The smarter subtitle translator. It reads your SRT, groups sequential lines for context, and uses GPT to produce translations that actually sound human.

  •  

context-aware-srt-translation is the translator your subtitles deserve. Stop feeding GPT one line at a time and getting robotic, disconnected results. This service groups sequential subtitle lines together, giving the AI the context it needs to understand the conversation and produce translations that actually flow naturally.

How it works:

• You: POST your SRT file to the API
• Service: Groups lines into context windows, translates concurrently
• Result: Natural translations that respect conversational flow
• Bonus: Full statistics on what happened

Line-by-line translation is a vibe-killer. Context windows make other methods look ancient.

"I think we should..."  →  "Sanırım biz..."
"...go there tomorrow"  →  "...yarın oraya git"

["I think we should...",
 "...go there tomorrow"]  →  
["Bence yarın oraya...",
 "...gitmeliyiz"]

The difference is context. When GPT sees the full thought, it understands the sentence structure, maintains speaker tone, and produces translations humans would actually write.

git clone https://github.com/yigitkonur/context-aware-srt-translation-gpt.git
cd context-aware-srt-translation-gpt
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

cp .env.example .env
# Add your OpenAI API key (required)
# Add DeepL API key (optional fallback)

python run.py

The API is now live at http://localhost:8000 🎉

Instead of translating each subtitle line individually (which loses context), this service groups sequential lines:

┌─────────────────────────────────────────────────┐
│  Traditional: Line 1 → Translate → Output 1    │
│               Line 2 → Translate → Output 2    │
│               Line 3 → Translate → Output 3    │
│               ❌ No context between lines       │
└─────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────┐
│  Context Window:                                │
│  [Line 1, Line 2, Line 3] → Translate Together  │
│               ↓                                 │
│  [Output 1, Output 2, Output 3]                 │
│               ✅ AI sees the full picture       │
└─────────────────────────────────────────────────┘

This allows GPT to:

• Maintain speaker continuity — Same character, same voice
• Preserve conversation flow — Questions match answers
• Handle split sentences — "I think..." + "...we should go" = coherent thought
• Respect cultural context — Idioms translated appropriately

curl -X POST "http://localhost:8000/subtitle-translate" \
  -H "Content-Type: application/json" \
  -d '{
    "srt_content": "1\n00:00:01,000 --> 00:00:04,000\nHello, how are you?\n\n2\n00:00:05,000 --> 00:00:08,000\nI am doing great, thanks!",
    "source_language": "en",
    "target_language": "tr"
  }'

{
  "translated_srt_content": "1\n00:00:01,000 --> 00:00:04,000\nMerhaba, nasılsın?\n\n2\n00:00:05,000 --> 00:00:08,000\nÇok iyiyim, teşekkürler!",
  "status": "success",
  "error_message": null,
  "stats": {
    "total_sentences": 2,
    "translated_sentences": 2,
    "failed_sentences": 0,
    "success_rate": 100.0,
    "openai_calls": 1,
    "deepl_calls": 0,
    "elapsed_seconds": 1.23
  }
}

curl http://localhost:8000/health
# {"status": "healthy", "version": "2.0.0"}

All settings via environment variables:

src/
├── config.py              # Environment configuration
├── models.py              # Pydantic request/response models
├── srt_parser.py          # SRT parsing & reconstruction
├── translator.py          # Main orchestration logic
├── main.py                # FastAPI application
└── services/
    ├── base.py            # Service interface
    ├── openai_service.py  # OpenAI implementation
    └── deepl_service.py   # DeepL fallback

Interactive docs available when running:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

# Setup
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

# Run tests
pytest tests/ -v

# Run with hot reload
python run.py