🤖 AI Speech-to-Text & Translation
Never worry about untranslated raw videos again! Kyna Player offers a complete automated AI subtitle workflow from speech extraction to bilingual subtitles.
⚡ Core Pipeline
[ Extract Audio ] ──> [ Local Whisper Offline ASR ] ──> [ LLM Context Translation ] ──> [ Real-Time Dual Subtitle Mount ]

- 100% Offline Privacy Protection Whisper speech recognition runs entirely on your local GPU or CPU without uploading any audio.
- Context-Aware LLM Translation Integrates with DeepSeek, ChatGPT, Claude, or local Ollama to produce natural translations based on cinematic context.

🧠 Whisper Model Specifications
| Model Size | VRAM Usage | Speed (RTX 4060) | Accuracy | Recommended Usage |
|---|---|---|---|---|
| Base | ~1.0 GB | ~20x Ultra Fast | Good | Short clips, clear English/Chinese speech |
| Small | ~1.8 GB | ~12x Fast | Excellent (Recommended) | TV dramas, anime series daily use |
| Medium | ~3.5 GB | ~6x Steady | Very High | Academic lectures, multi-language speech |
| Large-v3 | ~6.0 GB | ~3x Steady | Top-tier Peak | Complex background music, high-precision needs |