How Do Instructions Shape Speech? Cross-Attention Attribution for Style-Captioned Text-to-Speech

• URL: http://arxiv.org/abs/2606.20532v1
• Date: Thu, 18 Jun 2026 17:47:32 GMT
• Title: How Do Instructions Shape Speech? Cross-Attention Attribution for Style-Captioned Text-to-Speech
• Abstract summary: Style-captioned text-to-speech systems use natural language to control voice characteristics, but how individual words influence acoustic output remains unclear.<n>We propose cross-attention attribution for speech diffusion models, adapting the DAAM framework to the speech domain for the first time.<n>This is the first study of how natural language influences cross-attention in speech diffusion models.
• Score: 8.950816086845114
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Style-captioned text-to-speech systems use natural language to control voice characteristics, but how individual words influence acoustic output remains unclear. Understanding this is critical for diagnosing failure modes and improving controllability in expressive TTS. We propose cross-attention attribution for speech diffusion models, adapting the DAAM framework to the speech domain for the first time, and apply it to CapSpeech-TTS. Our method extracts per-token heatmaps across 25 layers and 24 ODE steps. We analyze 3,600 (style caption, text transcript) combinations comprising 120 style captions conditioning the generation of 30 text transcripts each, revealing how caption tokens shape waveforms. Results show: (1) style tokens have lower temporal variance than content/function tokens, confirming global conditioning; (2) style attention correlates with F0 and energy; (3) style conditioning peaks in early steps and deep layers; (4) attention entropy reaches its minimum at layer 17, co-occurring with the style importance peak, indicating maximal network selectivity at the most style-critical stage. This is the first study of how natural language influences cross-attention in speech diffusion models

Related papers

• Unlocking Fine-Grained and Within-Utterance Speaking Style Control in Prompt-Based Text-to-Speech Models [11.115821694268716]
  We propose techniques to achieve continuous style attribute across utterances and time-varying style transitions within a single utterance.<n>Our proposed inter-utterance achieves a 99-100% success rate in gender conversion, up to 36 Hz pitch variation, and up to 1.6 syllables-per-second speed change.
  arXiv  Detail & Related papers  (2026-04-09T01:06:26Z)
• DiffuSpeech: Silent Thought, Spoken Answer via Unified Speech-Text Diffusion [23.01044837428522]
  We introduce textbfSilent Thought, Spoken Answer'' -- a paradigm where speech LLMs generate internal text reasoning alongside spoken responses.<n>We present method, the first diffusion-based speech-text language model supporting both understanding and generation.<n>Experiments show method achieves state-of-the-art speech-to-speech QA accuracy, outperforming the best baseline by up to 9 points.
  arXiv  Detail & Related papers  (2026-01-30T12:08:33Z)
• OZSpeech: One-step Zero-shot Speech Synthesis with Learned-Prior-Conditioned Flow Matching [3.05024318465243]
  OZSpeech is the first TTS method to explore optimal transport conditional flow matching with one-step sampling.<n>Our approach operates on disentangled, factorized components of speech in token format, enabling accurate modeling of each speech attribute.<n> Experimental results show that our method achieves promising performance over existing methods in content accuracy, naturalness, prosody generation, and speaker style preservation.
  arXiv  Detail & Related papers  (2025-05-19T07:31:55Z)
• NaturalSpeech 3: Zero-Shot Speech Synthesis with Factorized Codec and Diffusion Models [127.47252277138708]
  We propose NaturalSpeech 3, a TTS system with factorized diffusion models to generate natural speech in a zero-shot way. Specifically, we design a neural with factorized vector quantization (FVQ) to disentangle speech waveform into subspaces of content, prosody, timbre, and acoustic details. Experiments show that NaturalSpeech 3 outperforms the state-of-the-art TTS systems on quality, similarity, prosody, and intelligibility.
  arXiv  Detail & Related papers  (2024-03-05T16:35:25Z)
• Natural language guidance of high-fidelity text-to-speech with synthetic annotations [13.642358232817342]
  We propose a scalable method for labeling various aspects of speaker identity, style, and recording conditions. We then apply this method to a 45k hour dataset, which we use to train a speech language model. Our results demonstrate high-fidelity speech generation in a diverse range of accents, prosodic styles, channel conditions, and acoustic conditions.
  arXiv  Detail & Related papers  (2024-02-02T21:29:34Z)
• TextrolSpeech: A Text Style Control Speech Corpus With Codec Language Text-to-Speech Models [51.529485094900934]
  We propose TextrolSpeech, which is the first large-scale speech emotion dataset annotated with rich text attributes. We introduce a multi-stage prompt programming approach that effectively utilizes the GPT model for generating natural style descriptions in large volumes. To address the need for generating audio with greater style diversity, we propose an efficient architecture called Salle.
  arXiv  Detail & Related papers  (2023-08-28T09:06:32Z)
• ContextSpeech: Expressive and Efficient Text-to-Speech for Paragraph Reading [65.88161811719353]
  This work develops a lightweight yet effective Text-to-Speech system, ContextSpeech. We first design a memory-cached recurrence mechanism to incorporate global text and speech context into sentence encoding. We construct hierarchically-structured textual semantics to broaden the scope for global context enhancement. Experiments show that ContextSpeech significantly improves the voice quality and prosody in paragraph reading with competitive model efficiency.
  arXiv  Detail & Related papers  (2023-07-03T06:55:03Z)
• token2vec: A Joint Self-Supervised Pre-training Framework Using Unpaired Speech and Text [65.04385919645395]
  token2vec is a novel joint pre-training framework for unpaired speech and text based on discrete representations of speech. Experiments show that token2vec is significantly superior to various speech-only pre-training baselines, with up to 17.7% relative WER reduction.
  arXiv  Detail & Related papers  (2022-10-30T06:38:19Z)
• GenerSpeech: Towards Style Transfer for Generalizable Out-Of-Domain Text-to-Speech Synthesis [68.42632589736881]
  This paper proposes GenerSpeech, a text-to-speech model towards high-fidelity zero-shot style transfer of OOD custom voice. GenerSpeech decomposes the speech variation into the style-agnostic and style-specific parts by introducing two components. Our evaluations on zero-shot style transfer demonstrate that GenerSpeech surpasses the state-of-the-art models in terms of audio quality and style similarity.
  arXiv  Detail & Related papers  (2022-05-15T08:16:02Z)
• AdaSpeech 3: Adaptive Text to Speech for Spontaneous Style [111.89762723159677]
  We develop AdaSpeech 3, an adaptive TTS system that fine-tunes a well-trained reading-style TTS model for spontaneous-style speech. AdaSpeech 3 synthesizes speech with natural FP and rhythms in spontaneous styles, and achieves much better MOS and SMOS scores than previous adaptive TTS systems.
  arXiv  Detail & Related papers  (2021-07-06T10:40:45Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.