I tried testing this with:

llm -m gemini-3-pro-preview -a 'https://www.youtube.com/watch?v=qgJ7x7R6gy0' 'Output a Markdown transcript of this meeting. Include speaker names and timestamps. Start with an outline of the key meeting sections, each with a title and summary and timestamp and list of participating names. Note in bold if anyone raised their voices, interrupted each other or had disagreements. Then follow with the full transcript.'

From https://simonwillison.net/2025/Nov/18/gemini-3/#analyzing-a-city-council-meeting

It said:

Error: Please use fewer than 10800 images in your request to this model

So I tried it with a shorter video instead:

llm -m gemini-3-pro-preview -a 'https://www.youtube.com/watch?v=9o1_DL9uNlM' 'Produce a summary with relevant URLs and code example snippets, then an accurate transcript with timestamps.'

This worked! https://gist.github.com/simonw/1b07aafb2bfc112b180ab68c864511cb

Weird that it logged it as text/html; charset=utf-8: https://www.youtube.com/watch?v=9o1_DL9uNlM though.

Unlike in the longer example the timestamps from the transcript appear to match the correct points in the video.