Learning Fine-Grained Controllability on Speech Generation via Efficient Fine-Tuning
This work addresses the need for efficient adaptation of large generative models in speech generation, offering a resource-efficient method for fine-grained control, though it is incremental as it builds on existing pre-trained models and fine-tuning techniques.
The paper tackled the problem of adding fine-grained controllability to pre-trained speech generation models by proposing Voicebox Adapter, which integrates conditions via a cross-attention module and uses efficient fine-tuning approaches like LoRA with bias-tuning, resulting in enhanced controllability without compromising speech quality across three tasks.
As the scale of generative models continues to grow, efficient reuse and adaptation of pre-trained models have become crucial considerations. In this work, we propose Voicebox Adapter, a novel approach that integrates fine-grained conditions into a pre-trained Voicebox speech generation model using a cross-attention module. To ensure a smooth integration of newly added modules with pre-trained ones, we explore various efficient fine-tuning approaches. Our experiment shows that the LoRA with bias-tuning configuration yields the best performance, enhancing controllability without compromising speech quality. Across three fine-grained conditional generation tasks, we demonstrate the effectiveness and resource efficiency of Voicebox Adapter. Follow-up experiments further highlight the robustness of Voicebox Adapter across diverse data setups.