Seed-TTS-Eval

The 'Seed-TTS-Eval' dataset/benchmark is used to evaluate the performance of text-to-speech models, specifically measuring word error rates (WERs) and similarity (SIM) scores across different test sets.

Papers using Seed-TTS-Eval (9)

dots.tts Technical Report2026Raon-OpenTTS: Open Models and Data for Robust Text-to-Speech2026PilotTTS: A Disciplined Modular Recipe for Competitive Speech Synthesis2026End-to-End Training for Discrete Token LLM based TTS System2026RobustSpeechFlow: Learning Robust Text-to-Speech Trajectories via Augmentation-based Contrastive Flow Matching2026X-VC: Zero-shot Streaming Voice Conversion in Codec Space2026LLaDA-TTS: Unifying Speech Synthesis and Zero-Shot Editing via Masked Diffusion Modeling2026JoyVoice: Long-Context Conditioning for Anthropomorphic Multi-Speaker Conversational Synthesis2025Eliminating stability hallucinations in llm-based tts models via attention guidance2025