Resemble Detect

Description:

Identifies AI generated audio through DNN based models and disects audio to examine clues of AI manipulation that are difficult for humans to detect. Puts audio data through “spectrogram-like time-frequency embeddings”, which goes through a classification process. Probability ranging from 0 to 1, with 1 insuating that there is a high chance of the audio being fake. [3]

 

Advantage:

Over 98% accuracy [3] Isolates voices for better accuracy [4]

 

Limitations:

The free option is limited to 1,000 seconds each month. [4] Not publicly available for development

 

References:

[3] Ahmed, Z. (2024, February 6). Top 3 Deepfake Detection Tools of 2023. Resemble AI. https://www.resemble.ai/top-deepfake-detection-tools/

[4] Resemble AI – Detect – DetectorTools.ai. (n.d.). DetectorTools.ai. https://detectortools.ai/tool/resemble-ai-detect/

Tool For:Audio


How to Access:

Purchase Service
(Request a Demo)


Resources:

Resource 1

Resource 2


Last Accessed:

11/16/2024

NSF Award #2346473