English | Tiแบฟng Viแป‡t

๐ŸŽฎ RSTGameTranslation

Real-time Screen Translation for Gaming

Version 4.6
โ˜• Buy me a coffee

โœจ Features

๐Ÿ”ฅ Real-time Translation

Using EasyOCR, RapidOCR, PaddleOCR, OneOCR or Windows OCR running locally for automatic real-time translation. Supports chat window for visual novels and full-screen translation.

๐Ÿค– Advanced AI Translation

Support for multiple LLMs: Gemini, Groq, ChatGPT, Ollama, Mistral, LM Studio and Google Translate. Context-aware translation for higher accuracy.

๐ŸŽฏ Smart Recognition

Optimized translation for specific games, considering previous context and recognizing character/location names.

๐Ÿ”Š Speech-to-Text

Recognize speech directly from game audio and translate it in real-time. Perfect for games with no subtitles or audio-only dialogue.

๐Ÿ“ข Additional Features

Text-to-speech functionality and ability to translate 100% locally with Ollama or LM Studio.

๐Ÿ“ Text Overlay

Display translated text overlaid on the original text in the selected translation area.

๐ŸŽฎ Game Optimized

By providing information about the game being translated and previous context, translations are more accurate than other methods.

โšก Multiple OCR Options

Choose from 5 OCR engines: Windows OCR (built-in), OneOCR, RapidOCR, EasyOCR, or PaddleOCR based on your hardware.

โš ๏ธ Important Notes

  • System Requirements: Windows 10 or higher. Game needs to be in windowed, fullscreen borderless, or windowed borderless mode.
  • Recommended Hardware: NVIDIA GPU for best performance with EasyOCR/PaddleOCR.
  • CPU Performance: Without an NVIDIA GPU, you'll use 30-50% CPU capacity (if choosing EasyOCR or PaddleOCR).
  • OCR Selection: RapidOCR, OneOCR, WindowsOCR (recommended for AMD and Intel GPUs), RapidOCR, EasyOCR and PaddleOCR for NVIDIA
  • Translation Area: Smaller translation area = faster translation speed.
  • API Keys: Some translation services require API keys. Google Translate works without keys, but AI services need keys.

๐Ÿš€ Quick Start

For Immediate Use (No Setup Required)

  1. Run rst.exe
  2. Go to Settings โ†’ OCR: Select OneOCR or Windows OCR
  3. Go to Settings โ†’ Language: Choose source and target languages
  4. Go to Settings โ†’ Translation: Select Google Translate
  5. Close settings and press Select Window button
  6. Press Alt+Q to select translation area
  7. Press Alt+F to enable overlay display
  8. Press Alt+G to start translation

For Best Quality (Requires API Key)

  1. Get a Gemini API Key from Google AI Studio
  2. Go to Settings โ†’ Translation
  3. Select Gemini and enter your API key
  4. Select PaddleOCR or RapidOCR in OCR settings
  5. Enter your Game Info for better context-aware translations

๐Ÿ“– Documentation

Table of Contents

๐Ÿ”„ How It Works

Translation Workflow

๐Ÿ“ท Capture Screen
โ†’
๐Ÿ”ค OCR Detection
โ†’
๐Ÿค– AI Translation
โ†’
๐Ÿ“ฑ Display Result

Step-by-Step Process

  1. Capture Screen: The application captures the selected area of your game screen at regular intervals or when text changes.
  2. OCR Detection: The selected OCR engine (Windows OCR, OneOCR, RapidOCR, EasyOCR, or PaddleOCR) extracts text from the captured image.
  3. Text Processing: The extracted text is filtered, deduplicated, and analyzed for context.
  4. AI Translation: The text is sent to your selected translation service (Gemini, ChatGPT, Google Translate, Ollama, etc.)
  5. Context Enhancement: Previous translations and game information are used to improve accuracy.
  6. Display Result: Translated text appears in the overlay or chat window.

Key Concepts

  • Translation Area: The rectangular region you select on screen. Smaller areas = faster processing.
  • Context: Previous text pieces that help the AI understand the conversation flow.
  • Game Info: Description of the game that helps AI translate character names and terminology correctly.
  • Block Detection: How OCR groups text elements together. Higher power = more text grouped together.

