Performance Optimization

🚀 Performance Optimization Guide

RenLocalizer is built for speed, capable of processing projects with over 100,000 lines. Use this guide to squeeze every bit of performance out of the tool.

---

🔹 Multi-Endpoint Google Architecture

Instead of using a single connection, RenLocalizer "races" multiple requests across different Google mirrors.

• Lingva Fallback: If the primary service rate-limits you, the app automatically switches to Lingva servers.
• Result: Translation speed jumps from ~3 strings/sec to ~10+ strings/sec.
• Important: Ensure "Aggressive Translation" is OFF in Settings. Enabling it forces double-checks for every line, reducing speed by ~50%.

---

⚙️ Key Tuning Settings

Setting	Recommended	Description
Parser Workers	4-8	Number of CPU threads used for file scanning. Match your CPU core count.
Concurrent Threads	32-64	Simultaneous translation requests. Set higher for fast fiber internet.
Batch Size	200-500	How many strings are sent in one block. Larger is faster but uses more RAM.
Request Delay	100ms	Pause between requests. Increase if you see HTTP 429 errors.

---

🧠 Memory & System Load

For low-end systems or very massive games:

1. Lower Batch Size: Reduces peak memory (RAM) usage.
2. Lower Workers: prevents the UI from freezing during the initial "Extracting" phase.
3. Use SIMPLE Format: Produces smaller .rpy files that are easier for both the tool and the game to handle.

---

🌐 Engine Selection

• Fastest: Google (via Multi-Endpoint).
• High Quality: DeepL (Requires API Key).
• Smartest: AI Engines (GPT/Gemini). Note: These are much slower due to the nature of Large Language Models.

---

💡 Tip: Use Google for the bulk of the game and switch to AI only for the most important story dialogue to save time.