Metraipolis › archive › dc_a317ed703d79

GrainSpeech: Less Context, More Detail for Compact Speech Synthesis — arXiv:2609.18856

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.18856
License (as archived)
CC0-1.0
Archived
2026-10-15T18:27:54.186Z
SHA-256 of the archived text
e6b9c0e594c87c0c92051bea94a1a54d2872fafe12fe7f9ff3ee12e40d818799
Stable URL of this page
https://metraipolis.com/documento/dc_a317ed703d79

How to cite

GrainSpeech: Less Context, More Detail for Compact Speech Synthesis — arXiv:2609.18856 — arXiv (en), https://arxiv.org/abs/2609.18856 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a317ed703d79 (Metraipolis archive dc_a317ed703d79, sha256 e6b9c0e594c87c0c92051bea94a1a54d2872fafe12fe7f9ff3ee12e40d818799)
BibTeX
@misc{dca317ed703d79,
  title = {GrainSpeech: Less Context, More Detail for Compact Speech Synthesis — arXiv:2609.18856},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.18856},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a317ed703d79; sha256 e6b9c0e594c87c0c92051bea94a1a54d2872fafe12fe7f9ff3ee12e40d818799},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.18856. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Compact acoustic models face a challenging quality-capacity trade-off. We investigate two factors in this regime: encoder context and Mel-spectrogram supervision. A receptive-field-scaling study shows that expanding self-attention beyond 15 phonemes provides no consistent gains in pitch, energy, or duration prediction. Guided by this finding, we introduce a fixed-receptive-field convolutional encoder that reduces the respective prediction errors by 36.0%, 17.3%, and 3.4%. We further show that directly transferring image-domain gradient-variance supervision restores fine-scale variation but degrades predicted quality, motivating a Mel-specific formulation with axis-specific gradients, overlapping local statistics, and log-domain variance matching. GrainSpeech contains only 264.8K parameters and achieves 17.9x real-time Mel generation on a microcontroller (MCU), while attaining UTMOS scores comparable to substantially larger models with less than 1.5% of their parameters. Source code and demos are available at this https URL .

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-17 in eess.SP).
Categorie arXiv: Audio and Speech Processing (eess.AS) ; Artificial Intelligence (cs.AI); Sound (cs.SD); Signal Processing (eess.SP).
Pagina del lavoro: https://arxiv.org/abs/2609.18856 — 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 2026-10-15T17:57:11.201Z dalla pagina https://arxiv.org/catchup/eess.SP/2026-09-17?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).