Search papers, labs, and topics across Lattice.
MOSS-TTS is a speech generation foundation model trained using discrete audio tokens, autoregressive modeling, and large-scale pretraining, leveraging the MOSS-Audio-Tokenizer for compressing 24kHz audio. Two models are released: MOSS-TTS, emphasizing scalability and long-context control, and MOSS-TTS-Local-Transformer, incorporating a frame-local autoregressive module for improved efficiency and speaker preservation. The models achieve zero-shot voice cloning, token-level duration control, and stable long-form generation across multilingual and open-domain settings.
Achieve controllable and scalable speech generation with MOSS-TTS, enabling zero-shot voice cloning and long-form synthesis.
This technical report presents MOSS-TTS, a speech generation foundation model built on a scalable recipe: discrete audio tokens, autoregressive modeling, and large-scale pretraining. Built on MOSS-Audio-Tokenizer, a causal Transformer tokenizer that compresses 24 kHz audio to 12.5 fps with variable-bitrate RVQ and unified semantic-acoustic representations, we release two complementary generators: MOSS-TTS, which emphasizes structural simplicity, scalability, and long-context/control-oriented deployment, and MOSS-TTS-Local-Transformer, which introduces a frame-local autoregressive module for higher modeling efficiency, stronger speaker preservation, and a shorter time to first audio. Across multilingual and open-domain settings, MOSS-TTS supports zero-shot voice cloning, token-level duration control, phoneme-/pinyin-level pronunciation control, smooth code-switching, and stable long-form generation. This report summarizes the design, training recipe, and empirical characteristics of the released models.
