According to Xiaomi Technology, the Next-Generation Kaldi team at Xiaomi Group's AI Lab recently released the ZipVoice series of text-to-speech (TTS) models, based on the Flow Matching architecture. These models include ZipVoice (a zero-shot single-speaker text-to-speech synthesis model) and ZipVoice-Dialog (a zero-shot conversational text-to-speech synthesis model). ZipVoice addresses the large number of parameters and slow synthesis speed of existing zero-shot text-to-speech synthesis models, while ZipVoice-Dialog addresses the stability and inference speed bottlenecks of existing conversational text-to-speech synthesis models.