JONAHKYAGABA/multilingual-african-speech-translation

Afrivoice is a robust audio processing pipeline designed to transcribe speech from African languages, translate it (via English), and synthesize the translated text back into audio. It integrates a wide range of ASR and TTS models to support African linguistic diversity. # multilingual-african-speech-translation Afrivoice is a robust audio processing pipeline designed to transcribe speech from African languages, translate it (via English), and synthesize the translated text back into audio. It integrates a wide range of ASR and TTS models to support African linguistic diversity. Features 🎙️ Automatic Speech Recognition (ASR) for over 20 African languages using models from SpeechBrain, HuggingFace, and Whisper. 🌐 Translation via Microsoft Azure or Google Translate APIs (configurable). 🔊 Text-to-Speech (TTS) for supported languages using VITS models. 🔄 End-to-End Pipeline: Input speech in one language → Transcription → Translation → Synthesized audio in another language. Supported Languages Language Code Language ASR Support TTS Support kin Kinyarwanda ✅ ✅ lug Luganda ✅ ✅ swh Swahili ✅ ✅ am Amharic ✅ ✅ ha Hausa ✅ 🚫* yo Yoruba ✅ ✅ ig Igbo ✅ 🚫* xho Xhosa ✅ ✅ zul Zulu ✅ ✅ tir Tigrinya ✅ ✅ so Somali ✅ ✅ ach, nyn, teo, lgg Sunbird languages (Acholi, Runyankole, Ateso, Lugbara) ✅ ✅