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.
# ποΈ 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.
---
## π Full Project Structure
```
african_speech_ai/
βββ main.py # Unified CLI entry point
βββ setup.py # Package installer
βββ requirements.txt # All dependencies
β
βββ config/
β βββ __init__.py
β βββ settings.py # All hyperparameters & paths (single source of truth)
β βββ languages.py # Vocab, tone maps, slang, intent/entity labels
β
βββ audio/
β βββ __init__.py
β βββ capture.py # Real-time mic capture + streaming utterance detection
β βββ vad.py # WebRTC Voice Activity Detection + energy fallback
β βββ preprocessing.py # Noise reduction, pre-emphasis, normalisation
β βββ features.py # Mel-spectrogram / MFCC extraction with delta features
β
βββ asr/
β βββ __init__.py
β βββ model.py # CNN + BiLSTM / Transformer ASR model with CTC
β βββ tokenizer.py # Character-level multilingual tokenizer
β βββ decoder.py # Greedy + beam-search CTC decoders
β βββ inference.py # Full ASR pipeline (file/bytes/stream β text)
β
βββ languages/
β βββ __init__.py
β βββ yoruba.py # Tonal processing, syllabification, IPA, slang
β βββ igbo.py # Igbo processing, tone, IPA
β βββ hausa.py # Hausa ejectives, tone, IPA
β βββ normalizer.py # Language detection, unified normaliser, transliterator
β
βββ nlp/
β βββ __init__.py
β βββ intent.py # Transformer intent classifier + keyword fallback
β βββ entities.py # Transformer NER + rule-based NER (phone/money/URL)
β βββ context.py # Multi-turn conversation context + session β¦