AI for the wine cellar?

AI slop is everywhere. We’re all annoyed by it. But there are use cases where AI actually helps.

To organize my wine cellar, I want to make the inventory searchable. But first I need the data: the wine’s name, grape varieties, the winemaker, and so on.

The simple solution: I send two photos of the wine to the OpenAI API, together with a matching JSON template. OpenAI’s AI not only reads the text from the images, it also assigns the data in a surprisingly reliable and well-structured way:

Here are example images I used to have the AI extract the wine’s data. As you can see, they are quick snapshots - a little blurry, with text that isn’t particularly easy to read.

Even so, this is already sufficient for the AI to produce a good, well-structured result. Very little manual cleanup is needed.

This way, I was able to catalog almost all the bottles in my wine cellar with virtually no manual effort.

As you can see in the example result image: All the data ist correct, even the dosage. And I did not change a single char, it's all the original data extracted by gpt5.

So yes - there are genuinely good AI applications after all.