I'm not affiliated with and have no interest in Recast, just think it's a useful product for some LessWrong readers.

I recently discovered Recast, an app and Chrome extension that I highly recommend. Recast uses text-to-speech and LLMs to convert articles into conversational podcasts, making them much more engaging and easier to consume than regular text-to-speech software.

I have always enjoyed listening to some interesting stuff while walking the dog, cooking dinner, or driving to work. Unfortunately, robotic voices reading an article has never done it for me, as the monotone and robotic voices are unable to keep my attention. Turns out, that if instead of having one robotic voice reading an article out loud, you have two robotic voices discussing an article, it's somehow much easier to keep paying attention!

I think the effect is best explained by giving it a try yourself, so here is a somewhat random collection of automatically generated dialogues about LessWrong posts:


Of course, a dialogue summary is not a replacement for reading the actual posts, but for me, it's a great way to determine whether I want to read the whole thing. Also, keep in mind that this is LLMs we're talking about, so sometimes the 'hosts' might hallucinate content that is actually not in the article. Unfortunately, Recast doesn't support very long articles right now, which probably has to do with the context length. Finally, it is free to use and I'm not sure what their business model is, which is generally a cause for concern.

New Comment