In addition to extensive AI capabilities that recognize and tag celebrities in our customers’ videos, AnyClip now introduces a new AI model capable of recognizing and tagging in a video any person it is trained on.

To train, please send the photos of the person you want recognized and tagged to your Customer Success manager, along with the name to be assigned as a People category tag for the videos where this person is recognized.

For optimal results, provide 3 to 20 photos where the person's face is clearly visible and takes up most of the picture. To achieve that you may crop your source picture.

There are no strict requirements regarding picture quality in terms of file size, resolution, or dimensions. However, higher picture quality directly reduces the occurrence of false positive recognition results. Hence, please provide the highest quality images available.

The confidence level, determining the filter for potential false positives, is customizable, with the default set at 90%. Inform your Customer Success manager if you wish to modify this level and specify the desired extent.

Once the model is established and trained, Custom People recognition will be applied to all newly uploaded videos. Existing videos require re-uploading to undergo processing by this algorithm.

When a person is identified in a video with the confidence level defined for your account, the name provided during training along with the photo set will be applied as a People category tag to the video.

This modification solely impacts videos within your account, regardless of the presence of the identified person in other accounts’ videos.

For further changes in the model application, like changing the confidence level or providing additional training images to refine the results, please contact your Customer Success manager. The changes will be applied to the newly uploaded videos only, and existing videos will have to be re-uploaded.