Fake No More

A platform that detects whether an audio recording is genuine or a deepfake, empowering individuals and businesses to mitigate the risks of voice-based scams. The platform leverages a Long-Short-Term Memory (LSTM) neural network model trained on 40 hours of audio recordings featuring real and fake voices from 54 prominent U.S. celebrities and politicians. By analyzing key sound wave features, such as tone and frequency, the model identifies subtle differences that reveal the authenticity of the voice. Currently optimized for native English speakers, the model's effectiveness can be broadened by incorporating diverse datasets, including non-native speakers and other languages. This expansion would enable the platform to generalize across multilingual and multicultural contexts.