Multiple category text prompt inferencing on video

Is there a way to add multiple text prompts (i.e. ['dog', 'cat']) for frame 0 and have the model detect + track any instances of both of these categories.

My question is specifically for the video inferencing SAM3 (with tracking), and not the single image segmentation.

I saw that many were pointed towards the example with the single image model, but that is not what I mean.

Thanks!

Start with the SAM3 video inference and tracking entry point, then compare it with the existing single-image example referenced in the issue. Trace how text prompts are supplied for frame 0; done means determining whether multiple categories can be supported and documenting or implementing the required video behavior with coverage for both requested categories.

Written by the indexing model from the issue text.