The algorithm can't tell because these experiments are run with literally empty videos/audios on a blank channel... You're feeding the algorithm an abnormal inp

The algorithm can't tell because these experiments are run with literally empty videos/audios on a blank channel... You're feeding the algorithm an abnormal input, with no data points to detect anything. Normally, it would look at the video, audio and channel status and so on. In this case, it's not getting any of that. The model wasn't built to run on edge cases like this and it isn't expected to be able to classify such an outlier input.