FlashTTS: Fast Streaming TTS with MTP Acceleration and X-pred Mean Flow Distillation
FlashTTS: 基于MTP加速和X-pred均值流蒸馏的快速流式TTS
机构 * Huawei Technologies Co., Ltd(华为技术有限公司)
AI总结 提出FlashTTS框架,通过滞后多轨架构、并行多令牌预测和X-pred均值流匹配解码器,实现低延迟流式TTS,首包延迟降至325ms,保持零样本语音克隆和跨语言可懂度。
Comments Accepted to Interspeech 2026