MOSS-TTSD

LatestSpeech generation

MOSS-TTSD (text to spoken dialogue) is an open-source bilingual (Chinese-English) spoken dialogue synthesis model that turns a two-speaker dialogue script into natural, expressive conversational speech. It supports zero-shot two-speaker voice cloning and long single-pass speech generation, making it ideal for podcasts, interviews, chat, and other dialogue scenarios.

View API reference

ENDPOINT

POST /v1/audio/speech/speakers

Current request path

Input

Text

Content accepted by the request

Output

Audio

Content returned on success

Model overview

Multi-speaker dialogue speech synthesis model. Use the stable model ID moss-ttsd-1.0 by default; to reproduce a fixed version, pass moss-ttsd-1.0-2026-03-20 in model.

Billing

This page does not include placeholder pricing. Refer to console pricing and bills for actual units and rates.

ItemUnit priceDescription

Model invocationSee pricingUsage is measured from actual requests and outputs. Refer to console bills for the final amount.

Modality support

TTextInput only

AAudioOutput only

IImageNot supported

VVideoNot supported

Supported endpoints

Single-speaker speech/v1/audio/speechNot supported

Multi-speaker dialogue speech/v1/audio/speech/speakersSupported

Audio transcription/v1/audio/transcriptionsNot supported

Voice design/v1/audio/voice/generationsNot supported

Multimodal understanding/v1/responsesNot supported

Capabilities

PerformanceHigh

Inference speedMedium

Streaming Not supported

Async Supported

Up to 40 minutes of audio per generation
Zero-shot two-speaker cloning

Versions and snapshots

Model IDmoss-ttsd-1.0

Snapshotmoss-ttsd-1.0-2026-03-20Latest