Auto Generated Text: A dark-themed workflow diagram illustrating an automation sequence starting with a Webhook trigger on the left. The process splits into two main paths: one connects to an AI Agent using the Qwen 3.5:9b model, and the other processes EXIF metadata through a JavaScript code block. These streams merge together, undergo manual field editing, and are then written to Obsidian via a POST request. A separate error trigger path leads to a message creation node at the bottom right.

If Bluesky taught me anything, it’s that alt text is necessary upon threat of death and bullying. I get it. I think it’s really important for our photographs to be able to be seen by everyone.

However, it is a bit onerous to type out alt text for every single photograph you publish. Especially if you’re a photographer.

So I’ve created a small agent that creates alt text.

Let’s walk through it.

First , I record a voice memo on my phone and send it through an Apple shortcut to a webhook on my N8N server.

Auto Generated Text: A menu interface displaying four circular icons on a dark background. From left to right: An icon of two overlapping documents labeled "Copy Photo," followed by a stylized robot head icon labeled "Alt Image Robot," then a folder icon with a person silhouette labeled "Add to Shared Album," and finally a folder icon with a plus sign labeled "Add to Album."

On that server, it instantly does two things.

  1. 1.

    It takes that file and runs it through Qwen 3.5:9b to analyze and ascertain alt text. 

  2. 2.

    Reads the EXIF metadata, so I can get photograph information.

I merge those two together, prepare the payload, and then append to a note in Obsidian - so I have a central place where all these alt texts go.

It’s super useful and super quick.

I can grab ten photos from my photo roll, stick them all into this, and they’ll all show up into Obsidian.

What I also like about it is I don’t have to automatically post. I’m just generating alt text, being ready for anything that I want to do with it. Some examples below.

freestone's avatar
1d

A close-up photo of a hand holding a white sign that reads "STOP THE MADNESS" in red, black, and pink handwritten text. In the blurred background, American flags are visible.
freestone's avatar
1d

That took 10 seconds with self hosted qwen3.5:9

A view down a narrow, paved residential alleyway or side street. On the left, a tall red brick wall features white-framed windows protected by black wrought-iron grilles with scroll designs. The wall transitions to a tan section with climbing ivy further back. To the right, the asphalt road has a white painted edge line, bordered by green trees and another red brick building in the distance under a blue sky with clouds and utility lines.
freestone's avatar
1d

It’s rad:

A wide river flows through a valley, featuring clear shallow water with visible rocks on the left bank and a dry, stony shoreline on the right. The left side of the riverbank rises steeply into a hill covered in green grass and evergreen trees, while distant mountains are visible in the background under a blue sky.