# Content is What Remains: Invariant Speech Tokenization from Parallel Utterances

arXiv: https://arxiv.org/abs/2607.19033 · HTML: https://mlsanity.com/paper/2607.19033/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 29 of its 42 references that are in the map.
- First posted on arXiv 21 Jul 2026.

## Builds on (most cited first, 29 of 29)

- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://mlsanity.com/paper/2106.07447/) (2021)
- [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://mlsanity.com/paper/2110.13900/) (2021)
- [AudioLM: a Language Modeling Approach to Audio Generation](https://mlsanity.com/paper/2209.03143/) (2022)
- [Moshi: a speech-text foundation model for real-time dialogue](https://mlsanity.com/paper/2410.00037/) (2024)
- [SpeechGPT: Empowering Large Language Models with Intrinsic Cross-Modal Conversational Abilities](https://mlsanity.com/paper/2305.11000/) (2023)
- [Libri-Light: A Benchmark for ASR with Limited or No Supervision](https://mlsanity.com/paper/1912.07875/) (2019)
- [W2v-BERT: Combining Contrastive Learning and Masked Language Modeling for Self-Supervised Speech Pre-Training](https://mlsanity.com/paper/2108.06209/) (2021)
- [Soft-DTW: a Differentiable Loss Function for Time-Series](https://mlsanity.com/paper/1703.01541/) (2017)
- [TED-LIUM 3: twice as much data and corpus repartition for experiments on speaker adaptation](https://mlsanity.com/paper/1805.04699/) (2018)
- [Seen and Unseen emotional style transfer for voice conversion with a new emotional speech dataset](https://mlsanity.com/paper/2010.14794/) (2020)
- [Textless Speech-to-Speech Translation on Real Data](https://mlsanity.com/paper/2112.08352/) (2021)
- [Efficient Self-supervised Learning with Contextualized Target Representations for Vision, Speech and Language](https://mlsanity.com/paper/2212.07525/) (2022)
- [VioLA: Unified Codec Language Models for Speech Recognition, Synthesis, and Translation](https://mlsanity.com/paper/2305.16107/) (2023)
- [ContentVec: An Improved Self-Supervised Speech Representation by Disentangling Speakers](https://mlsanity.com/paper/2204.09224/) (2022)
- [AudioDec: An Open-source Streaming High-fidelity Neural Audio Codec](https://mlsanity.com/paper/2305.16608/) (2023)
- [Vevo: Controllable Zero-Shot Voice Imitation with Self-Supervised Disentanglement](https://mlsanity.com/paper/2502.07243/) (2025)
- [XY-Tokenizer: Mitigating the Semantic-Acoustic Conflict in Low-Bitrate Speech Codecs](https://mlsanity.com/paper/2506.23325/) (2025)
- [Learning Sound Event Classifiers from Web Audio with Noisy Labels](https://mlsanity.com/paper/1901.01189/) (2019)
- [DualCodec: A Low-Frame-Rate, Semantically-Enhanced Neural Audio Codec for Speech Generation](https://mlsanity.com/paper/2505.13000/) (2025)
- [Augmentation Invariant Discrete Representation for Generative Spoken Language Modeling](https://mlsanity.com/paper/2209.15483/) (2022)
- [NAST: Noise Aware Speech Tokenization for Speech Language Models](https://mlsanity.com/paper/2406.11037/) (2024)
- [Estimating the Completeness of Discrete Speech Units](https://mlsanity.com/paper/2409.06109/) (2024)
- [DC-Spin: A Speaker-invariant Speech Tokenizer for Spoken Language Models](https://mlsanity.com/paper/2410.24177/) (2024)
- [SAC: Neural Speech Codec with Semantic-Acoustic Dual-Stream Quantization](https://mlsanity.com/paper/2510.16841/) (2025)
- [STAB: Speech Tokenizer Assessment Benchmark](https://mlsanity.com/paper/2409.02384/) (2024)
- [StableToken: A Noise-Robust Semantic Speech Tokenizer for Resilient SpeechLLMs](https://mlsanity.com/paper/2509.22220/) (2025)
- [Rethinking Discrete Speech Representation Tokens for Accent Generation](https://mlsanity.com/paper/2601.19786/) (2026)
- [Learning Disentangled Speech Representations](https://mlsanity.com/paper/2311.03389/) (2023)

## Abstract (arXiv)

Discrete speech tokenizers aim to disentangle semantic from acoustic information, yet targets from self-supervised learning (SSL) models like HuBERT retain non-linguistic variation: speaker identity, prosody, and channel conditions leak into the tokens, inflating entropy. Our key insight is that when enough speakers utter the same words under varying conditions, linguistic content is the only shared factor. We propose PINT (Parallel INvariant Tokenization), which fine-tunes an SSL encoder with alignment losses across parallel utterances and augmentations to distill this shared residual. PINT collapses identical words onto consistent token sequences, drastically reducing conditional entropy. Unlike ASR text, PINT tokens preserve frame-level temporal grounding and serve as drop-in semantic targets for audio codecs. Experiments show a 98.7% relative reduction in speaker probe accuracy (93.1% to 1.2%), a 42% lower ABX error rate, and 27-30% lower LM perplexity versus baselines, confirming that the right invariance is key to efficient learning.