ChatTTS: Text-to-Speech For Chat

ChatTTSisavoicegenerationmodelonGitHubat2noise/chattts,ChatTTSisspecificallydesignedforconversationalscenarios.Itisidealforapplicationssuchasdialoguetasksforlargelanguagemodelassistants,aswellasconversationalaudioandvideointroductions.ThemodelsupportsbothChineseandEnglish,demonstratinghighqualityandnaturalnessinspeechsynthesis.Thislevelofperformanceisachievedthroughtrainingonapproximately100,000hoursofChineseandEnglishdata.Additionally,theprojectteamplanstoopen-sourceabasicmodeltrainedwith40,000hoursofdata,whichwillaidtheacademicanddevelopercommunitiesinfurtherresearchanddevelopment.

电影视频 2025-07-17