The Claim
Machine learning models trained on speech pause features achieve 79% overall accuracy when evaluated on unseen participants, new paragraphs, and previously unused cloning tools.
What the research says
Not yet evaluated
We are still looking at what the research says.
These are independent scores, not a percentage. Higher-grade studies count more, so a single strong opposing study can outweigh several weaker ones.
Machine learning models that analyze pauses in speech can correctly identify deepfakes 79% of the time when tested on new speakers, new text, and new cloning tools.
See the scientific wording
Machine learning models trained on speech pause features maintain 79% overall accuracy when tested on unseen participants, new paragraphs, and previously unused cloning tools, suggesting potential resilience against out-of-domain deepfake generation.
Human speech naturally includes tiny pauses shaped by the timing of brain signals to vocal muscles, and these pauses stay consistent even when people say new words or use voice-changing tools, allowing machines to recognize real speech by detecting these stable timing patterns.
What the research says
1 studyScientists built a computer program that listens for tiny pauses in speech to tell real voices apart from fake ones. It worked well—even on voices, sentences, and AI tools it had never seen before.
Score breakdown, mechanism chain, raw evidence, ideal studies needed & 1 supporting studies
Not medical advice. For informational purposes only. Always consult a qualified healthcare professional before making health decisions.