Ask HN: How would you represent a downvoted comment in text-to-speech?

This is meta, but might still be interesting to think about.

On screen, a comment with significant downvotes has a different color. If it has enough downvotes, the contrast is such that the comment is harder to read; presumably that's the point. But when reading HN with text-to-speech (e.g. a screen reader for the blind), this information is lost. Sure, HN could be modified to add the word "downvoted" or the like such that only a screen reader can see it. But that doesn't have quite the same effect as changing the color. So, if you could solve this problem by modifying HN and/or the screen reader, how would you do it? Maybe add a background noise to the reading of the comment, and make the noise louder based on the number of downvotes?