How could one hour of video fit in 1M tokens? 1 hour at 30fps is 3600\*30=100k f... | Hacker News

Hacker News new | past | comments | ask | show | jobs | submit

login

p1esk on Feb 16, 2024 | parent | context | favorite | on: Our next-generation model: Gemini 1.5

How could one hour of video fit in 1M tokens? 1 hour at 30fps is 3600*30=100k frames. Each frame is converted in 256 tokens. So either they are not processing each frame, or each frame is converted into fewer tokens.

KTibow on Feb 16, 2024 [–]

The model can probably perform fine at 1 frame per second (3600*256=921600 tokens), and they could probably use some sort of compression.

Join us for AI Startup School this June 16-17 in San Francisco!
Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact