MagpieTTS-LF: Inference-Time Long-Form Speech Generation Without Training on Long-Form data
For TTS practitioners, this eliminates the need for retraining on long-form data, addressing a practical bottleneck in deploying high-quality TTS for long texts.
MagpieTTS-LF enables coherent long-form speech generation from a TTS model trained only on short utterances, using inference-time modifications that improve long-range intelligibility, prosodic coherence, speaker consistency, and boundary naturalness.
Neural Text-to-Speech (TTS) systems achieve remarkable quality on short utterances but long-form speech generation shows prosodic drift, speaker inconsistencies and sentence boundary artifacts. Existing approaches either compress sequences, increase context length or naively concatenate independently synthesized chunks. We present an inference-time approach called MagpieTTS-LF that enables MagpieTTS to produce coherent long-form speech without model retraining. Our method introduces three key innovations: (1) soft attention priors to guide monotonic alignment while preserving past and future context; (2) a stateful inference algorithm that maintains context across sentence chunks, ensuring prosodic continuity; (3) history-aware text encoding that uses past text for discourse-level prosodic planning. Experiments on long texts show significant improvements in long-range intelligibility, prosodic coherence, speaker consistency, and boundary naturalness compared to other baselines.