A native tone-aware Yoruba Text-to-Speech solution.
Project details
Kokoro Yorùbá TTS offers a sophisticated text-to-speech experience specifically designed for the Yoruba language. Utilizing advanced models like StyleTTS2 and Kokoro-82M, this tool provides high-quality, tone-aware speech synthesis, ensuring accurate representation of Yoruba tones and allowing for realistic vocal renditions.
Kokoro Yorùbá is a native tone-aware Yoruba Text-to-Speech (TTS) system that utilizes StyleTTS2 / Kokoro-82M and is powered by the oro-g2p engine, offering enhanced phonetic accuracy. This TTS solution is tailored for the complexities of the Yoruba language, capturing its tonal nuances and delivering a more authentic auditory experience.
The system allows for straightforward implementation. For instance, executing the following code snippet demonstrates how to synthesize speech:
from tts_yoruba import YorubaTTS
tts = YorubaTTS(styletts2_dir="./StyleTTS2") # Automatically downloads Akintunde102/kokoro-yoruba-v1
tts.speak("Ẹ kú àárọ̀ o! Báwo ni àwọn ará ilé?", voice="spk8", out_path="greeting.wav")
When using this system, ensure to pass Yoruba text with correct tone markers so that the engine can accurately represent tonal variations.
Kokoro Yorùbá has been trained on 2.70 hours of curated tone-marked Yoruba data, derived from BibleTTS and WaxalNLP, demonstrating solid performance especially for short phrases. However, longer multi-syllable compounds may sometimes encounter tonal inaccuracies due to a limited training dataset.
The project aims to enhance its capabilities further:
Kokoro Yorùbá represents a significant advancement in TTS technology for the Yoruba language, providing developers and researchers a powerful tool to enhance accessibility and interaction in Yoruba speech synthesis.
Comments
0Start the conversation
Share the first comment.