Spaces:
Runtime error
Runtime error
| title: Translation AI Agent | |
| emoji: π | |
| colorFrom: blue | |
| colorTo: purple | |
| sdk: gradio | |
| sdk_version: 5.43.1 | |
| app_file: app.py | |
| pinned: false | |
| license: mit | |
| # π Translation AI Agent | |
| Real-time multilingual speech-to-speech and text translation AI agent powered by state-of-the-art HuggingFace models. | |
| ## π― Features | |
| - **π Text Translation**: Translate between 15+ languages instantly | |
| - **π΅ Speech-to-Text**: Convert audio to text with Whisper accuracy | |
| - **π Text-to-Speech**: Generate natural speech from translated text | |
| - **π Speech-to-Speech**: Complete audio translation pipeline | |
| - **π΄ Live Translation**: Real-time microphone translation | |
| ## π€ AI Models | |
| - **Translation**: `facebook/nllb-200-distilled-600M` (Meta NLLB) | |
| - **Speech Recognition**: `openai/whisper-base` (OpenAI Whisper) | |
| - **Text-to-Speech**: `microsoft/speecht5_tts` (Microsoft SpeechT5) | |
| ## π Supported Languages | |
| English, Spanish, French, German, Italian, Portuguese, Russian, Japanese, Korean, Chinese, Arabic, Hindi, Vietnamese, Thai, Turkish | |
| ## π Usage | |
| 1. **Text Translation**: Enter text and select languages | |
| 2. **Audio Translation**: Upload audio file for translation | |
| 3. **Live Translation**: Use microphone for real-time translation | |
| ## π Performance | |
| - **Translation Quality**: BLEU score 25-35 | |
| - **Speech Recognition**: WER < 10% for clear audio | |
| - **Latency**: < 2 seconds end-to-end (GPU) | |
| Built with β€οΈ using HuggingFace Transformers, Gradio, and PyTorch. |