CodecFake+: Codec-Based Resynthesized Data as a Proxy for Detecting CodecFake Speech

Xuanjun Chen *, Jiawei Du *, Haibin Wu †, Lin Zhang, I-Ming Lin, I-Hsiang Chiu, Wenze Ren · 2025

Read CodecFake+: Codec-Based Resynthesized Data as a Proxy for Detecting CodecFake Speech by Xuanjun Chen *, Jiawei Du *, Haibin Wu †, Lin Zhang, I-Ming Lin, I-Hsiang Chiu, Wenze Ren (2025) 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-2501-08238.txt.

Description

Xuanjun Chen and co-authors built this 2025 data set for catching a new breed of fake speech. Modern audio codecs can now generate voices realistic enough to mimic a person and spread lies. Detectors trained on older synthesis fakes miss these codec fakes. The team's answer is a large, diverse training ground.

They re-synthesized speech through thirty-one open codec models for training, and gathered web output from seventeen advanced generators for testing. A taxonomy sorts codecs by their building blocks, letting analysts see which parts make fakes hard to spot. Training on re-synthesized speech proved a workable stand-in for real fakes. The data set is the largest and most varied of its kind, aimed at keeping detectors current.

Catalog record

CREATORXuanjun Chen *, Jiawei Du *, Haibin Wu †, Lin Zhang, I-Ming Lin, I-Hsiang Chiu, Wenze Ren
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE74 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-08238

Browse the full library · About the Tower of Babel