Applio

Last update: April 13, 2026

Introduction

Applio is a Fork of Original/Mainline RVC, developed by the IA Hispano team.
It's liked for its great UI, performance improvements and lots of extra features, such as TTS (with RVC models too), plugins, automatic model upload, customizable theme & more.
Because of its user-friendly experience & active development, it's considered to be one of the best forks.
Applio has it's own Applio Docs, which may have more info about the tool.

Is this Program Safe?

RVC Models are PyTorch Models, a Python library used for AI. PyTorch uses serialization via Pythons' Pickle Module, converting the model to a file. Since pickle can execute arbitrary code when loading a model, it could be theoretically used for malware, but this software has a built-in feature to prevent code execution along the model. Also, HuggingFace has a Security Scanner which scans for any unsafe pickle exploits and uses also ClamAV for scanning dangerous files.

Pros & Cons

The pros & cons are subjective to your necessities.

✔️ PROS

❌ CONS

Very complete
Has an active development
Currently stable
Faster interface
TTS features
Automatic model upload
User-friendly UI
TensorBoard included
Extra features: (plugins, model fusion, etc)

Doesn't support: Intel GPUs, ARM64 CPUs, NPUs.

System & Hardware Requirements

Check if you meet the requirements to run it locally.

Download & Installation

Before Downloading:

Don't put it in a folder with privileged access.
Don't run the run-install.bat as an administrator.
Make sure the path does not contain any spaces or special characters.
Deactivate your antivirus and firewall to avoid missing dependencies.

Nvidia on Windows (Precompiled)

RTX 5000 Series Users require version 3.3.0 or newer.

The easiest way to download Applio is by going to Applio's Hugging Face repo, and clicking the [ download ] button on the right-hand side.

Unzip the folder. This may take a few minutes.

Open Applio's folder & execute run-applio.bat.

‎
‎

A console tab will appear, and after a moment your default browser, with Applio ready to use.
‎

Don't close the console until you're done using it, or it will stop working.

Linux & macOS

The easiest way to download Applio is by going to Applio's Hugging Face repo, and clicking the [ download ] button on the right-hand side.

Unzip the folder. This may take a few minutes.

Make sure you have Python 3.10.12 or 3.11.x installed. You can check your version by running python --version.
Open a terminal in the Applio directory you just extracted.
Run the commands corresponding to your Linux distribution:

apt install python3.10-venv -y
python -m venv .venv
find ".venv" -type f -exec sed -i -e 's/\r$//' -e "s|/home/runner/work/Applio/Applio|$(pwd)|g" -e "s|/.venv/bin/python|/.venv/bin/$(basename $(which python))|g" {} +

sudo pacman -S python-virtualenv --noconfirm
python -m venv .venv
find ".venv" -type f -exec sed -i -e 's/\r$//' -e "s|/home/runner/work/Applio/Applio|$(pwd)|g" -e "s|/.venv/bin/python|/.venv/bin/$(basename $(which python))|g" {} +

sudo dnf install python3-virtualenv -y
python -m venv .venv
find ".venv" -type f -exec sed -i -e 's/\r$//' -e "s|/home/runner/work/Applio/Applio|$(pwd)|g" -e "s|/.venv/bin/python|/.venv/bin/$(basename $(which python))|g" {} +

Run Applio

In the terminal, run the following commands to make the script executable and launch the application:

chmod +x run-applio.sh
./run-applio.sh

‎

A console tab will appear, and after a moment your default browser, with Applio ready to use.
‎

Don't close the console until you're done using it, or it will stop working.

AMD on Windows (Precompiled Fix)

"AMD Compatibility Note"

AMD drivers update frequently and often break Zluda (the CUDA bridge).

If you are on very new drivers (Adrenalin 25.5.1+), you must use the latest available HIP SDK.
If your setup breaks after a driver update, check the AI HUB Discord or the Applio Assets repo for updated patch scripts.

For Adrenalin 25.5.1 driver or newer

Download a compiled version of Applio v3.5.0 or newer from the Hugging Face repo, and unzip it.
Download and install the latest stable HIP SDK from the AMD ROCm Hub.
- Important: Install components but exclude/deselect the video driver at the bottom of the installer list.
Add the bin folder of your installed HIP SDK to your System Environment Variables (Path): C:\Program Files\AMD\ROCm\<YOUR_VERSION>\bin

Open a command line (CMD) inside the Applio folder and run:

env\python -m pip uninstall torch torchvision torchaudio
env\python -m pip install torch torchvision torchaudio --upgrade --index-url https://download.pytorch.org/whl/cu118

Download the patch file corresponding to your installed HIP SDK version from the Applio Assets repo and run-applio-amd.bat.