๐Ÿ”ค OCR Engine Guide

OCR Engine Comparison

Built-in Fast

Windows OCR

Resources: ~2% CPU

Accuracy: Medium

Setup: None

Best for: AMD/Intel GPUs, quick setup

Built-in Fast

OneOCR

Resources: ~10% CPU

Accuracy: High

Setup: None

Best for: All users, manga support

External Fast

RapidOCR

Resources: ~10% GPU

Accuracy: Very High

Setup: 5-15 min

Best for: NVIDIA GPUs, Western languages

External Medium

EasyOCR

Resources: ~20% GPU

Accuracy: Good

Setup: 5-15 min

Best for: NVIDIA GPUs, many languages

External Fast

PaddleOCR

Resources: ~15% GPU

Accuracy: Very High

Setup: 5-15 min

Best for: NVIDIA GPUs, Asian languages

OCR Settings Explained

Setting Description Recommended Value
Block Detection Power Controls how text blocks are grouped. Higher values merge more text together. 5 (default) - Adjust lower for small text, higher for paragraphs
Settle Time Seconds to wait for text to stabilize before capturing. 0.5s - Increase if text flickers or updates slowly
Text Similarity Threshold Skips OCR if new text is similar to previous (0.5-1.0). 0.75 - Lower to skip more duplicates
Character Level OCR Splits results into individual characters for better spacing handling. Enabled - Recommended for tightly spaced text
Min Fragment Size Ignores text fragments smaller than this many characters. 2 - Filters out noise and small UI elements
Min Letter/Line Confidence Filters OCR output by confidence score (0.0-1.0). 0.1 / 0.2 - Lower to accept more results
Multi-Selection Area Allows selecting multiple translation regions simultaneously. Optional - Useful for HUD + chat translation

๐Ÿค– Translation Services Guide

Service Comparison

Service API Key Required Quality Speed Privacy
Google Translate Optional (Free tier available) Good Fast Cloud
Gemini Yes (free tier available) Excellent Fast Cloud
ChatGPT Yes (paid) Excellent Fast Cloud
Groq Yes Excellent Very Fast Cloud
Ollama No Good-Very Good Depends on hardware 100% Local
LM Studio No Good-Very Good Depends on hardware 100% Local
Mistral Yes Excellent Fast Cloud

Context Settings

Setting Description Recommended Value
Max Context Pieces Number of previous translations to include as context. 3 - Provides good context without overwhelming
Min Context Size Minimum characters for text to be included as context. 20 - Filters out menu labels and buttons
Game Info Description of the game for better terminology handling. Brief description of genre, setting, characters
Leave Translation Onscreen Always show translation instead of original text. Personal preference
๐Ÿ’ก Tip: Using context significantly improves translation quality, especially for conversations. The AI can understand character relationships and story flow better.

โš™๏ธ Detailed Settings Reference

Language Tab

Setting Description
Source Language The language of text in the game (e.g., ja for Japanese, ko for Korean).
Target Language The language you want translations in (e.g., en for English, vi for Vietnamese).
Swap Languages Quickly swap source and target languages.
Check Language Pack Verify Windows OCR language pack is installed for your source language.

OCR & Screen Tab

Setting Description
OCR Method Select OCR engine: EasyOCR, Windows OCR, PaddleOCR, RapidOCR, or OneOCR.
Select Screen Choose which monitor to capture if using multiple displays.
Auto OCR Enable automatic continuous translation. Disable for manual control.
HDR Support Enable for HDR games (may affect text detection accuracy).
Remove OCR Data Delete downloaded OCR models to free disk space.

Ignore Phrases Tab

Setting Description
Phrases to Ignore List of text patterns to skip during translation. Useful for removing UI elements, watermarks, or repetitive text.
Exact Match When enabled, ignores entire text if it matches. When disabled, removes the phrase from text.

Exclude Regions Tab

Setting Description
Exclude Regions Define rectangular areas within the translation area that should never be captured or translated.

Text-to-Speech (TTS) Tab

Setting Description
Enable TTS Read translated text aloud using synthesized voice.
TTS Service Select: ElevenLabs (best quality), Google Cloud TTS, or Windows TTS.
Voice Selection Choose voice for TTS output.
Exclude Character Names Skip reading character names in dialogue.

