Saved in:
Bibliographic Details
Main Authors: Chien, Chung-Ming, Tjandra, Andros, Vyas, Apoorv, Le, Matt, Shi, Bowen, Hsu, Wei-Ning
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2406.06251
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914830675345408
author Chien, Chung-Ming
Tjandra, Andros
Vyas, Apoorv
Le, Matt
Shi, Bowen
Hsu, Wei-Ning
author_facet Chien, Chung-Ming
Tjandra, Andros
Vyas, Apoorv
Le, Matt
Shi, Bowen
Hsu, Wei-Ning
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2406_06251
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Learning Fine-Grained Controllability on Speech Generation via Efficient Fine-Tuning
Chien, Chung-Ming
Tjandra, Andros
Vyas, Apoorv
Le, Matt
Shi, Bowen
Hsu, Wei-Ning
Audio and Speech Processing
Computation and Language
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.
title Learning Fine-Grained Controllability on Speech Generation via Efficient Fine-Tuning
topic Audio and Speech Processing
Computation and Language
url https://arxiv.org/abs/2406.06251