Edit the file located at rvc/lib/zluda.py. Replace the content with the following:

import torch

if torch.cuda.is_available() and torch.cuda.get_device_name().endswith("[ZLUDA]"):
    # disabling unsupported cudnn
    torch.backends.cudnn.enabled = False
    torch.backends.cuda.enable_flash_sdp(False)
    torch.backends.cuda.enable_math_sdp(True)
    torch.backends.cuda.enable_mem_efficient_sdp(False)

Run your downloaded patch script, then run run-applio-amd.bat.

Download and install the VC++ Runtime.

First, check the official System Requirements on the AMD ROCm™ documentation site.

Install the HIP SDK version recommended in your documentation from the AMD ROCm Hub.

Install v5.7.1 HIP SDK from the AMD ROCm Hub.
Download the correct archive for your GPU:
- For 6700, 6700XT, 6750XT, download the gfx1031 archive.
- For 6600, 6600XT, 6650XT, download the gfx1032 archive.
Navigate to C:\Program Files\AMD\ROCm\5.7\bin\rocblas\ and rename the library folder to library.old.
Create a new, empty folder named library in its place.
Unzip the content of the archive you downloaded into this new library folder.

Find your GPU's gfxNNNN value (search "techpowerup your_gpu_name" and check "Shader ISA").
Follow the steps for your corresponding gfx value:
1. Install v5.7.1 HIP SDK.
2. Download this archive.
3. Navigate to C:\Program Files\AMD\ROCm\5.7\bin\rocblas\ and rename the library folder to library.old.
4. Unzip the content of the archive directly into the C:\Program Files\AMD\ROCm\5.7\bin\rocblas\ folder.
- Visit this repository and follow the instructions.

Download a compiled version of Applio (v3.2.5 or higher).

Open a Command Prompt in the Applio folder. Run the following to install the correct PyTorch version:

env\python -m pip uninstall torch torchvision torchaudio
env\python -m pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --upgrade --index-url https://download.pytorch.org/whl/cu118

Navigate to the assets\zluda folder inside your Applio directory.
Move all .bat files from this folder to the main (root) Applio folder.
Run the patch file that corresponds to your installed HIP SDK version.
Add your HIP SDK bin directory to your Path environment variable.
Run run-applio-amd.bat.

"Check your GPU Index"

It's assumed your primary AMD GPU has an index of 0. If you have an iGPU listed first in Device Manager, edit run-applio-amd.bat and change the index from "0" to "1".

"Initial Compilation Will Be Slow"

The first time you run a task, Applio may appear to freeze for 15-20 minutes. This is normal. Zluda is compiling kernels.

‎

A console tab will appear, and after a moment your default browser, with Applio ready to use.
‎

Don't close the console until you're done using it, or it will stop working.

‎

Upload Models

‎

1. Upload voice model.

Go to the Download tab.
You have two ways of uploading it: through its link or manually inputting its files.
1. Go to the Download tab & paste the link of the model in the Model Link bar. It must be from Hugging Face or Google Drive.
  ‎
  
  ‎
2. Press Download Model.
1. Drag & drop the model's .PTH in the Drop files box below.
  ‎
  
  ‎
2. Then drag the .INDEX.

‎

Inference

‎

1. Select voice model.

Return to the Inference tab & click the Refresh button on the right.

‎
Select your model in the Voice Model dropdown.

2. Input vocals.

With Applio you can convert audios individually or in batches:
1. Drag & drop the audio or click the upload box to search it.
  ‎
  
  ‎
2. Then select it in the dropdown below.
  ‎
1. Go to the Batch tab.
  ‎
2. In the Input Folder bar, paste the path folder containing the audios.
  
  In Output Folder you can paste a path folder for the results.
  
  Ensure the paths don't contain spaces/special characters.

‎

3. Modify settings. (optional)

Unfold Advanced Settings if you wish to modify the inference settings for better results, or to determine the output folder.

‎

‎

4. Convert.

Click Convert at the bottom. The audio will begin to process.
The processing time will mainly depend on your specs, length of audio & the algorithm picked.
Once it's done, you can hear the results in the Export Audio box below.

By default the output files will be in the "audios" folder: \ApplioV3.0.7\assets\audios

‎

Training

‎

The training guide will be centered around using TensorBoard. Read about it first if you haven't already.
If you encounter an issue, be sure to read the Troubleshooting chapter.

‎

a. Model Name

‎

Go to the Train tab. Input a name for your model in Model Name.
Don't include spaces/special characters.

‎

‎

b. Dataset Path

‎

Paste the path file of your dataset in the Dataset Path bar. Ensure the path doesn't contain spaces/special characters.

