This message was deleted.

Hi @Joy Arulraj, video is not yet a default IO descriptor but a File is. Using the File IO descriptor, you can get the input in the form of

io.BytesIO

. You can load in the bytes with a library that handles video decoding based on your choice.

Thanks for the helpful suggestion, @Sean. We will look into the File I/O descriptor. I have a follow-up question that is somewhat open-ended. Could you share your thoughts on whether combining multiple Bento-served models in a SQL-like query shown above (i.e.,

FaceDetector

,

EmotionDetector

) will address any challenges Bento users currently face? Thanks for taking the time to think about this.

I think declarative model inference using SQL fills an important skill gap of users who knows SQL but not any programming language. Reminded me of Predibase.

Great, thanks for that helpful pointer.