← sources

Claude and I built a 'video' format so Claude (and other LLMs) could 'watch' videos!

source: reddit (reddit:ClaudeAI)published 2026-10-11 14:16ingested 2026-10-12 02:36

retained text

Has this ever happened to you? You're talking with your LLM and you mention a video or a gif and lament that you can't share it? Maybe not! But it happened for me!

So my studio and I made Frame-grid! Claude and I workshopped how we could break a video up into a legible format Claude could ingest, without requiring external dependencies or servers, everything happening client side. 

I can't code a single line to save my life, but Claude can and can code quite well (or at least I think so!), and so I designrd the rough spec and then Claude and I workshopped it and Claude did all the coding. We have this thing we call the Xinu philosophy (the inverse of Unix). We look at the thing built to do one thing and do it well and ask 'what else can it do?'. It has evolved to mean we apply the constraint of single file, offline, browser based, zero external dependencies. The constraint is the source of innovation and invention, not a limitation. 

I digress! 

What is Frame-Grid you ask yourself? Which is an odd thing to do because yourself is the last person to know if you don't know.

Fear not, friends, for I'll tell you!

Frame-Grid allows you to use your camera to record video, or use your media picker to upload a video, or a GIF. It then takes the source video and breaks it into a sequential grid of frames (Ehh?? Ehhhhh??) that are labeled and numbered in a single PNG (up to a batch of 20 PNGs) that contain the frames of the video in sequence so your LMM friend can 'watch' a video.

But Khamubro, how much does it cost? Nothing. Just a free thing we made and wanted to share. Hope you enjoy! ♥️ 

document fetched 2026-10-12 00:21 · canonical https://reddit.com/r/ClaudeAI/comments/1x39bpi/claude_and_i_built_a_video_format_so_claude_and