‎

‎

c. Sampling Rate

‎

Select your dataset's sample rate. If you don't know the amount, click here.

‎

‎

d. Preprocess Dataset

‎

Ensure RVC Version is set as V2 & click Preprocess Dataset.

It'll finish when the output box says preprocessed successfully.

‎

a. Pitch extraction algorithm

‎

Select the algorithm you want. Use either RMVPE (most suggested) or Crepe. Applio removed pm. dio and harvest as they are outdated.

‎

b. Embedder Model

‎

Select the Embedder Model you want. Contentvec is the most used.

‎

c. Custom Embedder Model

‎

If you select "custom" for embedders, you can add your own, like Spin or Spin V2 (which seems to have better pronunciation than contentvec and better for realtime as it handles context differently)
Give it a Folder Name, like "spin".
Upload the .bin and .json files, which for example you can find them at https://huggingface.co/IAHispano/Applio/tree/main/Resources/embedders/spin for spin.
Click "Move files to custom embedder folder".
After you added your custom embedder, Refresh Embedders and select it from the Dropdown menu at the left of the refresh embedders button.

‎

d. Extract Features

‎

Press Extract Features.
It'll finish when it says extracted successfully.

‎

a. Batch Size

‎

If you are a newbie, use 8. But in case your dataset is short (around 2 minutes or less), use 4.

‎

b. Save Every Epoch

‎

Frequency of the saving checkpoints, based on the epochs.
‎
If you are a newbie, simply leave it at 15, but if you wish to be percise set it to 1.

‎
‎
‎
E.g: with a value of 10, they will be saved after the epoch 10, 20, 30, etc.

‎

c. Total Epoch

‎

Input the total amount of epochs (training cycles) for the model.
‎
But since we'll use TensorBoard, use an arbitrarily large value like 1000

‎

‎

d. Use Pretrained Models

‎

RVC uses the Orignal Pretrain by default (Pretrained is always checked, unchecking it is highly not recommended as you won't use even the Original Pretrain and train from scratc) to significantly reduce training duration and enhance overall quality. You can use the original pretrain, or community made models downloaded via the Download tab or upload them yourself.

Pretrained and Custom Pretrained options in Applio

Download Custom Pretrained Models (Optional):
- Go to the Download tab, go to the download custom pretrain, and select the community made ones like TITAN and for which sample rate you need.
- To use a pretrained model that you downloaded from the Download tab, simply check the Pretrained box.
- If you can't find the pretrain you want, you can check AI HUB's #pretrain-models or here
Use Custom Pretrained Models (Optional):
- Check the Custom Pretrained box to use your own files. This will open the Pretrained Custom Settings section.
- Upload: Click Upload Pretrained Model to open a file dialog. Here, you can drag and drop your files or click to upload the Generator (G) and Discriminator (D) .pth files. This is for when you want to upload your own pretrains which aren't in the community download pretrains tab.
- Select: After uploading, click Refresh Custom Pretraineds. Then, select your custom generator and discriminator from the Custom Pretrained G and Custom Pretrained D dropdown menus. These dropdowns will also show any custom pretrained models you have downloaded from the Download tab.

‎

e. Sync Graphs

‎

Sync graphs trains a single epoch and sets the log interval to the amount of steps that epoch trained for.
‎
Doing this makes the Tensor Board's graphs accurate.

‎

‎

f. GPU Settings

‎

If you have multiple GPUs, tick GPU Settings to use a specific one for the training.

‎

‎

g. Generate Index

‎

Click Generate Index. This will create the model's .INDEX file.

‎

h. Start Training

‎

Press Start Training to begin the training process.
‎
To open TB, execute run-tensorboard in Applio's folder. Remember to monitor it, as well as the console just in case.
‎
The latter will show you errors if they happen, and information about the epochs & checkpoints.

‎

a. Stop training

‎

When you're very sure of overtraining, you can stop training by going to the Settings tab & press Restart Applio.

‎

‎

b. Get the INDEX

‎

Create a new folder anywhere named as the model.
‎
Open Applio's folder, go to logs, and open the folder named as the model.
‎
Select the .INDEX named added_ & move it to your newly made folder.
‎
‎ ‎

‎

c. Get the PTH

‎

In said folder you'll also find all the checkpoints.
‎
Select the one closest to before the overtraining point, and move it to the new folder.

The checkpoints will be organized with this format: ModelName_Epoch_Step.pth
Example: arianagrande_e60_s120.pth

‎
‎

And that's all, have fun with your model. To test it, do a normal inference as usual.

