Metraipolis › archive › dc_8a6d3373c8d4

GEPARD - Generative, Prosody-aware, Autoregressive text-to-speech model for Realtime Dialogue — arXiv:2609.04222

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.04222
License (as archived)
CC0-1.0
Archived
2026-10-16T22:35:43.484Z
SHA-256 of the archived text
1677b65ec37138a0035eb41abee2cdc439cdbb88e041e035eaf30307b520872c
Stable URL of this page
https://metraipolis.com/documento/dc_8a6d3373c8d4

How to cite

GEPARD - Generative, Prosody-aware, Autoregressive text-to-speech model for Realtime Dialogue — arXiv:2609.04222 — arXiv (en), https://arxiv.org/abs/2609.04222 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8a6d3373c8d4 (Metraipolis archive dc_8a6d3373c8d4, sha256 1677b65ec37138a0035eb41abee2cdc439cdbb88e041e035eaf30307b520872c)
BibTeX
@misc{dc8a6d3373c8d4,
  title = {GEPARD - Generative, Prosody-aware, Autoregressive text-to-speech model for Realtime Dialogue — arXiv:2609.04222},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.04222},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8a6d3373c8d4; sha256 1677b65ec37138a0035eb41abee2cdc439cdbb88e041e035eaf30307b520872c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.04222. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We present GEPARD (Generative, Prosody-aware, Autoregressive text-to-speech model for Realtime Dialogue), a streaming text-to-speech model for real-time spoken dialogue. GEPARD generates speech autoregressively with an LLM backbone - text and audio embeddings are trained together in a single decoder-only model - and decodes it to a waveform with an FSQ-based neural codec, streaming audio chunk-by-chunk as text arrives. Our central goal is a TTS architecture served by a standard LLM engine (vLLM) without modifying its compute kernels. This defines the overarching design principle: the backbone is a standard full-attention transformer, while all non-trivial auxiliary mechanisms - zero-shot voice cloning, text augmentation, and classifier-free guidance - are moved out of the autoregressive decode loop into prefill, or distilled directly into the weights. On streaming end-to-end inference, a single stream reaches a Real-Time Factor of about 0.067 (roughly 15x faster than real-time); under 256 concurrent streams the system reaches an aggregate speedup of about 204x on a single server-class GPU. We detail: (1) system-level solutions for vLLM-native serving; (2) the "short register" (1-2 word) failure mode of autoregressive speech decoders, with diagnostic probes and a mitigation; and (3) distillation of two-pass classifier-free guidance over text into single-pass weights via Direct Preference Optimization (DPO).

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-07 in cs.CL).
Categorie arXiv: Audio and Speech Processing (eess.AS) ; Computation and Language (cs.CL); Machine Learning (cs.LG); Sound (cs.SD).
Pagina del lavoro: https://arxiv.org/abs/2609.04222 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-16T22:21:16.307Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).