Voice Generation Guide
Learn how to generate voice lines for your favorite anime and gacha game characters using Style-Bert-VITS2

Getting Started
If you're a fan of anime and gacha games, you might have wondered how to generate voice lines for your favorite characters. This comprehensive guide will walk you through the entire process, from downloading the necessary tools to creating your own voice lines.
What You'll Find Here
- Voice Line Generation Guides: Step-by-step instructions on how to generate voice lines for your favorite characters
- Audio Demonstrations: Listen to sample voice lines and audio demos in our Library
- Prompt Guidelines: Comprehensive guidelines for creating effective prompts to generate appropriate voice lines
- Voice Models: Explore different voice models and their capabilities
How to Download and Install Style-Bert-VITS2
To get started, download Style-Bert-VITS2 from the official GitHub release page below:
Download Options:
- Official Release Page(Download sbv2.zip)
- Direct Download Link
After downloading the zip file, extract it to a location that does not contain Japanese characters or spaces in the path.
Installation Instructions:
- •If you have a graphics card: Double-click on
Install-Style-Bert-VITS2.bat
(Note: This version supports training your own voice models, but some training errors may require manual fixes.) - •If you're using a CPU-only system: Double-click on
Install-Style-Bert-VITS2-CPU.bat
(Note: The CPU version can synthesize voices but cannot train new models.)
Video Tutorial
How to Generate Voice Lines
After downloading and installing Style-Bert-VITS2, follow these steps to generate voice lines:
- 1
Open the Style-Bert-VITS2 folder and double-click on
App.bat. The webpage will open in your default browser.Navigate to the 音声合成 tab (Speech Synthesis) — this is the default tab for voice generation.
- 2
Scroll down until you see モデル一覧 (Model List). Select the voice model or character you want to generate voice lines for.
- 3
If you see モデルファイル (Model File) with multiple options, select one. These options represent different training iterations of the model. Generally, models with more training iterations produce better quality results.
- 4
Enter the text you want to convert to voice in the text box. Important: The text should be in Japanese.
- 5
Click the ロード button (Load). You only need to click this once per character/model. For subsequent voice lines with the same character, you can skip this step.
- 6
Click the 音声合成 button (Speech Synthesis). The voice line will be generated and you can play it directly in the browser. You can also download the audio file for later use.
Note: The first generation may take a bit longer as the model loads and initializes.
Video Tutorial
Pro Tip: Creating Authentic Voice Lines with AI
When writing your voice lines, try using AI tools like ChatGPT or other text generators to create natural and fitting dialogue.
You can base the text on a specific character's personality — for example, generate lines as if spoken by Rem from Re:Zero or Raiden Shogun from Genshin Impact.
Don't forget to include signature phrases or expressions (like "Ufufu~", "Ara~", or other character quirks) that match the character's tone.
This makes the generated voice sound much more authentic and in-character!
How to Add New Voice Models
To add your own voice models, follow these steps:
- 1
Navigate to the
Style-Bert-VITS2/model_assetsfolder.Add your voice model folder. Each voice model should be a folder containing:
- A
config.jsonfile - A
style_vectors.npyfile - One or more
safetensorsfiles
- A
- 2
Return to the Style-Bert-VITS2 webpage and click the 更新 button (Renew). Your new voice model will now appear in the model list, and you can select it to generate voice lines.
Try Example Models
Want to test it out? You can try the Emilia voice model from the anime "Re:Zero" as an example:
Emilia Voice Model (Free)
More Free Models to Try:
Video Tutorial
Pro Tips
- Free Voice Models: You can find more free voice models on Hugging Face or other websites.
- Training Your Own Model: Once the training errors are fixed, you'll be able to train your own voice models. Alternatively, browse our Library to purchase pre-trained voice models.
- Explore Our Library: Visit the Library to browse and purchase professionally trained voice models for your favorite characters for just $2.