A complete, production-ready Python-based multilingual speech AI system supporting English, Yoruba, Igbo, Hausa, and Nigerian Pidgin. Built with PyTorch, FastAPI, and modern deep learning practices.
# 🎙️ African Speech AI
A complete, production-ready Python-based multilingual speech AI system supporting **English, Yoruba, Igbo, Hausa, and Nigerian Pidgin**. Built with PyTorch, FastAPI, and modern deep learning practices.
---
## 📁 Full Project Structure
```
african_speech_ai/
├── main.py # Unified CLI entry point
├── setup.py # Package installer
├── requirements.txt # All dependencies
│
├── config/
│ ├── __init__.py
│ ├── settings.py # All hyperparameters & paths (single source of truth)
│ └── languages.py # Vocab, tone maps, slang, intent/entity labels
│
├── audio/
│ ├── __init__.py
│ ├── capture.py # Real-time mic capture + streaming utterance detection
│ ├── vad.py # WebRTC Voice Activity Detection + energy fallback
│ ├── preprocessing.py # Noise reduction, pre-emphasis, normalisation
│ └── features.py # Mel-spectrogram / MFCC extraction with delta features
│
├── asr/
│ ├── __init__.py
│ ├── model.py # CNN + BiLSTM / Transformer ASR model with CTC
│ ├── tokenizer.py # Character-level multilingual tokenizer
│ ├── decoder.py # Greedy + beam-search CTC decoders
│ └── inference.py # Full ASR pipeline (file/bytes/stream → text)
│
├── languages/
│ ├── __init__.py
│ ├── yoruba.py # Tonal processing, syllabification, IPA, slang
│ ├── igbo.py # Igbo processing, tone, IPA
│ ├── hausa.py # Hausa ejectives, tone, IPA
│ └── normalizer.py # Language detection, unified normaliser, transliterator
│
├── nlp/
│ ├── __init__.py
│ ├── intent.py # Transformer intent classifier + keyword fallback
│ ├── entities.py # Transformer NER + rule-based NER (phone/money/URL)
│ ├── context.py # Multi-turn conversation context + session …