Hot Keys Tab

Setting Description
Enable Hot Keys Master switch for keyboard shortcuts.
Function Selection Choose which action to assign a hotkey to.
Combine Keys Set modifier key (CTRL, SHIFT, ALT) + letter/number.

Audio Processing Tab

Setting Description
Auto-Translate Audio Start Whisper service for speech-to-text translation from game audio.
Model Selection Choose local Whisper model size (larger = better accuracy, slower).
Silence Threshold Audio level below which is considered silence (0.0-1.0).
Max Buffer Samples Maximum audio buffer size for processing.

Server Controls Tab

Setting Description
Install/Start/Remove Server Manage OCR backend servers (EasyOCR, PaddleOCR, RapidOCR).
Send Data to Server Allow diagnostic data posting to remote servers (used by some OCR engines).

Game Profiles Tab

Setting Description
Create/Load Profiles Save translation area rectangles and settings as named profiles for different games.

โŒจ๏ธ Keyboard Shortcuts

Shortcut Action Description
Alt + Q Select Area Select translation area on screen
Alt + F Toggle Overlay Show/hide translation overlay
Alt + G Start/Stop Start or stop translation
Alt + C ChatBox Open translation history window
Alt + S Settings Open settings window
Alt + L Log Open log window
Alt + A Audio Service Toggle Whisper audio service
Alt + X Clear Areas Clear all selected translation areas
Alt + 1-5 Select Area N Quick switch to area 1-5
Alt + R Retry Retry last failed translation
Alt + W Swap Languages Swap source and target languages

๐Ÿ“š Use Cases & Examples

Case 1: Visual Novel Translation

Recommended Setup:
OCR: OneOCR or Windows OCR
Translation: Gemini with context enabled
Display: ChatBox mode
Settings: Enable "Leave Translation Onscreen" for easy reading

Steps:

  1. Select the dialogue box area using Alt+Q
  2. Set Max Context to 3-5 pieces for conversation flow
  3. Enter game info: "Visual novel, Japanese school setting"
  4. Start translation with Alt+G

Case 2: Action RPG Game HUD Translation

Recommended Setup:
OCR: RapidOCR (if NVIDIA) or OneOCR (if AMD/Intel)
Translation: Google Translate (fast) or Gemini (accurate)
Display: Overlay mode with auto background

Steps:

  1. Select the HUD area containing item names, quest text, etc.
  2. Set Block Detection Power lower (3-4) for individual text elements
  3. Enable "Auto Set Overlay Background Color"
  4. Use Alt+F to toggle overlay as needed during gameplay

Case 3: Manga/Comic Translation

Recommended Setup:
OCR: OneOCR (has manga support)
Translation: Gemini for best quality
Display: ChatBox mode
Settings: Lower Text Similarity Threshold (0.6) for speech bubbles

Steps:

  1. Select the speech bubble area
  2. Disable Character Level OCR for connected text
  3. Enable "Leave Translation Onscreen"
  4. Use Multi-Selection Area if manga has multiple text boxes

Case 4: Audio-Only Games (Speech-to-Text)

Recommended Setup:
OCR: Not needed
Translation: Gemini or Ollama (local)
Audio: Whisper with "base" or "small" model
Display: ChatBox mode

Steps:

  1. Enable "Auto-Translate Audio" in Audio Processing tab
  2. Select Whisper model (ggml-small for balance, ggml-base for speed)
  3. Adjust silence threshold to avoid processing background noise
  4. Translations appear in ChatBox as audio is recognized

Case 5: 100% Offline Translation

Recommended Setup:
OCR: OneOCR or Windows OCR (both work offline)
Translation: Ollama or LM Studio (100% local)
No internet connection required after initial setup

Steps:

  1. Install Ollama or LM Studio
  2. Download a model (llama3, phi3, or mistral)
  3. Enter URL and model name in Translation settings
  4. All translations stay on your computer

๐Ÿ’พ Installation & Usage (Windows)

Prerequisites:

  1. Download the latest version (zip file) here and extract it.

Setup Options:

