainyKTV
    @RainyKTV ปีที่แล้ว

    So for that making the data set do ii just make the vocaloid sing anything and use it as samples or does it have to be songs

  • @Krumplebumble
    @Krumplebumble ปีที่แล้ว

    Stuck on step 3 It says: /content/Retrieval-based-Voice-Conversion-WebUI
    python3: can't open file '/content/Retrieval-based-Voice-Conversion-WebUI/infer-web.py': [Errno 2] No such file or directory

  • @RAISHANIC
    @RAISHANIC ปีที่แล้ว

    The End Got Me LAUGH HAHAHAH XD

  • @soviet_ren4412
    @soviet_ren4412 11 หลายเดือนก่อน

    Hey, i keep trying to do it but every time i do i gives me an error saying runtime disconnected and that its not free but im finding nowhere to buy it. pls help

  • @catcelot_p2nd
    @catcelot_p2nd ปีที่แล้ว +1

    do we need words or just letters in the samples

  • @bastyboyomg2.0
    @bastyboyomg2.0 11 หลายเดือนก่อน

    HELP THIS FELT LIKE PIRATING VOCALOID

  • @reisakashiya
    @reisakashiya ปีที่แล้ว +3

    This is possibly the most painful video I have ever watched and I learned nothing.

  • @duckriver7448
    @duckriver7448 ปีที่แล้ว +1

    I got ValueError while the 3rd step is running; It said: "need at least one array to concatenate". What should I do?

    • @averayuki
      @averayuki  ปีที่แล้ว +2

      When you open a GUI in Step3 make sure that ''input training folder path'' is named correctly
      It should be:
      /content/dataset/X
      and X = your dataset name you wrote in Step2 but without ''.zip''
      (and also make sure that all previous cells have been installed correctly, this is very important)

    • @waguri_sama
      @waguri_sama ปีที่แล้ว +1

      @@averayuki ValueError: at least one array is needed to combine google collab 4 Steps RC (reject) - is it rarely supported Is there a tutorial on how to fix the error?

    • @averayuki
      @averayuki  ปีที่แล้ว +2

      You were right. I checked it. 😮 Maybe for some reason this link doesn't work for everyone??? 🤔
      But I tried to fix the error and created a new Colab Notebook 🤓🤓🤓
      Here is a link, use it:
      colab.research.google.com/drive/1a_2Whhwv98aBAw17RKG4mHddV-gerpqH?usp=sharing
      In this new link you may get an red error in Step:1 but don't worry about it. Just click on ''Restart Runtime'' and then reinstall ''Install Dependencies' cell''. Repeat this step twice and it should install properly. The next steps are same as before. There should be no ''Value Error'' anymore :) Please, let me know if it works for you ^^

    • @DeadMemeMan
      @DeadMemeMan ปีที่แล้ว

      @@averayuki this was a few days ago but this link did in fact work (for me at least)
      i was having the same issue and decided to look in the replies to check if anybody was having the same issue
      just giving back feedback :)

    • @fluffynix6076
      @fluffynix6076 ปีที่แล้ว

      @@averayuki there's still the same error for me, i think i did everything right tho? idk. i was able to make one model but the other one just doesn't wanna be made

  • @cresha8779
    @cresha8779 ปีที่แล้ว

    It says that the cell 3 runned succesfully but it doesnt give me any links help

    • @cresha8779
      @cresha8779 ปีที่แล้ว +1

      /content/Retrieval-based-Voice-Conversion-WebUI
      python3: can't open file '/content/Retrieval-based-Voice-Conversion-WebUI/infer-web.py': [Errno 2] No such file or directory
      ???? help

  • @s.ekkii_
    @s.ekkii_ ปีที่แล้ว

    for some reason colab disconnects me not even passing 30 minutes, is there any way to fix that?

  • @KaliopaTsutomu
    @KaliopaTsutomu 10 หลายเดือนก่อน

    what happens if i do above 10 seconds for each file
    and i have 100s of wav for datasets :|
    (imma make a GUMi model :3)

  • @Kushas
    @Kushas ปีที่แล้ว +1

    "ValueError: need at least one array to concatenate" ?????

