REAL CONVERSATION PREVIEW
声道 1声道 2
00:0000:00
CHINESE CONVERSATIONAL SPEECH DATA
Real Chinese
conversations
for voice AI.
Recorded human conversations that capture natural turn-taking, pauses, overlap and emotion—for teams training and evaluating voice AI.
Chinese speech data for training and evaluation.Explore the conversations
AUDIO SAMPLES
Hear how the
conversation unfolds.
Listen to featured conversations across domains, explore full recordings, and find data for your model tasks.
对话分类公开样例
More conversation samples are coming soon.
精选试听用于快速了解场景。进入分类，试听完整录音或下载样例包。
SPEECH & INTERACTION
Signals for
conversational intelligence.
Timing, prosody and overlapping speech offer a closer look at how real conversations work—and what voice models need to learn.
Timing & turn-taking
Pauses, interruptions and responses reveal how people share a conversation, informing research into real-time voice interaction.
TURN-TAKING & VOICE AGENTSProsody & expression
Tone, laughter and shifts in emotion capture the range of human expression for research into expressive speech generation.
EXPRESSION & SPEECH GENERATIONSpeakers & overlap
Explore overlapping speech and speaker-separated audio, where available, for diarization, separation and multi-party interaction.
DIARIZATION & SEPARATIONDATA SPECIFICATIONS
Know what
you’ll receive.
Review audio quality, transcripts and channel formats. We’ll help you define the dataset scope and delivery requirements for your project.
Discuss your requirements- Audio
- 48 kHz / 16 bit
- Transcripts
- Machine-generated transcripts
- Channels
- Speaker-separated audio on supported datasets
- Custom scope
- Domains, volume, annotations and delivery
Product specifications shown. Featured excerpts are MP3; full recordings are available as WAV. Confirm delivery format, annotations and licensing for your selected dataset.
OPEN RESEARCH
Research into
spoken interaction.
Our collaboration with ASLP at Northwestern Polytechnical University brings together detailed evaluation and large-scale conversational data.
Research releases follow their published licenses. Commercial datasets are licensed separately.Read the paper
ABOUT QUALIALABS
Built around
real conversation.
QualiaLabs provides Chinese conversational speech data for teams building voice models and real-time voice agents. We work with you to match conversation domains, audio and transcript formats, and delivery scope to your project.
Meet your data partnerBUILD WITH QUALIALABS
Let’s find the right data for your project.
Tell us what you’re building and which Chinese conversations you need. We’ll work with you to review samples, define the scope and agree on delivery.
微信 15300730521
01Share requirements02Evaluate samples03Agree on scope04Receive data