๐Ÿ”ต Simple Setup (No Installation)

  1. Run `rst.exe`
  2. Go to Settings โ†’ OCR: Select OneOCR
  3. Go to Settings โ†’ Language: Choose source and target languages
  4. On Language tab, If you are using Windows OCR, please click the "Check" button to verify the language pack before starting
  5. Go to the Translation tab, select the translation service you want to use
  6. Now you can close the settings window
  7. Press button Select Window: Choose window which you want to capture
  8. Press Alt+Q to select area, then Alt+F to turn on Overlay
  9. Press Alt+G to start/stop

๐Ÿ”ด Advanced Setup (Need Installation)

  1. Server Setup (Only do once for each OCR method):
    • Click the SetupOCR button to start setting up the server based on the OCR method you selected in settings (If you selected Windows OCR, you can skip this step)
    • Setup can take 5-15 minutes, depending on your internet speed and computer power
    • Wait until the server setup is complete, you will see a message "... environment setup completed"
  2. Start Translating:
    • Click the StartOCR button and wait until you see the message "Successfully connected to .... server" (If you selected Windows OCR, you can skip this step)
    • Select translate region (ALT+Q or Click SelectArea button)
    • Click Start button (ALT+G) to begin translation
    • Translation results are displayed in the chat window (ChatBox button, ALT+C) or in the Overlay (Overlay button, ALT+F)
  3. LLM Setup:
    • Go to settings and add your Gemini API key. There are instructions on how to get the API key there.
    • You can enter multiple API keys, press Enter after entering each API key.
    • Check the other settings; the default values should be fine. Note that there is a place to enter the name of the game, which is important! The LLM knowing this will help it fix errors and create better dialogue.

Note: The first time you use EasyOCR or PaddleOCR with a new language, it has to download first! So it might seem like it's not working, just wait a minute or two and start/stop the application's translation, it will work.

๐Ÿ”„ How to Update

RSTGameTranslation will automatically check for updates when you start it. If there's a new version, you'll see a notification asking if you want to download it. To update:

โš™๏ธ Advanced Settings

Recommended Settings

Performance Tips

โ“ Frequently Asked Questions (FAQ)

Q: Why is the translation slow?

Possible causes and solutions:

  • Large translation area: Reduce the area size for faster processing
  • OCR engine: Try OneOCR or Windows OCR for faster speeds
  • API response time: Google Translate is fastest, followed by Gemini
  • CPU/GPU load: Close other resource-heavy applications
Q: OCR is not detecting text correctly

Troubleshooting steps:

  • Check language selection: Make sure source language matches the game
  • Adjust OCR settings: Lower the confidence thresholds
  • Try different OCR: Some engines work better with certain text styles
  • Clear context: Sometimes old context can interfere
  • Check game mode: Ensure game is in windowed or borderless mode
Q: How do I get an API key?

For different services:

Q: Can I use this offline?

Yes! Options for offline use:

  • Windows OCR / OneOCR: Built-in, works without internet
  • Ollama: Download from ollama.com for 100% local AI
  • LM Studio: Download from lmstudio.ai for local models
Q: How to translate multiple areas at once?

Multi-selection setup:

  1. Go to Settings โ†’ OCR Settings
  2. Enable "Multi-selection area"
  3. Select your first area with Alt+Q
  4. Press Alt+Q again to add another area
  5. Repeat for up to 5 areas
Q: The overlay covers game UI

Solutions:

  • Make your translation area more precise
  • Use "Exclude Regions" to avoid capturing UI elements
  • Toggle overlay with Alt+F when not needed
  • Try ChatBox mode instead of overlay
Q: Can I save different settings for different games?

Yes! Use Game Profiles:

  1. Configure settings for your first game
  2. Go to Settings โ†’ Game Profiles
  3. Click "Create Profile" and name it
  4. Repeat for other games
  5. Load profiles quickly when switching games
Q: Why does translation quality vary?

Factors affecting quality:

  • Context: Enable context for better conversation flow
  • Game Info: Enter game description for terminology accuracy
  • OCR quality: Clearer text = better translations
  • AI model: Gemini and GPT-4 provide best results
Q: How to report bugs or get help?

Support channels: