Real-Time VC: Mastering Voice Conversion

RVC Real-Time VC Screenshot

01. Basic Concepts & Hardware

Real-time Voice Conversion is demanding. It slices your voice into fragments, processes them through AI, and plays them back instantly.

💡 Is your voice stuttering?

If you see Using CPU mode in red, the AI load is on your processor. An NVIDIA RTX graphics card is strongly recommended for smooth conversion.

02. Audio Routing (How others hear you)

By default, only you can hear the changed voice. To use it in Discord or OBS, you must install a "virtual audio cable."

Setup & Routing Steps

  1. In our software:
    • Input Device (Mic): Your physical microphone.
      👉 Recommended: Physical Microphone (e.g., Realtek Audio or USB Device).
      ⚠️ Critical: NEVER select CABLE Output as the mic here.
    • Output Device (Speaker): Where the converted voice goes.
      👉 To Discord/OBS: Select CABLE Input (VB-Audio Virtual Cable).
      👉 To hear yourself: Select your speakers or headset.
  2. In Discord/OBS Settings:
    • Input Device: Select CABLE Output (VB-Audio Virtual Cable).

📋 Common Audio Devices Explained

Device Type Description
Microsoft Sound Mapper ✅ Default Points to your current Windows default recording/playback device.
CABLE Input (VB-Audio Cable) ❓ Missing Virtual cable. If not visible, please install as described below.
Microphone (Realtek / USB Audio) ⭐ Rec. Mic Physical microphone. Best quality and lowest latency.
Headset (Hands-Free) ❌ Avoid Bluetooth call mode. Poor quality (8kHz) causes severe AI glitches.

📡 VB-Cable Installation Guide

If you don't see CABLE in the list, follow these steps:

  1. Download: Visit the VB-Audio Website and download **VB-CABLE Driver**.
  2. Install: Extract, right-click VBCABLE_Setup_x64.exe, and "Run as administrator".
  3. Refresh: Once installed, **restart our app** and click "Rescan".

💡 Flow: You speaking ➔ Studio0808 (Output: CABLE Input) ➔ Discord (Input: CABLE Output).

03. Core Parameters

1. Pitch Shift

2. Input Gain

If you hear popping, lower the Gain (e.g., 0.8) to keep the signal clean.

3. Silence Threshold

Filters background noise.AI won't run when your volume is below this level. Move to the right if you hear constant noise.

4. Release Time (Rel)

Controls how long the AI keeps generating sound after you stop speaking. Recommended: 0.2s - 0.5s for smooth sentence endings. ⚠️ Note: If you experience an echo loop when using "Sound Mapper," set this to 0.0s.

5. Latency (Sec)

The size of each audio chunk. Faster GPUs (RTX 40/50 series) can handle lower values (0.2s ~ 0.4s). Increase this value if you hear stuttering or dropouts.

🚀 Lag-Prevention Technology

Our system features a smart buffer-clearing mechanism. If the GPU falls behind, it automatically discards old audio chunks to ensure that the output is always real-time, preventing any perceived lag or "delayed echo."