MOSS Transcribe Diarize Technical Report

MOSI.AI * · 2026

Read MOSS Transcribe Diarize Technical Report by MOSI.AI * (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-01554.txt.

Description

MOSI AI and 25 co-authors present a single model that writes down speech, marks who spoke, and stamps when each part was said. Meeting records need all three jobs done well together. Older systems split the jobs, lose track of speakers over long talks, and often fail to give exact times.

Their model, MOSS Transcribe Diarize, handles talks up to ninety minutes with a very long context window. It was trained on large amounts of real, messy recordings. Across public and in-house tests, the authors report it beat leading paid systems. The work points toward meeting notes that need less hand fixing after the call ends.

Catalog record

CREATORMOSI.AI *
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE21 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-01554

Browse the full library · About the Tower of Babel