Triggersift

TriggerSift lets you keep the ASMR sounds you love and avoid the ones you can't stand. ASMR videos mix many sounds. One viewer loves tapping but hates mouth sounds; another relaxes to crinkling until a mic bump ruins it. Today you can either sit through those moments or leave the video. TriggerSift listens to a video and marks the sounds it finds on a timeline, using 15 sound classes such as sharp tapping, crinkling, mouth sounds and unwanted noises. You mark what you love and what you want to avoid. A long section you dislike is skipped. A short sound inside something you love is turned down in the audio while the rest keeps playing. It is a local web app with an open-source repo. I hand-labelled 766 clips from 27 videos by what you hear, not by which object makes the sound. In the demo, a mic bump's peak drops by about 15 dB in the first example and about 8 dB in the second, while the surrounding taps stay about the same. Detection is still imperfect: it can miss sounds or flag the wrong section, so improving it is the main next step.