When Fine-Tuning Fails and when it Generalises: Role of Data Diversity and Mixed Training in LLM-based TTS

TL;DR

This paper investigates how data diversity and mixed training influence the effectiveness of fine-tuning large language models for neural text-to-speech, demonstrating that LoRA fine-tuning enhances voice quality and speaker fidelity.

Contribution

It shows that LoRA fine-tuning improves speech quality and speaker adaptation in LLM-based TTS systems, especially with diverse training data, outperforming the base model.

Findings

LoRA fine-tuning increases DNS-MOS scores by up to 0.42 points.

Speaker fidelity improves with consistent voice similarity gains.

Signal-to-noise ratio improves by up to 34% with diverse data.

Abstract

Large language models are increasingly adopted as semantic backbones for neural text-to-speech systems. However, frozen LLM representations are insufficient for modeling speaker specific acoustic and perceptual characteristics. Our experiments involving fine tuning of the Language Model backbone of TTS show promise in improving the voice consistency and Signal to Noise ratio SNR in voice cloning task. Across multiple speakers LoRA finetuning consistently outperforms the non-finetuned base Qwen-0.5B model across three complementary dimensions of speech quality. First, perceptual quality improves significantly with DNS-MOS gains of up to 0.42 points for speakers whose training data exhibits sufficient acoustic variability. Second, speaker fidelity improves for all evaluated speakers with consistent increases in voice similarity indicating that LoRA effectively adapts speaker identity representations without degrading linguistic modeling. Third, signal level quality improves in most cases with signal to noise ratio increasing by as much as 34 percent. Crucially these improvements are strongly governed by the characteristics of the training data. Speakers with high variability in acoustic energy and perceptual quality achieve simultaneous gains in DNS-MOS voice similarity and SNR. Overall this work establishes that LoRA finetuning is not merely a parameter efficient optimization technique but an effective mechanism for better speaker level adaptation in compact LLM-based TTS systems. When supported by sufficiently diverse training data LoRA adapted Qwen-0.5B consistently surpasses its frozen base model in perceptual quality speaker similarity with low latency using GGUF model hosted in quantized form.