# Frame-Stacked Local Transformers For Efficient Multi-Codebook Speech Generation

arXiv: https://arxiv.org/abs/2509.19592 · HTML: https://mlsanity.com/paper/2509.19592/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 21 of its 23 references that are in the map.
- First posted on arXiv 23 Sep 2025.

## Builds on (most cited first, 21 of 21)

- [MaskGIT: Masked Generative Image Transformer](https://mlsanity.com/paper/2202.04200/) (2022)
- [SoundStream: An End-to-End Neural Audio Codec](https://mlsanity.com/paper/2107.03312/) (2021)
- [Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers](https://mlsanity.com/paper/2301.02111/) (2023)
- [LibriTTS: A Corpus Derived from LibriSpeech for Text-to-Speech](https://mlsanity.com/paper/1904.02882/) (2019)
- [Moshi: a speech-text foundation model for real-time dialogue](https://mlsanity.com/paper/2410.00037/) (2024)
- [Simple and Controllable Music Generation](https://mlsanity.com/paper/2306.05284/) (2023)
- [Finite Scalar Quantization: VQ-VAE Made Simple](https://mlsanity.com/paper/2309.15505/) (2023)
- [Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions](https://mlsanity.com/paper/2502.06768/) (2025)
- [Fast Conformer with Linearly Scalable Attention for Efficient Speech Recognition](https://mlsanity.com/paper/2305.05084/) (2023)
- [VALL-E 2: Neural Codec Language Models are Human Parity Zero-Shot Text to Speech Synthesizers](https://mlsanity.com/paper/2406.05370/) (2024)
- [The T05 System for The VoiceMOS Challenge 2024: Transfer Learning from Deep Image Classifier to Naturalness MOS Prediction of High-Quality Synthetic Speech](https://mlsanity.com/paper/2409.09305/) (2024)
- [SpeechX: Neural Codec Language Model as a Versatile Speech Transformer](https://mlsanity.com/paper/2308.06873/) (2023)
- [TitaNet: Neural Model for speaker representation with 1D Depth-wise separable convolutions and global context](https://mlsanity.com/paper/2110.04410/) (2021)
- [Improved Vector Quantized Diffusion Models](https://mlsanity.com/paper/2205.16007/) (2022)
- [Halton Scheduler For Masked Generative Image Transformer](https://mlsanity.com/paper/2503.17076/) (2025)
- [VibeVoice Technical Report](https://mlsanity.com/paper/2508.19205/) (2025)
- [Masked Generative Video-to-Audio Transformers with Enhanced Synchronicity](https://mlsanity.com/paper/2407.10387/) (2024)
- [Koel-TTS: Enhancing LLM based Speech Generation with Preference Alignment and Classifier Free Guidance](https://mlsanity.com/paper/2502.05236/) (2025)
- [Improving Robustness of LLM-based Speech Synthesis by Learning Monotonic Alignment](https://mlsanity.com/paper/2406.17957/) (2024)
- [NanoCodec: Towards High-Quality Ultra Fast Speech LLM Inference](https://mlsanity.com/paper/2508.05835/) (2025)
- [Text-Conditioned Sampling Framework for Text-to-Image Generation with Masked Generative Models](https://mlsanity.com/paper/2304.01515/) (2023)

## Abstract (arXiv)

Speech generation models based on large language models (LLMs) typically operate on discrete acoustic codes, which differ fundamentally from text tokens due to their multicodebook structure. At each timestep, models must predict N codebook entries jointly, introducing dependencies that challenge simple parallel prediction approaches. Parallel prediction assumes independence among codebooks, yielding efficient decoding but often at the cost of reduced fidelity. To address this, hierarchical strategies employ a local transformer (LT) to refine predictions and capture intra-timestep dependencies. In this work, we systematically investigate two LT architectures: an autoregressive transformer that generates codebooks sequentially, and a MaskGIT-based transformer that performs iterative masked prediction. Both designs further enable frame stacking, where the primary transformer predicts multiple frames jointly, and the LT decodes their codebooks, offering improvements in speed without compromising perceptual quality. Through extensive analysis, we characterize the tradeoffs between parallel and iterative sampling strategies across different throughput and quality regimes. Finally, we propose practical guidelines for selecting decoding strategies based on deployment priorities such as computational efficiency and synthesis fidelity.