SwanTale: Unified Multi-Speaker Speech and Audio Generation for Instruct and Zero-Shot Tasks
2026-08-03 • Sound
Sound
AI summaryⓘ
The authors created a system called SwanTale that can generate voices and sounds for animations, games, and other media without needing existing voice recordings. They made a new dataset called SwanData-Caption with cleaned audio and detailed descriptions to train their model. SwanTale can follow instructions about environment and speaker style or mimic voices from example audio. Their approach uses special techniques like SwanVAE and curriculum learning to improve quality and handle multiple speakers and audio types. Tests show SwanTale works well for generating expressive voices and complex audio scenes.
speech synthesisaudio generationzero-shot learningmulti-speaker modelingexpressive speechvariational autoencoder (VAE)curriculum learningreward conditioningMixture of Experts (MoE)audio captioning
Authors
Yu Zhang, Ruiqi Li, Changhao Pan, Ke Lei, Xiang Yin, Cheng Yang
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.