In case the training finished but the model still needed training, you don't have to start from scratch.
‎
Simply enter the same settings & criteria that you've previously inserted. You don't have to do the preprocess or train the .INDEX again.
‎
You can change the save frequency, or increase the Total Epoch amount in case you didn't input enough before.
‎
Begin training again & remember to monitor TB & console like before.

‎

TTS

+ with any RVC model

‎

Applio is also known for having one TTS tool by default, with plenty of voices to choose for.
You can also use it with RVC models & apply the inference settings if you wish.
Aditionally, you can download the Eleven Labs TTS plugin.

‎

Instructions:

Go to the TTS tab.

‎

‎

If you want to use an RVC model, download it, go to TTS, click Refresh & select it in Voice Model & Index File.

‎
‎

To modify the inference settings or the output folder for the TTS/RVC audio, unfold Advanced Settings.

‎

In TTS Voices select the voice of your desired language, accent & gender.

In Text to Synthesize input your text. Then click Convert.

‎
‎

If you are using an RVC model, select a voice that matches the model the most, to guarantee great results.

‎

Once it's done, you'll be able to hear the result in the Export Audio box. By default, the output audio will be in the "audios" folder. < \ApplioV3.0.7\assets\audios >

‎

‎

Update

‎

To Update Applio, you need to firstly Save your audios and models, then Delete the current Applio folder and reinstall the latest version.

‎

Voice Blender

‎

The Voice Blender (also known as Model Fusion) allows you to combine the weights of two different RVC models to create an entirely new, hybrid voice. This is ideal for creating unique characters or fine-tuning a model's tonal characteristics.

1. Setup Models

Go to the Voice Blender tab.
Model Name: Enter the name for your new hybrid model.
Upload Models: You can either drag and drop the two .pth files into the upload boxes or paste their local file paths in the "Path to Model" bars.

2. Blend & Fuse

Blend Ratio: Adjust the slider to determine the influence of each model.
- A value of 0.5 is a perfect 50/50 split.
- Moving it closer to 0 favors the first model, while 1 favors the second.
Fusion: Click the Fusion button to begin the process.

3. Export Information

Output Information: Once finished, the console log and the output box will confirm the successful creation of the model.
Download/Save: The new fused model will be generated and can be found in your Applio logs or downloaded directly from the interface.

Index Blending

Note that Voice Blender only merges the .pth (weight) files. It does not merge .index files. To get the best results with a fused model, you may need to use it without an index or create a new index using the training tab with a combined dataset.

‎

Realtime Voice Conversion

‎

Beyond offline inference, Applio features a powerful built-in Realtime tab, allowing you to convert your voice live during calls, streaming, or recordings.

It uses the same RVC technology but is optimized for low-latency performance.
Features include Voice Activity Detection (VAD), input/output device selection, and advanced performance tweaking.

For a full step-by-step setup and configuration guide, please visit the Applio Realtime Guide.

‎

Extra

‎

Applio has an Extra menu, containing an audio analyzer.
Making it convenient for determining the sample rate of datasets when training models.
It also contains the model fusion tool, ideal for advanced users.

‎

Audio Analyzer:

Go to the Extra tab & press the upload box to input your audio. Or simply drag & drop.

‎

‎

Once it's done uploading, click Get information about the audio.

‎

In Sampling rate you'll see the audio's full sample rate. Use said value for training.

‎

‎

WARNING:

If the frequencies don't reach the top of the spectrogram, see at which number peaks & multiply it by 2.

‎

Example:
‎
‎

Here it reached 20 kHz. Doubling it gives 40kHz. Therefore the ideal target sample rate would be 40k

‎

Plugins

Plugins are components that you can add to Applio, that add new features & enhance your experience.
These are made by the public, and are free & easy to install.
You can find them on their GitHub page. More will be added in the future.

‎

Installation:

Access their GitHub page & click on the name of the plugin you want.

‎

‎

Click on the ZIP file.

‎
‎

Click on the download button on the right. This will download the ZIP file of the plugin.

‎

‎

Open Applio & head over to the Plugins tab. Drag & drop the ZIP file to the upload box.

‎
‎

You will be able to see its installation process in the console.

‎

‎

Go to the settings tab & click Restart Applio at the bottom. Then you'll be able to see the plugin in the Plugins tab.

‎

Troubleshooting

If you are experiencing lag, stuttering, or slow training speeds, disable HAGS (Hardware-Accelerated GPU Scheduling) on Windows 10/11. It is known to interfere with VRAM management for Local AI apps. Read the HAGS Glossary Entry for the full explanation and how to disable it.

If it's lower than 32k: select 32k.
If it's 44.1k: select 40k.
If i'ts higher than 48k: select 48k.

This a phenomenon called artifacting. To fix it, read here.

Report your issue here.

Communities

‎

`You have reached the end.`

Report Issues