# How to Install Voice-Pro on Windows: The Complete Setup Guide > Easily install Voice-Pro on Windows by running start.bat. Get the complete setup guide for this powerful tool without admin rights. - Repository: [ABUS/voice-pro](https://github.com/abus-aikorea/voice-pro) - Tags: how-to-guide - Published: 2026-08-03 --- **You can install Voice-Pro on Windows by simply running `start.bat`, which automatically downloads a portable Python environment, installs all dependencies from `uv.lock`, and launches the Gradio web interface without requiring administrator privileges.** Voice-Pro is a Windows-first open-source application developed by abus-aikorea that provides AI-powered voice processing through a self-contained Gradio web UI. The installation process bundles every dependency it needs, meaning you can run it on corporate or restricted machines without system-wide changes. ## Prerequisites and System Requirements Voice-Pro is designed to run on any modern Windows PC without administrative rights. The only mandatory requirement is an internet connection for the initial download of the runtime and AI models. While the application can function without system dependencies, having **git** and **ffmpeg** already installed on your PATH may speed up the first launch. If these are missing, the installer handles them automatically by downloading portable versions. ## Step-by-Step Installation Process The installation consists of two lightweight batch scripts located in the repository root. ### Optional System Setup with configure.bat The `configure.bat` script performs one-time system-wide setup such installing *git* and *ffmpeg* via Chocolatey. However, this step is **not required** for Voice-Pro to function, as the application downloads portable versions of necessary tools on demand. Run this only if you want to install these utilities permanently on your system: ```bat configure.bat ``` *Source:* [`CLAUDE.md`](https://github.com/abus-aikorea/voice-pro/blob/main/CLAUDE.md) line 13 in the abus-aikorea/voice-pro repository. ### Core Installation with start.bat The `start.bat` script is the core installer and launcher. When executed, it performs the following actions according to the source code: 1. Downloads **uv** (a fast Python package manager) into `installer_files/uv/` 2. Uses the committed `uv.lock` to create a portable Python 3.12 environment inside `installer_files/` 3. Installs every Python package (Torch, Gradio, Faster-Whisper, etc.) in that isolated environment 4. Auto-downloads a **portable ffmpeg** binary if ffmpeg is not already on the system PATH 5. Detects a CUDA-capable NVIDIA GPU or falls back to CPU configuration 6. Launches the application by running `python start-abus.py voice` Execute the installer: ```bat start.bat ``` Typical console output during installation: ```text > start.bat Downloading uv... Creating portable Python 3.12 environment... Installing dependencies from uv.lock... Downloading ffmpeg... (if not present) Detecting GPU... NVIDIA RTX 4090 Launching Voice-Pro UI... ``` *Source:* [`README.md`](https://github.com/abus-aikorea/voice-pro/blob/main/README.md) lines 94-108. ## Understanding the Installation Architecture Voice-Pro uses a **portable Python environment** strategy to avoid conflicts with existing system Python installations. ### The uv Package Manager The installer downloads **uv** (an extremely fast Python package manager written in Rust) to handle dependency resolution. It reads the `uv.lock` file committed in the repository to ensure reproducible installations of packages like PyTorch and Gradio. ### Portable FFmpeg and GPU Detection As implemented in `start.bat`, the installation routine automatically downloads a portable ffmpeg binary if the command is not found on the system PATH. It then probes for NVIDIA CUDA capabilities, configuring the runtime to use GPU acceleration when available or falling back to CPU processing otherwise. ### Entry Point Chain The launch sequence follows this path: 1. `start.bat` calls `python start-abus.py voice` 2. [`start-abus.py`](https://github.com/abus-aikorea/voice-pro/blob/main/start-abus.py) verifies the environment and delegates to [`start-voice.py`](https://github.com/abus-aikorea/voice-pro/blob/main/start-voice.py) 3. [`start-voice.py`](https://github.com/abus-aikorea/voice-pro/blob/main/start-voice.py) imports from [`app/abus_app_voice.py`](https://github.com/abus-aikorea/voice-pro/blob/main/app/abus_app_voice.py) to create the Gradio UI The [`app/abus_config.py`](https://github.com/abus-aikorea/voice-pro/blob/main/app/abus_config.py) module loads user configuration from `config-user.json5`, while [`app/abus_path.py`](https://github.com/abus-aikorea/voice-pro/blob/main/app/abus_path.py) provides centralized path management for the `workspace/`, `model/`, and `installer_files/` directories. ## First Launch and Model Download After `start.bat` completes the environment setup, it automatically opens the Voice-Pro interface. Navigate to the printed address in your browser, normally `http://127.0.0.1:7870`. The first launch requires downloading AI models (approximately 10 GB) which are stored under the `model/` directory. Subsequent runs reuse these cached models and the installer files, making subsequent launches instantaneous. ## Troubleshooting and Recovery If any error occurs during installation or launch, the recommended recovery method is to delete the temporary installer folder while preserving your downloaded models: ```bat rmdir /S /Q installer_files start.bat ``` This forces a rebuild of the Python environment while keeping your `model/` directory intact. According to the source code analysis, this resolves most environment-related issues. ## Uninstalling Voice-Pro To completely remove the portable runtime while keeping your models and workspace data: ```bat uninstall.bat ``` For unattended removal without confirmation prompts: ```bat uninstall.bat silent ``` This removes the `installer_files/` directory and other temporary resources but preserves the `model/` and `workspace/` folders containing your data. ## Summary - **No admin rights required**: Voice-Pro installs entirely within the user directory using `start.bat`. - **Self-contained**: The installer downloads `uv`, creates a portable Python 3.12 environment in `installer_files/`, and manages its own ffmpeg binary. - **Automatic GPU detection**: Configures CUDA support automatically or falls back to CPU. - **First-time model download**: Expect approximately 10 GB download on first launch to the `model/` folder. - **Easy recovery**: Delete `installer_files/` and rerun `start.bat` to rebuild the environment without losing models. - **Clean uninstall**: Use `uninstall.bat` to remove runtime files while preserving user data. ## Frequently Asked Questions ### Does Voice-Pro require Python to be installed on my system? No. The `start.bat` script downloads and configures a portable Python 3.12 environment inside the `installer_files/` folder. It uses the `uv` package manager to install dependencies from the `uv.lock` file without affecting any system Python installations. ### Can I install Voice-Pro without administrator privileges? Yes. The installation is designed for corporate or restricted Windows PCs. You only need user-level permissions to run `start.bat`. The optional `configure.bat` requires admin rights to install system-wide tools like git and ffmpeg via Chocolatey, but these are not mandatory since Voice-Pro downloads portable versions automatically. ### How do I fix installation errors or a corrupted environment? Delete the `installer_files` folder and rerun `start.bat`. This forces the installer to recreate the portable Python environment and reinstall dependencies from `uv.lock` while preserving your downloaded AI models in the `model/` directory. This is the recommended recovery method mentioned in the repository documentation. ### Where are the AI models stored, and will I lose them when updating? The AI models are stored in the `model/` directory at the repository root. When updating Voice-Pro or rebuilding the environment, only the `installer_files/` folder is modified. Your models remain cached, making subsequent launches much faster after the initial download.