# Ilaria RVC

Last update: Mar 8, 2024


# Introduction

  • Ilaria RVC is a port of EasyGUI (Mangio) to Google Colab. Made by Ilaria.

  • Works for inferencing only, has a pretty UI, huge speed, & the great tools that Mangio has (such as Mangio-Crepe algorithm).

  • And for this it's considered one of the best alternatives for doing inference through the cloud.

# Pros & Cons

✔️ PROS:
  • Model download through links.
  • Two extra TTS tools.
  • Great UI.
  • Has Mangio-Crepe.
  • Doesn't take too long to load.
  • Very quick.
  • Usage limit for free users.
  • Takes 3 minutes to set up.


# Inference

# See video tutorial


# 1. Enter the space.

  1. First log in to your Google account here.

  2. Then access the Colab space.


# 2. Set up space.

  • Execute the Install Ilaria-RVC cell, by pressing the play button, then Run anyway.

  • Ilaria RVC will begin to set up.

    image‎ ‎ ‎image

  • It'll finish when the cell says Done Cloning Repo.



# 3. Open Gradio.

  1. Now execute the Start cell below.

  2. After a bit it's going to show you two links. Open the gradio.live one in a separate tab.



# 4. Download voice model.

  • image
  1. In the Gradio, go to the Download Voice Models tab.

  2. Paste the link of the model in the Hugging Face Link bar. It must be a public link from either Hugging Face/Google Drive.

  3. In Name of the model, insert a name for it. Don't include spaces/special characters.

  4. Then click Download.


# 5. Select model.

  1. Return to the Inference tab & click the upper pink Refresh button.

  2. Unfold the Choose the model dropdown & select your model.



# 6. Select your audio.

  • Below it, in Drag your audio file and click refresh, click it to load your audio manually. Or just drag the file into it.



# 7. Adjust settings. (optional)

  • If you wish, you can modify the inference settings in Index Settings & Advanced Options for better results.
    Tap them to unfold.



# 8. Convert.

  • Click the upper Convert button & wait for the inference to finish.



# 9. Download output.

  • Once it's done processing, there will be a playable audio in the Final Result! box.

  • Tap the three buttons on the right of the audio and then Download.



# and with an RVC model

  1. First access the Gradio. If you don't know how, follow the first three steps of the previous chapter.

  1. Go to the IlariaTTS section.


    If you want, you can use the other ElevenLabs / Google TTS.

  1. In Voice pick a voice of the gender, language & accent that you wish. Under it, insert the text.

    If you're going to use an RVC model, pick one that sounds similar to it.

  1. Press Speak. The TTS will begin to process.

  1. Once it's done processing, there will be a playable audio in the Final Result! box.
    At the right press the three dots & then Download.


  1. To use an RVC model, simply upload the output to Ilaria RVC & convert it using your model. (Optional)
    If you don't know, learn here.


# Troubleshooting

  • It's normal. Try repeating your action.
  • If it persists, reload the Gradio page.
  • Ensure your audio is selected in the Choose the audio file. If it's not in the list, click Refresh on its right & select it.
  • If it's still persisting, restart the Colab space:
    • Go to the Colab space and press the downward arrow ( ) at the top.
    • Click Disconnect and delete all data.
    • Reload the page and load RVC again.
  • You'll have to create the link yourself. Learn how here.
  • This could be due to a few reasons:

    1. Link is private:
      • If it's from GD, ensure the General access is set as Anyone with the link.
      • If it's from HF, ensure the repo is set as Public. Learn more here.
    2. Invalid HF link:
      • The HF link must contain the word ".zip".
      • If it contains the word blob, replace it for resolve & paste it.
    3. Incorrect files:
      • The model must be zipped in a .ZIP file. Not .RAR or .7ZIP.
      • It must contain its correct .PTH & .INDEX files. Learn more here.
  • This is an anomaly called "artifacting". Learn how to fix it here.
  • Report your issue here.


# You have reached the end.

Report Issues