    • @pytholicism
      @pytholicism ปีที่แล้ว +1

      same here

    • @averayuki
      @averayuki  ปีที่แล้ว

      When you open a GUI in Step3 make sure that ''input training folder path'' is named correctly.
      It should be:
      /content/dataset/X
      and X = your dataset name you wrote in Step2 but without ''.zip''
      (and also make sure that all previous cells have been installed correctly, this is very important)

    • @EdwinVargas-x4e
      @EdwinVargas-x4e ปีที่แล้ว

      @@averayuki I got the same error and already checked like 10 times im following along and put same as you, can i just send you the dataset $5 paypal lmk lol

    • @kuuraisu
      @kuuraisu ปีที่แล้ว

      @@EdwinVargas-x4e i could do this for you

  • @Lairisan
    @Lairisan ปีที่แล้ว +4

    Please, could you add English subtitles? Some parts are hard to understand.

    • @averayuki
      @averayuki  ปีที่แล้ว +1

      Done 😉 Sorry that you couldn't understand. English isn't my first language and I never speak it 😅 but should be ok with the subtitles.

    • @Lairisan
      @Lairisan ปีที่แล้ว +2

      @@averayuki Thank you for adding subtitles. English isn't my first language either and I was having some difficulties understanding the explanation with the robotic voices.

  • @Aishikio03
    @Aishikio03 ปีที่แล้ว +1

    Can you please say where you get your voice samples from? I'm having a hard time making them or finding them.

    • @averayuki
      @averayuki  ปีที่แล้ว +1

      I made them myself using the VOCALOID software 🙂

    • @Aishikio03
      @Aishikio03 ปีที่แล้ว

      @@averayuki Did you have to purchase the full software and voicebanks to get the voices. If so that's problably what I'll do.

    • @madara-mada-mada-go
      @madara-mada-mada-go ปีที่แล้ว +2

      @@Aishikio03 I think I can help? I made an UTAU of Miku once and I just made a project file of all the samples I needed. So, yes, you'll likely need the software. It's easier to get samples without the instrumental tracks or other compression. Just make sure your export quality is good. And for separating samples into chunks, I use Audacity since it's free and what I'm used to, but in another comment OP said he uses Audio Slicer.

  • @Nayaitem
    @Nayaitem 3 หลายเดือนก่อน

    Link is not open. How to open it.

  • @hey-bd1vf
    @hey-bd1vf ปีที่แล้ว

    i don’t understand what should be in my dataset folder on desktop😭😭 i just used a clip of me speaking and put it in a folder and zipped it,,,

    • @LilKobra
      @LilKobra ปีที่แล้ว

      10 seconds takes of raw vocal, more - better

  • @h.s.4508
    @h.s.4508 ปีที่แล้ว

    Thank you for the tutorial. I wish you had used a voice that was more understandable and not put background music which interferes with the poor-quality voice even further.

  • @hitmusicworldwide
    @hitmusicworldwide 7 หลายเดือนก่อน

    I wish i didn't have to do accent error correction in order to listen and understand this. I mean I appreciate it, but all the effects, music and accents etc means there is too much noise reduction and audio processing overhead into comprehensible English for me to parse the content. Humans are LllMs. We need quality data to learn. Imagine if I rode an article and just about every single word was misspelt or written in a patois. That's the experience here.

  • @tthree831
    @tthree831 ปีที่แล้ว

    I can't run rvc gui, can you help me

  • @mathismaltz407
    @mathismaltz407 ปีที่แล้ว

    How do you cut your audio sample into 10-second chunks?

    • @averayuki
      @averayuki  ปีที่แล้ว

      I use ''audio-slicer'', it's the easiest way 😉 ---> download link in video description :)

    • @Dgubz11
      @Dgubz11 11 หลายเดือนก่อน

      I installed ffmpeg from a video, and then i ran some terminal code to automatically split it

  • @322pk
    @322pk ปีที่แล้ว

    FileNotFoundError on a 2 step please help

  • @Happycactustree
    @Happycactustree ปีที่แล้ว

    What program do you use for making the covers?

    • @averayuki
      @averayuki  ปีที่แล้ว +1

      RVC-GUI (to which I converted voicebanks from VOCALOID4)
      So let's say it's RVC & VOCALOID4 😉

  • @ralphmacchios1fan
    @ralphmacchios1fan 11 หลายเดือนก่อน

    Isnt this illegal??😭😭

  • @thundershyofficial-2769
    @thundershyofficial-2769 8 หลายเดือนก่อน +2

    thank you the anime is completely cancer

  • @alessius89x38
    @alessius89x38 11 หลายเดือนก่อน +1

    bro ur too fast chill

  • @Junior-bz8or
    @Junior-bz8or ปีที่แล้ว

    Promo`SM

  • @mihaiexploratoru
    @mihaiexploratoru ปีที่แล้ว

    emo