Bagpiper-TTS: Natural Language Guided Universal Speech Synthesis

Bagpiper-TTS is a newly introduced universal speech synthesis system that interprets natural language prompts to generate comprehensive captions, which guide speech synthesis. This model supports a wide range of applications beyond traditional TTS, such as multi-talker synthesis and singing voice synthesis, achieving a 1.7% Word Error Rate (WER) on the Seed-TTS-Eval benchmark and matching performance with specialized models in evaluations. Its flexibility in handling diverse user requests is significant for practitioners seeking to develop more adaptive and capable TTS systems.