I had a go at this using the on-device models in edge and chrome, phi4-mini and gemini nano, worked surprisingly well for such small models.

https://ryanseddon.com/ai/how-to-build-an-agent-on-device/

I built https://ffprompt.ryanseddon.com using the chrome ai (Gemini nano). Allows you to do ffmpeg operations on videos using natural language all client side.

What are the prerequisites for this? I keep getting "Bummer, looks like your device doesn't support Chrome AI" on macOS 15.2 Chrome 132.0.6834.84 (Official Build) (arm64)

[Edit] Found it. I had to enable chrome://flags/#prompt-api-for-gemini-nano

Yeah the instructions are not clear. They're on the github repo[1] linked in the header.

1. Install Chrome Dev: Ensure you have version 127. [Download Chrome Dev](https://google.com/chrome/dev/).

2. Check that you’re on 127.0.6512.0 or above

3. Enable two flags: chrome://flags/#optimization-guide-on-device-model - BypassPerfRequirement chrome://flags/#prompt-api-for-gemini-nano - Enabled

4. Relaunch Chrome

5. Navigate to chrome://components

6. Check that Optimization Guide On Device Model is downloading or force download if not Might take a few minutes for this component to even appear

7. Open dev tools and type (await ai.languageModel.capabilities()).available, should return "readily" when all good

[1]: https://github.com/ryanseddon/FFprompt

Ah bummer I've been posting my earthin24 timelapses[1] to this for quite a while now.

[1] https://botsin.space/@earthin24

In my home state of Victoria Australia the government had a program to give out these powerpal[1] units for free that could measure your usage in realtime using the flashing led on our smart meters, we also require all energy grid operators (the people who own the poles and wires) to have an energy portal where users can get near realtime data to the nearest 30mins, soon to be 5 with some new legislation.

The former most people have no idea about but the powerpal has been a smashing success for consumers to understand what is using energy.

[1] https://www.powerpal.net/

I've got an oven and induction cooktop freestanding unit[1] that has big chunky knobs to change the induction power levels. Would never bother with gas again.

[1] https://www.fisherpaykel.com/au/cooking/freestanding-cookers...

I did something similar using filter_complex to create a 14x14 grid showing 196 days of earth full disc shots for my earthin24 Twitter bot. It's truly impressive what ffmpeg can do https://ryanseddon.com/javascript/an-earth-mosaic/

I combine this and 3 other satellites on my twitter bot into daily videos if you're interested.

https://twitter.com/earthin24

Cool! Got any multi-day/week/month videos?

I have an original copy of this game in my garage! Including the vmu microphone.

Hopefully the next step is to take image decoding off the main thread.

I think multi-threaded (and thus off main thread) image decoding landed on Aurora 22, and I assume made it into that release.

According to a few people both IE11[1] and Firefox[2] disable hit testing on scroll if you don't move your cursor while scrolling.

[1] http://www.thecssninja.com/javascript/pointer-events-60fps/c... [2] http://www.thecssninja.com/javascript/pointer-events-60fps/c...