News for hackers. Stuff that matters.

A comment on The killer app of Gemini Pro 1.5 is video. Parent comment

Re:The killer app of Gemini Pro 1.5 is video by ankeshanand on Wednesday February 21, 2024 @03:00PM (#39458704)
We've done extensive comparisons against GPT-4V for video inputs in our technical report: https://storage.googleapis.com/deepmind-media/gemini/gemini_....

Most notably, at 1FPS the GPT-4V API errors out around 3-4 mins, while 1.5 Pro supports upto an hour of video inputs.

[ Parent | on HN ]
Re:The killer app of Gemini Pro 1.5 is video by jxy on Wednesday February 21, 2024 @03:23PM (#39458990)
So that 3-4 mins at 1FPS means you are using about 500 to 700 tokens per image, which means you are using `detail: high` with something like 1080p to feed to gpt-4-vision-preview (unless you have another private endpoint).

The gemini 1.5 pro uses about 258 tokens per frame (2.8M tokens for 10856 frames).

Are those comparable?

[ Parent | on HN ]
Re:The killer app of Gemini Pro 1.5 is video by verticalscaler on Wednesday February 21, 2024 @04:13PM (#39459628)
The average shot length in modern movies is between 4 and 16 seconds and around 1 minute for a scene.
[ Parent | on HN ]
Re:The killer app of Gemini Pro 1.5 is video by moralestapia on Wednesday February 21, 2024 @04:18PM (#39459688)
>while 1.5 Pro supports upto an hour of video inputs

At what price, tho?

[ Parent | on HN ]