This is a large-scale, high-quality Bulgarian Text-to-Speech (TTS) dataset containing approximately 1300 hours of transcribed audio.
The original raw dataset consisted of 3000 hours of audio. To ensure the highest quality for training TTS models (and specifically for optimizing model context windows), a rigorous filtering and cleaning pipeline was applied. The final dataset is 1300 hours, with the… See the full description on the dataset page:
https://huggingface.co/datasets/beleata74/Bulgarian-TTS-1300h.
