Documentation
Complete guide to WhisperSubTranslate - from installation to advanced features.
Installation
Windows (Portable)
- Download the latest .zip file from GitHub Releases
- Extract the ZIP file to your desired location
- Run WhisperSubTranslate.exe
The application includes all necessary components (whisper.cpp, ffmpeg). No additional installation required.
Quick Start
- Launch WhisperSubTranslate
- Drag and drop your video/audio file or click "Select File"
- Select the Whisper model (small recommended for most users)
- Choose the source language (or select "Auto Detect")
- Click "Extract Subtitles" and wait for processing
- Save the generated subtitle file
System Requirements
| Component | Minimum | Recommended |
|---|---|---|
| OS | Windows 10 | Windows 10/11 |
| RAM | 4GB | 8GB+ |
| Storage | 2GB | 5GB+ |
| GPU | Optional | NVIDIA GTX 1060+ |
Whisper Models
WhisperSubTranslate supports various Whisper models with different accuracy and speed trade-offs:
Source: whisper.cpp GitHub - VRAM requirements are lower than PyTorch Whisper due to GGML optimization
| Model | Size | VRAM | Speed |
|---|---|---|---|
| tiny | 75MB | ~1GB | Fastest |
| base | 142MB | ~1GB | Fast |
| small | 466MB | ~2GB | Balanced |
| medium | 1.5GB | ~4GB | Accurate |
| large | 3GB | ~5GB | Best Quality |
| large-v2 | 3GB | ~5GB | Best Quality |
| large-v3 | 3GB | ~5GB | Best Quality |
| large-v3-turbo | 809MB | ~4GB | Fast & Accurate ⭐ |
For most users, the "small" or "large-v3-turbo" model provides the best balance of speed and accuracy.
Model Selection by Hardware
| Hardware | Recommended |
|---|---|
| CPU Only / Integrated Graphics | tiny, base |
| Low RAM (4GB or less) | tiny |
| 8GB RAM (CPU mode) | small |
| GPU 4GB VRAM | small, medium, large-v3-turbo ⭐ |
| GPU 6GB+ VRAM | Any model (including large-v3) |
For CPU-only users: Start with "tiny" or "base" for faster processing. "small" is usable but slower.
Output Formats
WhisperSubTranslate generates subtitles in the most compatible format:
- SRT - Most widely compatible format, works with all video players. The app automatically generates SRT files with millisecond-precision timestamps.
Note: The underlying whisper.cpp engine supports additional formats (VTT, TXT, JSON) via command line, but the app interface focuses on SRT for maximum compatibility.
Translation
WhisperSubTranslate can translate your subtitles using external APIs:
DeepL
- Create an account at deepl.com
- Get your API key from the account settings
- Enter the API key in WhisperSubTranslate settings
MyMemory (Free)
- No API key required for basic usage
- Select MyMemory as translation service in settings
- Free tier available with daily limits
GPT-5-nano (OpenAI)
Pricing: $0.05 input / $0.40 output per 1M tokens
- Create an account at openai.com
- Get your API key from the API settings
- Enter the API key in WhisperSubTranslate settings
Gemini 3 Flash (Google)
Free: 250 subs/day (~20-30min video), Paid: unlimited
- Go to Google AI Studio
- Create an API key (Google account required)
- Enter the API key in WhisperSubTranslate settings
Translation APIs may incur costs based on usage. Check the pricing of each service.
GPU Acceleration
GPU acceleration can dramatically speed up subtitle extraction. WhisperSubTranslate supports NVIDIA CUDA.
Requirements
- NVIDIA GPU (GTX 1060 or higher recommended)
- Latest NVIDIA drivers
- CUDA Toolkit (included with drivers)
GPU acceleration can provide 5-10x faster processing compared to CPU-only mode.
Command Line
For advanced users, the underlying whisper.cpp can be used directly from the command line:
whisper-cli.exe -m _models/ggml-small.bin -l auto input.wav
Common Options
-m- Model path-l- Source language (auto, en, ko, ja, etc.)-osrt- Output SRT format-ml- Maximum segment length in ms
Data Storage
WhisperSubTranslate stores your settings and API keys locally on your computer. This data is never uploaded to any server.
Storage Location
All settings are stored in your Windows AppData folder:
%APPDATA%\whispersubtranslate\translation-config-encrypted.json
Typical path:
C:\Users\[YourUsername]\AppData\Roaming\whispersubtranslate\
What is Stored
- API Keys - DeepL, OpenAI, Google Gemini keys (encrypted with AES)
- App Settings - Selected model, language, device, translation service
- UI Preferences - Interface language
- Error Logs - Translation error logs stored in
logs/subfolder
Log Files
Translation error logs are saved to help with troubleshooting:
%APPDATA%\whispersubtranslate\logs\translation-errors.log
These logs contain only error information for debugging purposes and do not include any personal data or API keys.
Your API keys are encrypted before storage. The settings file is stored outside the project folder and will NOT be uploaded when you share the project on GitHub.
Reset Settings
To reset all settings to defaults, delete the config folder:
- Close WhisperSubTranslate
- Press
Win + R, type%APPDATA%and press Enter - Delete the
whispersubtranslatefolder - Restart the application
Backup Settings
To backup your settings (including API keys), copy the entire whispersubtranslate folder from AppData to a safe location.
Keep your backup secure! The file contains your encrypted API keys.
Troubleshooting
App won't start
- Make sure you have Windows 10 or later
- Try running as administrator
- Check if antivirus is blocking the application
Poor accuracy
- Try a larger model (medium or large-v3)
- Make sure the audio is clear with minimal background noise
- Specify the source language instead of using auto-detect
GPU not detected
- Update NVIDIA drivers to the latest version
- Verify GPU compatibility (CUDA 11.x support required)
- Restart the application after driver update