# Cross-modal Consistency Guidance for Robust Emotion Control in Auto-Regressive TTS Models

arXiv: https://arxiv.org/abs/2510.13293 · HTML: https://mlsanity.com/paper/2510.13293/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 36 references that are in the map.
- First posted on arXiv 15 Oct 2025.

## Builds on (most cited first, 25 of 25)

- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Classifier-Free Diffusion Guidance](https://mlsanity.com/paper/2207.12598/) (2022)
- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [LibriTTS: A Corpus Derived from LibriSpeech for Text-to-Speech](https://mlsanity.com/paper/1904.02882/) (2019)
- [AudioLDM: Text-to-Audio Generation with Latent Diffusion Models](https://mlsanity.com/paper/2301.12503/) (2023)
- [UTMOS: UTokyo-SaruLab System for VoiceMOS Challenge 2022](https://mlsanity.com/paper/2204.02152/) (2022)
- [CosyVoice 2: Scalable Streaming Speech Synthesis with Large Language Models](https://mlsanity.com/paper/2412.10117/) (2024)
- [DNSMOS: A Non-Intrusive Perceptual Objective Speech Quality metric to evaluate Noise Suppressors](https://mlsanity.com/paper/2010.15258/) (2020)
- [emotion2vec: Self-Supervised Pre-Training for Speech Emotion Representation](https://mlsanity.com/paper/2312.15185/) (2023)
- [NISQA: A Deep CNN-Self-Attention Model for Multidimensional Speech Quality Prediction with Crowdsourced Datasets](https://mlsanity.com/paper/2104.09494/) (2021)
- [UniAudio: An Audio Foundation Model Toward Universal Audio Generation](https://mlsanity.com/paper/2310.00704/) (2023)
- [Seen and Unseen emotional style transfer for voice conversion with a new emotional speech dataset](https://mlsanity.com/paper/2010.14794/) (2020)
- [Guided Flows for Generative Modeling and Decision Making](https://mlsanity.com/paper/2311.13443/) (2023)
- [Step-Audio: Unified Understanding and Generation in Intelligent Speech Interaction](https://mlsanity.com/paper/2502.11946/) (2025)
- [Can Large Language Models Understand Real-World Complex Instructions?](https://mlsanity.com/paper/2309.09150/) (2023)
- [HierSpeech++: Bridging the Gap between Semantic and Acoustic Representation of Speech by Hierarchical Variational Inference for Zero-shot Speech Synthesis](https://mlsanity.com/paper/2311.12454/) (2023)
- [Emo-DPO: Controllable Emotional Speech Synthesis through Direct Preference Optimization](https://mlsanity.com/paper/2409.10157/) (2024)
- [Koel-TTS: Enhancing LLM based Speech Generation with Preference Alignment and Classifier Free Guidance](https://mlsanity.com/paper/2502.05236/) (2025)
- [EmoSphere-TTS: Emotional Style and Intensity Modeling via Spherical Emotion Vector for Controllable Emotional Text-to-Speech](https://mlsanity.com/paper/2406.07803/) (2024)
- [Towards Controllable Speech Synthesis in the Era of Large Language Models: A Systematic Survey](https://mlsanity.com/paper/2412.06602/) (2024)
- [Classifier-free Guidance with Adaptive Scaling](https://mlsanity.com/paper/2502.10574/) (2025)
- [Preference Alignment Improves Language Model-Based TTS](https://mlsanity.com/paper/2409.12403/) (2024)
- [Dynamic Classifier-Free Diffusion Guidance via Online Feedback](https://mlsanity.com/paper/2509.16131/) (2025)
- [A Review of Human Emotion Synthesis Based on Generative Technology](https://mlsanity.com/paper/2412.07116/) (2024)
- [Style Mixture of Experts for Expressive Text-To-Speech Synthesis](https://mlsanity.com/paper/2406.03637/) (2024)

## Abstract (arXiv)

While Text-to-Speech (TTS) systems enable emotional control via natural-language instructions, expressiveness, naturalness, and speech quality degrade when the target emotion conflicts with the textual semantics. We propose a Cross-modal Consistency Guided Classifier-Free Guidance (CCG-CFG) method with dynamic scales based on the degree of inconsistency between the text emotion and the explicit speech emotion, replacing the dropout condition with the text emotion. We also distill the CCG-CFG guidance signal using a hard-sample mining strategy, improving the TTS model's emotional alignment capability. Evaluations on five emotional corpora and two TTS benchmarks show that our approaches applied to CosyVoice2 achieve up to a 12% absolute improvement in emotion-recognition accuracy and a 10% relative improvement in subjective scores, outperforming baselines including HierSpeech++, Qwen3-TTS, and original CosyVoice2, while preserving intelligibility, naturalness, and high speech quality.