
🎙️ Exciting breakthrough in AI! Our new study evaluates the Audio Spectrogram Transformer (AST) for detecting synthesized speech. Achieving a low EER of 0.91%, it excels in generalizing across techs like ElevenLabs & NotebookLM, even with minimal training samples. #AI #VoiceTech
arXiv Sound@ArxivSound
``Cross-Technology Generalization in Synthesized Speech Detection: Evaluating AST Models with Modern Voice Generators,'' Andrew Ustinov, Matey Yordanov, Andrei Kuchma, Mikhail Bychkov, ift.tt/85VyfA0
English

















