Authors: Nicolas M. Müller, Piotr Kawa, Shen Hu, Matthias Neu, Jennifer Williams, Philip Sperl, Konstantin Böttinger

Published on: February 09, 2024

Impact Score: 8.2

Arxiv code: Arxiv:2402.06304

Summary

• What is new: Proposes a shift from considering audio as simply ‘fake’ or ‘real’ to identifying specific types of ‘voice edits’, including TTS and VC alterations.
• Why this is important: The challenge of identifying and categorizing voice fakes, beyond the binary fake vs. real paradigm, to address societal challenges.
• What the research proposes: A conceptual shift to focus on ‘voice edits’ with a detailed categorization and the creation of a new challenge dataset for detection.
• Results: Baseline systems for the new dataset show the feasibility of detecting various types of voice edits beyond the binary distinction.

Technical Details

Technological frameworks used: M-AILABS corpus for dataset curation

Models used: Baseline detection systems for identifying voice edits

Data used: Curated challenge dataset categorized into 6 types of voice edits

Potential Impact

Media and news reporting sectors, Security services, Podcasting platforms, and Text-to-Speech technology providers

Want to implement this idea in a business?

We have generated a startup concept here: Authentica.