Reading Was Never the Whole Story: Listening Just Landed in Manabit.ai

For a while now, manabit.ai has been mostly a reading app. You open a text, you read at your level, unknown words get glossed, and the ones you're shaky on come back later through spaced repetition. That part works, and I use it myself every day.

Reading and listening are not the same skill

But there's a gap in extensive reading that's easy to ignore until you sit for the JLPT listening section, or try to follow a conversation with a friend, and it's this: reading comprehension and listening comprehension are not the same skill. You can read a passage smoothly and still lose the thread the moment someone says the same sentence out loud at a normal pace. I hit this wall myself, more than once, on the way to N2.

What shipped this week

So the update I shipped this week is simple on paper: every text in your reading session can now also be listened to. Same content, same vocabulary you're already working through, but read aloud at a natural pace instead of the slowed-down, over-enunciated speed a lot of learning audio uses.

The idea isn't new. Extensive listening alongside extensive reading is a known method, and plenty of people already do it manually, pairing graded readers with audio found elsewhere, or shadowing recordings. What I wanted was to remove the friction of hunting down matching audio for whatever you happen to be reading. If the text is already in your queue, the audio is right there next to it.

How I'd actually use it

How I'd actually use it, and how I've been using it while testing: read the text once normally, the way you always do. Then play the audio, either following along with the text or just listening with your eyes off the screen. The second pass is where it gets uncomfortable in a useful way — you'll notice how much slower your ear is than your eye, especially with anything above N4-N3 level. That gap is exactly what natural-speed listening is supposed to close over time.

A couple of honest caveats

A couple of honest caveats, because I'd rather say this now than have you find out later. This is a first version. The audio is generated, not recorded by a voice actor, so it's consistent but it doesn't have the full range of intonation a native speaker doing a dramatic reading would bring. Pacing is natural, but "natural" for a text-to-speech voice still isn't identical to a real conversation with pauses, filler words and interruptions — that's a different kind of listening practice, and one I'm not trying to replace. And right now it works text by text, it's not bundled into a "listening mode" for your whole session yet — that's on the list, not built.

Where to start

If you've been using manabit.ai for the reading and spaced repetition side, the listening feature is already live on your existing texts, no separate section to go hunt for. If you're specifically JLPT-prep-focused, pairing today's reading queue with the audio version is probably the fastest way to feel the difference — I noticed it within a few days of testing it on myself.