In the evening I pushed version 2.0.0 of tumblelog to GitHub. This version renders text after an image as a caption. Or, the following Markdown:
![Cat](cat.jpg) Photo of a cat resting.
Is rendered as HTML as follows:
<figure> <img alt="Cat" src="cat.jpg" /> <figcaption> Photo of a cat resting. </figcaption> </figure>
This allows for styling using CSS of both the image and the caption. You can see examples on this blog.