mlsanityData to 14 Oct 2026

Same Bit Width, Different Outcomes: Post-Training Quantization of Text-to-Speech Across Architectures

arXiv 2609.28974 ↗First posted 24 Sep 2026eess.AS cs.LG cs.SDcited by 0

Not on the map. Why?

Abstract (arXiv)

Post-training quantization (PTQ) reduces the cost of on-device text-to-speech (TTS), but published evaluations cover one system or method. We evaluate PTQ across TTS architectures under one protocol with three core models, weight and activation ablations of eight more, and two held-out models quantized blind. Four-bit per-channel weights reduce UTMOS, a predicted mean opinion score, by 2.8 on Supertonic and 0.07 on Kokoro, and per-tensor scaling can cause severe degradation even at 8 bits. The same bit width yields different outcomes, because the sensitive component is model-specific and not reliably predicted from the model class. A staged ablation procedure identifies it, and per-layer GPTQ can restore it to within 0.1 UTMOS. Real int8 and int4 kernels reproduce the simulated ordering at hardware-dependent cost. On a Mac mini, a 4-bit weight kernel runs Supertonic at 0.60x the fp32 latency while int8 is slower, so each configuration requires validation on the target runtime.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 23 of its 32 references that are in the map. First posted on arXiv 24 Sep 2026.

Builds on

23 of 32 references.

References · 23

  1. 2022Robust Speech Recognition via Large-Scale Weak Supervisioncited by 4,057
  2. 2022SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Modelscited by 1,386
  3. 2022No Language Left Behind: Scaling Human-Centered Machine Translationcited by 1,312
  4. 2023AWQ: Activation-aware Weight Quantization for LLM Compression and Accelerationcited by 1,298
  5. 2023Scaling Speech Technology to 1,000+ Languagescited by 411
  6. 2019HAWQ: Hessian AWare Quantization of Neural Networks with Mixed-Precisioncited by 402
  7. 2022UTMOS: UTokyo-SaruLab System for VoiceMOS Challenge 2022cited by 390
  8. 2024F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matchingcited by 361
  9. 2023Q-Diffusion: Quantizing Diffusion Modelscited by 273
  10. 2022Post-training Quantization on Diffusion Modelscited by 267
  11. 2021DNSMOS P.835: A Non-Intrusive Perceptual Objective Speech Quality Metric to Evaluate Noise Suppressorscited by 172
  12. 2023StyleTTS 2: Towards Human-Level Text-to-Speech through Style Diffusion and Adversarial Training with Large Speech Language Modelscited by 152
11 more
  1. 2025VoxCPM: Tokenizer-Free TTS for Context-Aware Speech Generation and True-to-Life Voice Cloningcited by 42
  2. 2021Deep Learning Based Assessment of Synthetic Speech Naturalnesscited by 35
  3. 2021LightSpeech: Lightweight and Fast Text to Speech with Neural Architecture Searchcited by 28
  4. 2025Streaming Sequence-to-Sequence Learning with Delayed Streams Modelingcited by 27
  5. 2026OmniVoice: Towards Omnilingual Zero-Shot Text-to-Speech with Diffusion Language Modelscited by 25
  6. 2024MPQ-DM: Mixed Precision Quantization for Extremely Low Bit Diffusion Modelscited by 21
  7. 2021On-device neural speech synthesiscited by 9
  8. 2024PTQ4ADM: Post-Training Quantization for Efficient Text Conditional Audio Diffusion Modelscited by 3
  9. 2025Post-Training Quantization for Audio Diffusion Transformerscited by 3
  10. 2025Compact Neural TTS Voices for Accessibilitycited by 2
  11. 2025BitTTS: Highly Compact Text-to-Speech Using 1.58-bit Quantization and Weight Indexingcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Quantizing Whisper-small: How design choices affect ASR performance
  2. 2024Optimizing Large Language Models through Quantization: A Comparative Analysis of PTQ and QAT Techniques
  3. 2025Edge-ASR: Towards Low-Bit Quantization of Automatic Speech Recognition Models
  4. 2025PTQ1.61: Push the Real Limit of Extremely Low-Bit Post-Training Quantization Methods for Large Language Models
  5. 2025Exploring Layer-wise Information Effectiveness for Post-Training Quantization in Small Language Models
  6. 2026D2Quant: Accurate Low-bit Post-Training Weight Quantization for LLMs
  7. 2025Layer-Wise High-Impact Parameter Ratio Optimization in Post-Training Quantization for Large Language Models
  8. 2026Post-Training Ternarization of Qwen3-4B Capability, Effective Bit Budget, Storage Compression, and Deployment