Unverified80% confidenceFactTime unknown
MiMo VL TTS在预训练阶段采用了上亿小时的语音数据，支持同一句话内完成语气转折和情感递变
1
Sources
80%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedXiaomi officially released its self-developed speech synthesis large model MIMO VR-PTS, which is based on a self-developed Audio Tokenizer and a multi-codebook joint speech-text modeling architecture, and was pre-trained on hundreds of millions of hours of speech data.72% similarUnverified现代神经网络VAD如Silero VAD基于LSTM在数千小时多语言多噪声环境音频上训练70% similarUnverifiedTraditional TTS systems require recording hours of voice data to train a speaker-specific model.70% similarUnverifiedCurrent flagship voice products generally adopt a joint training approach combining large-scale pretrained language models with dedicated speech decoders.68% similarUnverifiedLibriSpeech、CommonVoice 等主流语音识别训练集大多来自朗读录音或受控环境，信噪比高、语速均匀、句子完整68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/22824API
curl https://kongchang.com/api/v1/knowledge/claims/22824MCP
get_claim(id=22824)