Claude Talks Too Much? Let It Paint the Big Picture Instead
Claude can be incredibly verbose. Whether it's explaining something or delivering a comprehensive result, you get a lot of fluff, but also very useful information you don't want to miss.
But the more verbose it gets, the harder it is to grasp the big picture at a glance.
I've written before about strategies to make Claude more concise. This time I took a whole different strategy: let Claude stay verbose, and use image generation to paint the big picture as an infographic, rendered right inside the chat, for a few pennies.
Here is what that looks like in my Claude Code chat:

Notice: the picture sits directly under the reply, with "View larger" and "Show prompt" buttons and the path where the file was saved. No detour, no separate window.
The funnel
Think about it as a funnel. You have a massive document, or even an entire repository, and you want an executive summary of how it works. You could ask Claude to write that summary.
Or, more powerful, and what we're doing here: you let Claude create a description (a prompt) for an infographic, and have it painted on the fly by a very cheap yet very capable image generator, Nano Banana Lite (openrouter.ai).

Notice: nothing gets summarized away. The verbose answer stays, the funnel just adds the map on top of it.
The wide end keeps everything Claude already does well: reading it all, keeping every detail. The narrow end is one picture you grasp in two seconds.
The ingredients
- macOS (the mod leans on the Keychain,
sipsand Preview) /usr/bin/python3, the one every Mac already ships with- the Claude Code desktop app with mods, the same mods I used to fix every UX problem I had with Claude Code
- an OpenRouter account with Nano Banana Lite enabled, model ID
google/gemini-3.1-flash-lite-image(openrouter.ai) - your OpenRouter key, stored once with
security add-generic-password -s openrouter -a "$USER" -w
What it costs: OpenRouter lists Nano Banana Lite at $30 per million image-output tokens, $0.25 per million input tokens and $1.50 per million text-output tokens (checked 21 Nov 2026, openrouter.ai). In practice, the infographic further down this post measured $0.03365975.
About 3.4 cents, so a dollar paints roughly 30 infographics.
How it works
Here is the part I like most: the AI makes no tool call. It writes one ordinary line of text in its reply, a marker: