† Our 99% identification accuracy is measured under internal benchmarks across image, video, audio, and text assets using our proprietary, privacy-first AI and machine learning. Idem is trained on terabytes of video and audio and millions of images, delivering forensic-grade matching at enterprise scale even after heavy modifications and noise.
* The 10% partial match threshold applies to image content: Idem can correctly identify an image when up to 90% of the original has been removed or altered. Video and audio resilience is measured by different metrics. This compares favourably to conventional matching approaches, which typically require a substantially higher proportion of the original asset to be present for a match.
Partial Match (Images, Audio & Video)
Match even with a small fraction of the original image, audio, or video remaining*
Video Resilience
Can accurately match even after compression and editing.
Audio Detection
Resilient to speed/pitch changes; withstands standard and AI audio editing (e.g. stem removal, AI voice editing).
Scale
Scalable to billions of comparisons.
Accuracy
99%† (forensic grade)