EmoShift: Lightweight Activation Steering for Enhanced Emotion-Aware Speech Synthesis

TL;DR

EmoShift introduces a lightweight activation-steering method for emotion-aware speech synthesis, enabling precise emotional control with minimal parameters and outperforming traditional fine-tuning approaches.

Contribution

The paper proposes EmoShift, a novel low-parameter framework with an EmoSteer layer that effectively models emotion-specific latent features in speech synthesis.

Findings

Outperforms zero-shot and fully fine-tuned baselines in evaluations

Uses less than 1/30 of the parameters compared to full fine-tuning

Enhances emotional expressiveness while maintaining naturalness and speaker similarity

Abstract

Achieving precise and controllable emotional expression is crucial for producing natural and context-appropriate speech in text-to-speech (TTS) synthesis. However, many emotion-aware TTS systems, including large language model (LLM)-based designs, rely on scaling fixed emotion embeddings or external guidance, limiting their ability to model emotion-specific latent characteristics. To address this gap, we present EmoShift, a lightweight activation-steering framework incorporating a EmoSteer layer, which learns a steering vector for each target emotion in the output embedding space to capture its latent offset and maintain stable, appropriate expression across utterances and categories. With only 10M trainable parameters,less than 1/30 of full fine-tuning, EmoShift outperforms zero-shot and fully fine-tuned baselines in objective and subjective evaluations, enhancing emotional expressiveness while preserving naturalness and speaker similarity. Further analysis confirms the proposed EmoSteer layer's effectiveness and reveals its potential for controllable emotional intensity in speech synthesis.