Any way to prompt the video model without an object id?

Hi is there any intention to add the ability (or any way to do it currently) to prompt the model without an object id?

In instances where masklet tracking fails, it would be useful to be able to prompt without an object-id and have the model figure out which masklet the prompted mask belongs to. This way you can use an object detector algorithm to hone in on subjects and tell sam to only focus in certain areas when looking to extend existing masklets.