# The Evolving Bottleneck in Speech Generation: Interface Co-design and Staged Alignment from CosyVoice to Qwen-Audio-3.0-TTS

arXiv: https://arxiv.org/abs/2609.16514 · HTML: https://mlsanity.com/paper/2609.16514/

## Key facts

- No citing papers in the map yet (data through 19 Oct 2026).
- Placed in Vocoders and speech synthesis, one of 11 directions in Speech and audio language models.
- Builds on 14 of its 15 references that are in the map.
- First posted on arXiv 15 Sep 2026.

## Placed in

- Speech and audio language models / [Vocoders and speech synthesis](https://mlsanity.com/speech/vocoders-and-speech-synthesis/)

## Builds on (most cited first, 14 of 14)

- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers](https://mlsanity.com/paper/2301.02111/) (2023)
- [Finite Scalar Quantization: VQ-VAE Made Simple](https://mlsanity.com/paper/2309.15505/) (2023)
- [F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching](https://mlsanity.com/paper/2410.06885/) (2024)
- [Voicebox: Text-Guided Multilingual Universal Speech Generation at Scale](https://mlsanity.com/paper/2306.15687/) (2023)
- [CosyVoice: A Scalable Multilingual Zero-shot Text-to-speech Synthesizer based on Supervised Semantic Tokens](https://mlsanity.com/paper/2407.05407/) (2024)
- [Seed-TTS: A Family of High-Quality Versatile Speech Generation Models](https://mlsanity.com/paper/2406.02430/) (2024)
- [CosyVoice 2: Scalable Streaming Speech Synthesis with Large Language Models](https://mlsanity.com/paper/2412.10117/) (2024)
- [MaskGCT: Zero-Shot Text-to-Speech with Masked Generative Codec Transformer](https://mlsanity.com/paper/2409.00750/) (2024)
- [CosyVoice 3: Towards In-the-wild Speech Generation via Scaling-up and Post-training](https://mlsanity.com/paper/2505.17589/) (2025)
- [E2 TTS: Embarrassingly Easy Fully Non-Autoregressive Zero-Shot TTS](https://mlsanity.com/paper/2406.18009/) (2024)
- [DiTAR: Diffusion Transformer Autoregressive Modeling for Speech Generation](https://mlsanity.com/paper/2502.03930/) (2025)
- [JoyVoice: Long-Context Conditioning for Anthropomorphic Multi-Speaker Conversational Synthesis](https://mlsanity.com/paper/2512.19090/) (2025)
- [Qwen-Audio-3.0-TTS: Freely Controllable and Highly Robust Speech Synthesis with Multi-Stage Training Paradigm](https://mlsanity.com/paper/2607.23938/) (2026)

## Abstract (arXiv)

Speech synthesis systems are commonly narrated as a sequence of larger models, better tokenizers, and broader data. This technical retrospective offers a different account of the CosyVoice lineage, from CosyVoice through CosyVoice 2 and CosyVoice 3 to Qwen-Audio-3.0-TTS: progress came from repeatedly relocating the system's dominant bottleneck. Across the lineage, a stable decomposition separates an autoregressive language model that plans speech from a flow-matching model that renders acoustics. What changes is the contract between them. CosyVoice establishes supervised semantic tokens as a content-aligned interface; CosyVoice 2 makes that interface causally available for streaming and removes the utterance-level speaker embedding from the language model; CosyVoice 3 improves the learnability and coverage of the interface through multitask supervision, scaling, and differentiable reward optimization; and Qwen-Audio-3.0-TTS reduces token rate, conditions its renderer on continuous language-model hidden states instead of token embeddings, and progressively aligns the coupled system. We formalize this history through four interface dimensions---representation, ownership, availability, and gradient reach---and separate within-paper evidence from cross-paper comparison. The resulting synthesis connects discrete autoregressive, continuous non-autoregressive, hybrid, and continuous autoregressive speech-generation paradigms, and yields practical principles for diagnosing and training modular speech generators.