How I Write Alt Text

[aleteoryx]

These are obviously not rules, but I see a lot of alt-text that misses the mark in one way or another, and I try to follow these general guidelines when writing alt text.

Pronunciation

Remember, alt text will not be parsed by human eyes, it will be spoken by whatever TTS the viewer's screen reader uses. I see a lot of criticism of DaShareZone's alt-text, but if you actually run it through TTS, it sounds almost normal. "WHUT R U DEW EENG" and "What are you doing?" sound almost the same when run through TTS, and the main difference is in annunciation.

Consider your username. TTS pronounces mine as "ay l tier eeks", but a closer pronunciation can be gotten with "uh lee tee ore ikhs". I recently wrote alt text for a screenshot in which a username was "sexhaver". The TTS I test with pronounces that as "seksaver", instead of "sex haver". Screen readers are camel-case aware, so I generally just change any strings like that to a camelcased form, e.g. "sexHaver"

Structure

Mastodon provides in-built OCR for image uploads, to make alt text easier to add. This is not a perfect function. If you're uploading, for example, a variation on the "Is this a pigeon?" meme, and replacing the butterfly with an image, the generated alt text may just be "Is this a XYZ?"

My general strategy is to start with a broad description of the image context, followed by each important element. The opening might look like:

• A screenshot of the Misskey admin panel.
• A photo of a field covered in snow.
• An edit of the drake meme format.

Then, each element might look like:

• The federation graph has a large deadzone in the middle of it, lasting multiple days.
• There is a trail of small footprints across the canvas.
• The top thing, which drake is opposed to, is the Line Webtoon logo.

Feel free to omit stuff! If a screenshot includes tens of things, but your focus is on only one, you can just describe the bit that matters. Alt text is not about perfectly describing an image so that someone can see it in its entirety in their mind's eye, it's about getting the point of the image across. Especially in the case of a joke post, terse alt text is critical to making the punchline land.

Context

I generally try and provide extra context when putting alt text into images. Many fediverse frontends provide a way to view alt text as a sighted individual, and I find that nice for making my screenshots simple. I don't need to have a lot of extra visual data, I can just crop to a small area and introduce the alt text with "it's this Minecraft modpack", "it's this website", etc. This is the main curb-cut effect from alt text, and I am a huge fan of it. There are people far more technically talented than me on the internet, and it gives me just that much more of a foothold to try and understand out what they're doing.

Anyways...

That's about it. I just wanted to share that, cause I see pretty bad alt text a lot, esp for memes. Also, it doesn't need to be nearly as structured as the above for most cases. You can always just do something like "A photo of my new black tank top.", for simpler images and that will suffice. Structure is only necessary for complicated screenshots and shit.

That's all from me! Toodles!

Comments

[blimix]

Thanks! I'll try to remember about omitting the unimportant parts. I'm not sure how to judge that when I post a nature photo. I do work at brevity.