The Year in AI Papers
The year/Independent research

Paper 2608.02023

SwanTale: Unified Multi-Speaker Speech and Audio Generation for Instruct and Zero-Shot Tasks

Yu Zhang, Ruiqi Li, Changhao Pan, Ke Lei, Xiang Yin, Cheng Yang

Published
Aug 2026
Research lab
Independent
Citations
0
GitHub
Not linked

01 In brief

Summary

SwanTale is a unified model for multi-speaker speech and audio generation supporting both zero-shot and instruct tasks.

It introduces SwanData-Caption, a data pipeline that cleans raw audio, adds targeted synthetic coverage (elderly speech, short utterances, challenging pronunciations), and annotates multi-level captions (environment, speakers, content).

The model uses SwanVAE, a 48 kHz waveform-latent autoencoder, and a flow-based Transformer with reward-conditioned quality control, Engram conditioning, and Unified MoE for multi-task and multi-audio-modality modeling.

Curriculum learning and GRPO post-training progressively enhance capabilities.

Experiments show SwanTale leads on multiple zero-shot and instruct metrics, achieving best expressiveness scores and supporting complex instruct generation with multi-speaker speech and audio.

It ranks first in Timbre Consistency, Expressive Richness, and Expressive Hierarchy in zero-shot monologue and dialogue, and achieves the highest Mean MOS on SwanBench-Scene.

However, it shows weaknesses in English DSD and Role-Play tasks, and challenges remain in complex background music, long-form generation, and precise local style control.

02 From the paper

Abstract

Speech and audio generation is often needed in animation dubbing, audio drama, movies, advertising, games, podcasts, and short-video production. In these scenarios, creators may need to design voices without reference recordings, control speaker styles with natural language, support acoustic scenes with environments and audio effects, and later reuse the designed voices. Therefore, it is important to support multi-speaker speech and audio generation for both instruct and zero-shot tasks. The instruct task requires a caption of the environment, speaker styles, and fine-grained content, while the zero-shot task uses reference audio together with the same fine-grained content. We address these tasks from both the data and model sides. First, we propose SwanData-Caption, which cleans raw speech and audio data, adds targeted synthetic coverage, and annotates diverse and accurate multi-level captions. Then, we propose SwanTale, a multi-speaker expressive speech and audio generation model that supports both zero-shot and instruct tasks. We introduce SwanVAE to support high-quality multi-audio-modality generation. Then, we adopt reward-conditioned quality control and Engram conditioning, along with Unified MoE for multi-task and multi-audio-modality modeling. In addition, we use curriculum learning and GRPO post-training to let the model progressively learn and strengthen its capabilities. Experimental results show that SwanTale leads on multiple key zero-shot and instruct metrics, achieves the best expressiveness scores in both tasks, and supports complex instruct generation involving multi-speaker speech and audio. Demos can be found at https://swanaigc.github.io/\#swantale.