MOSS-TTSD

MOSS-TTSD is an open source bilingual dialogue synthesis model that supports natural and expressive speech generation. It converts conversation scripts into high-quality voice and is suitable for podcast production and AI conversation applications. The model features zero-shot speech cloning and long-term speech generation, with a high level of expression and sense of reality. The training basis of MOSS-TTSD includes large-scale language data and speech data, ensuring the naturalness and accuracy of the generated speech. The technology is suitable for commercial use and is completely open source.

Demand population:

"This product is suitable for developers of voice synthesis, podcasting and conversational AI applications, especially content creators and researchers who need high-quality voice generation. MOSS-TTSD provides a flexible and powerful platform that enables users to generate natural and smooth conversational audio that meets business and educational needs."

Example of usage scenarios:

Podcast audio generated by MOSS-TTSD improves the audibility of content.

An interactive voice answering system for online education platforms.

In entertainment applications, add real sound performance to character conversations.

Product Features:

Supports Chinese and English dialogue voice generation.

Achieve zero- shot Two-person voice clones to accurately switch speakers.

Long-term voice generation, suitable for AI podcast production.

Highly expressed dialogue voice, close to the tone of human natural dialogue.

It provides two inference methods: local and API, which are convenient for users to use.

Supports batch processing tools and can handle multiple generation requests at the same time.

Includes podcast generation tools that convert long text or web content into audio.

Provides simple fine-tuning scripts for user-defined models.

Tutorials for use:

Install the required dependency libraries and set up a Python environment.

Download and prepare the XY Tokenizer model weights.

Prepare input files in JSONL format, including conversation scripts and speaker audio references.

Run the inference script, specifying the input file path and output directory.

View the generated audio files for subsequent processing or publishing.