Metraipolis › archive › dc_842fe01ebfc2

Whisper-Flash: Acoustically Conditioned Parallel Drafting for Faster Whisper Decoding — arXiv:2609.32869

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.32869
License (as archived)
CC0-1.0
Archived
2026-10-08T01:37:21.827Z
SHA-256 of the archived text
f6140c92788b7f6ee085053db07575f1e07822087af50682a11e5bd77f66645e
Stable URL of this page
https://metraipolis.com/documento/dc_842fe01ebfc2

How to cite

Whisper-Flash: Acoustically Conditioned Parallel Drafting for Faster Whisper Decoding — arXiv:2609.32869 — arXiv (en), https://arxiv.org/abs/2609.32869 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_842fe01ebfc2 (Metraipolis archive dc_842fe01ebfc2, sha256 f6140c92788b7f6ee085053db07575f1e07822087af50682a11e5bd77f66645e)
BibTeX
@misc{dc842fe01ebfc2,
  title = {Whisper-Flash: Acoustically Conditioned Parallel Drafting for Faster Whisper Decoding — arXiv:2609.32869},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.32869},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_842fe01ebfc2; sha256 f6140c92788b7f6ee085053db07575f1e07822087af50682a11e5bd77f66645e},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.32869. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Whisper is a widely used encoder-decoder model for speech recognition. Its encoder reads an utterance in one parallel pass, but its decoder writes the transcript one token at a time, which dominates inference time. Speculative decoding shortens such loops without changing their output: a small drafter guesses several upcoming tokens, and the original model verifies them all in one forward pass. We present Whisper-Flash, a two-layer drafter built on a property of speech recognition: the words still to be written have already been spoken. It reads Whisper's encoded audio and accepted decoder states and proposes eight tokens in a single forward pass. On the complete LibriSpeech test sets, Whisper-Flash processes $3.16\times/2.85\times$ as much audio per second as greedy decoding with identical outputs, and it remains faster at batch sizes up to 96 and under temperature sampling. Ablations show that direct access to the audio matters most.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in eess.AS).
Categorie arXiv: Sound (cs.SD) ; Audio and Speech Processing (eess.AS).
Pagina del lavoro: https://arxiv.org/abs/2609.32869 — 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 2026-10-08T01:11:22.086Z dalla pagina https://arxiv.org/catchup/eess.AS/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).