Powered by

# AICoverGen

Last update: Mar 4, 2024 ‎

# ‎

# Introduction

  • The AICoverGen Colabs are a port of the AICoverGen RVC fork to Google Colab.

  • These are ideal for users who want 'quick & dirty' AI covers, as the whole process of inputting audio, vocal isolation & song mixing is automated.

  • The pitch control is limiting & inconvenient, so if you want a Colab with more control over it, use Ilaria RVC instead.

  • # The two versions of the port are: AICoverGen NO UI and AICoverGen UI.
# ‎

# AICoverGen NO UI

# ‎

# Description

  • This is the version of the port that doesn't have the premade user interface. Credits to Eddy & Raid. Base notebook by Ardha27.

  • It's an upgrade from the original Colab space, bringing bug fixes, improvements, & extra features.

  • The NO UI counterpart is mainly preferred due to being more stable compared with the UI version.
    ‎

# Pros & Cons

✔️ PROS:
❌ CONS:
  • Automatic vocal extraction.
  • Song mixing tool.
  • Has Mangio-Crepe.
  • Automatic model upload.
  • Input audio with YouTube links.
  • Can get the stems files.
  • More stable.
  • Usage limit for free users.
  • Takes 6 mins to load.
  • UI is incovenient.
  • No control over the stem separation.
  • The extraction will always run, you'll waste time if you input clean vocals.
  • Little control of the mixing tools.
  • Limited pitch control.
# ‎

# Setting Up

# 1. Enter the space

  1. Access the Colab space here. Then Sign in to your Google account.

    image
# ‎

# 2. Clone and Install

  • Execute the Clone and Install. This will install RVC.

    image‎
    ‎

  • It will take around 15 minutes.
    It'll be done when you see a check symbol (✔️) on the corner.

    image‎ ‎

Don't worry if red text appears, it's normal.

# ‎

# Inference

# 1. Download model

  1. Go to Model Download Function cell. Paste the model's link in the url bar.

  2. In dir_name name the model. Don't include spaces/special characters.

  3. Then execute the cell.

    image‎

# ‎

# 2. Input the audio

  • Input the vocals/song in the Generate Cover cell.

  • You can go about it with either a YouTube link or a Google Drive file:

    • Copy a YouTube link and paste it in the SONG_INPUT bar.

      image‎
    1. Execute the Mount Drive cell below.
      ‎
      image‎
    # ‎
    1. Click Connect to Google Drive & select your Account.
      ‎
      image
    # ‎

    1. Click the folder symbol ( ) on the left.
      (For mobile users: tap the three lines on the top left & Show file browser)

      Go to drive & you'll find your GD storage.
      Right-click your audio & press Copy path.
      ‎
      image

    # ‎
    1. Paste the path in the SONG_INPUT bar.
      ‎
      image
# ‎

# 3. Select model

  • In RVC_DIRNAME insert the model's name you assigned before.

    image
# ‎

# 4. Modify settings (optional)

  • Below RVC_DIRNAME until Audio Mixing Options you'll find the inference settings.
    Tweak them accordingly for better results if you wish.

    image‎

# ‎

# 5. Modify mix & reverb (optional)

  • In Audio Mixing Options you can modify the values to define the volume of main/backing vocals & instrumental.

    image‎
    ‎

  • In Reverb Control you can add reverb to the output vocals.
    ‎
    image ‎
    ‎

REVERB_SIZE
How "wide" the reverb sounds, like the size of a room.
REVERB_WETNESS
Volume of the reverb itself.
REVERB_DRYNESS
Volume of the vocals.
REVERB_DAMPING

Level of absorption of the reverb's high frequencies:

  • Higher values yield a warmer, natural-sounding reverb.
  • Lower ones sound brighter & more present.
# ‎

# 6. Begin inferring

  • Execute the Generate Cover cell to begin the conversion.

  • It'll be done when the logs say Cover generated at followed by the file path.

    image
# ‎

# 7. Download output

  1. Click the folder symbol ( ) on the left.
    (For mobile users: tap the three lines on the top left & Show file browser)

  2. Open song_output folder & the one inside it. Right-click the first file & press Download.
    The other audios are the stems. Download them too if you wish.

    image‎

# ‎

  • # If the voice glitches out, click here.

# ‎

# AICoverGen UI

# ‎

# Description

  • This is the version that's built with the premade user interface. Port by Hina. Original repo by SociallyIneptWeeb.

  • It's characterized & preferred due to its intuitive UI, ideal for beginners.

  • But on the downside, it may be more buggy, so consider using the NO UI version if issues arise.
    ‎

# Pros & Cons

✔️ PROS:
❌ CONS:
  • Automatic vocal extraction.
  • Song mixing tool.
  • Has Mangio-Crepe.
  • Automatic model upload.
  • Input audio with YouTube links
  • Great UI.
  • Can get the stems files.
  • Quicker audio file upload.
  • Usage limit for free users.
  • Takes 6 mins to load.
  • No control over the stem separation.
  • The extraction will always run, you'll waste time if you input clean vocals.
  • Little control of the mixing tools.
  • Inconvenient pitch control.
  • More unstable.
# ‎

# Setting Up

# 1. Enter the space

  • Access the Colab space here.
  • Then login to your Google account.
# ‎

# 2. Define pitch

  • In the Clone Repository cell, select the pitch change.

    image‎
    ‎

  • If you're going to use models of the same gender as the vocals, use 1.

  • If they're of opposite gender, use 12

  • To change this, you'll have to start the Colab from scratch.

# ‎

# 3. Run cells

  • Then go to Runtime on top & press Run all.

    image
# ‎

# 4. Open UI

  • After a moment the Run WebUI cell will show two links. Open the public URL one.

    image‎

# ‎

# Inference

# 1. Upload model

  • You can upload it using its link or manually inputting its files.

    1. Go to the Download model tab & paste the model link in the bar.

    2. Name it in Name your model. Don't introduce spaces/special characters.

    3. Click Download.

      image

    1. Zip the model into a .ZIP file, if you haven't already.

    2. Go to the Upload model tab, press the upload box & submit the ZIP.

    3. Name it in Model name. Don't introduce spaces/special characters.

    4. Press Upload model.

      image
# ‎

# 2. Select model

  • Return to the Generate tab & press Refresh Models. Then select it in the Voice Models dropdown.

    image
# ‎

# 3. Select audio

  • On its right, input the YouTube link.

    image‎
    ‎

  • Or, press Upload file instead, click Upload & input the audio file.

# ‎

# 4. Modify settings (optional)

# ‎

# 5. Select output format

  • Unfold Audio mixing options & set Output file type as wav, for a better quality output.

    image‎
    ‎

  • Modify the mix too if you wish.

# ‎

# 6. Convert

  • Press Generate. The audio will begin processing.

  • Once it's done, you can hear the results in the AI Cover output box.

  • To download it, press the download symbol ( ) on its right.

    image
# ‎

# 7. Download stems (optional)

  • To download the stems, go to Colab & click the folder symbol ( ) on the left.
    (For mobile users, tap the three lines ( ) on the top left & Show File Explorer)

  • Go to Hina_RVC, song_output, open the new folder, right-click the stem you wish & click Download.

    image‎
    ‎

  • # If the voice glitches out, click here.

# ‎

# You have reached the end.

Report Issues

© Copyright 2024. All rights reserved.