# Neural networks for Text-to-Speech evaluation

arXiv: https://arxiv.org/abs/2604.08562 · HTML: https://mlsanity.com/paper/2604.08562/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 11 references that are in the map.
- First posted on arXiv 17 Mar 2026.

## Builds on (most cited first, 8 of 8)

- [The Unreasonable Effectiveness of Deep Features as a Perceptual Metric](https://mlsanity.com/paper/1801.03924/) (2018)
- [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://mlsanity.com/paper/2106.07447/) (2021)
- [SpeechGPT: Empowering Large Language Models with Intrinsic Cross-Modal Conversational Abilities](https://mlsanity.com/paper/2305.11000/) (2023)
- [UTMOS: UTokyo-SaruLab System for VoiceMOS Challenge 2022](https://mlsanity.com/paper/2204.02152/) (2022)
- [DNSMOS: A Non-Intrusive Perceptual Objective Speech Quality metric to evaluate Noise Suppressors](https://mlsanity.com/paper/2010.15258/) (2020)
- [NISQA: A Deep CNN-Self-Attention Model for Multidimensional Speech Quality Prediction with Crowdsourced Datasets](https://mlsanity.com/paper/2104.09494/) (2021)
- [MOSNet: Deep Learning based Objective Assessment for Voice Conversion](https://mlsanity.com/paper/1904.08352/) (2019)
- [DistilMOS: Layer-Wise Self-Distillation For Self-Supervised Learning Model-Based MOS Prediction](https://mlsanity.com/paper/2601.13700/) (2026)

## Abstract (arXiv)

Ensuring that Text-to-Speech (TTS) systems deliver human-perceived quality at scale is a central challenge for modern speech technologies. Human subjective evaluation protocols such as Mean Opinion Score (MOS) and Side-by-Side (SBS) comparisons remain the de facto gold standards, yet they are expensive, slow, and sensitive to pervasive assessor biases. This study addresses these barriers by formulating, and implementing a suite of novel neural models designed to approximate expert judgments in both relative (SBS) and absolute (MOS) settings. For relative assessment, we propose NeuralSBS, a HuBERT-backed model achieving 73.7% accuracy (on SOMOS dataset). For absolute assessment, we introduce enhancements to MOSNet using custom sequence-length batching, as well as WhisperBert, a multimodal stacking ensemble that combines Whisper audio features and BERT textual embeddings via weak learners. Our best MOS models achieve a Root Mean Square Error (RMSE) of ~0.40, significantly outperforming the human inter-rater RMSE baseline of 0.62. Furthermore, our ablation studies reveal that naively fusing text via cross-attention can degrade performance, highlighting the effectiveness of ensemble-based stacking over direct latent fusion. We additionally report negative results with SpeechLM-based architectures and zero-shot LLM evaluators (Qwen2-Audio, Gemini 2.5 flash preview), reinforcing the necessity of dedicated metric learning frameworks.