You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
Right now, what I understand when we want to analyze video, llm vision is take 5 (default) frame with most movement in the video. It seems that it been fed to the llm vision based on which frame with most movement first. So the video lost it context.
Describe the solution you'd like
Change this behaviour to pick most movement first, but when fed to be analyzed, do with chronological order from these 5 frame
Describe alternatives you've considered
Additional context
I deduce that because most of times the result is indicating these behaviour of feeding the data.
Correct me if I'm wrong tho. Thanks dev valentinfrich
The text was updated successfully, but these errors were encountered:
Is your feature request related to a problem? Please describe.
Right now, what I understand when we want to analyze video, llm vision is take 5 (default) frame with most movement in the video. It seems that it been fed to the llm vision based on which frame with most movement first. So the video lost it context.
Describe the solution you'd like
Change this behaviour to pick most movement first, but when fed to be analyzed, do with chronological order from these 5 frame
Describe alternatives you've considered
Additional context
I deduce that because most of times the result is indicating these behaviour of feeding the data.
Correct me if I'm wrong tho. Thanks dev valentinfrich
The text was updated successfully, but these errors were encountered: