FlaevoGet the app
← All posts
Flae, the Flaevo study companion

Talk to Flae: How Flaevo's Live Voice Tutor Works

Flaevo's AI tutor has always been something you type to. Now you can talk to it.

Open the AI tutor, tap the wave button beside the message box, and you are in a conversation. You speak, Flae answers out loud, and you can cut in whenever you like — the way you would with a study partner sitting next to you.

Here is how it actually works, and how to get something out of it beyond the novelty.

Why talking is worth doing at all

This is not a gimmick bolted onto a chat box. Speaking changes what your brain has to do.

When you type a question, you can be vague and still get a useful answer. When you say it out loud, you have to commit to words in real time — and the moment you try to explain something aloud and stumble, you have found a gap you did not know was there. That stumble is the point. It is the same principle behind active recall: retrieving an idea is what strengthens it, and re-reading a page is not retrieval.

Voice also unlocks study time you cannot type in. Walking home. Waiting for a trotro. Lying down with your eyes closed after a long day of reading.

What happens when you speak

A single turn has four steps, and knowing them explains most of what you will notice.

1. She waits for you to finish. Flae listens until you have stopped talking for about a second and a half. That pause is deliberate — cut it shorter and she interrupts you every time you pause to think. If you would rather end your turn immediately, tap the screen and she starts working straight away.

2. Your words become text. What you said is turned into text, which is also what makes the conversation appear in your normal chat history afterwards.

3. She starts answering before she has finished thinking. This is the part that makes it feel like a conversation rather than a recording. Flae does not write out a whole answer and then read it. She reads the first sentence as soon as it exists, while the rest is still being written, and the sentences after it are being prepared while she is still saying the first one. Waiting for a complete answer before speaking would add several seconds of silence to every single turn.

4. You can stop her at any point. Just start talking. She stops mid-sentence, drops everything she was about to say, and listens. Tapping the screen does the same thing. This matters more than it sounds: without it, a wrong turn costs you the whole answer before you can redirect her.

Why her spoken answers are so short

If you have used the typed tutor, you will notice Flae answers very differently out loud — usually one or two sentences, rarely more.

That is on purpose, for two reasons.

You cannot scroll back through speech. A written answer can have headings, a worked example, and three bullet points, because your eye can go back over it. A spoken answer has to be short enough to hold in your head the first time, because there is nothing to re-read.

Short answers keep the conversation moving. Instead of a lecture, you get one step, and then it is your turn again. If you want more, ask for more — "go deeper", "give me an example", "say that simpler". A tutor who talks for two minutes without stopping is not tutoring you, they are broadcasting at you.

When she looks something up

Most questions do not need the internet — the answer to "explain photosynthesis" has not changed this week. But some do.

If you ask something that genuinely depends on current information, Flae will tell you she is checking, look it up, and answer from what she finds. She will mention where it came from in passing, but she will not read out links or citation numbers, because hearing "bracket one" read aloud is unbearable. The sources appear as tappable links in your chat instead, so you can check them after the call.

You control this in Settings → Web search, which has three positions: When needed (the default — she decides), Always, or Never.

Everything ends up in your chat

Nothing said in a voice conversation is lost. Every question you asked and every answer she gave is written into the same tutor chat you would have got by typing — with the full formatting, working, and any sources.

So the pattern that works well is: talk through it, then read it back. Have the conversation to find out what you do not understand, then scroll the chat afterwards for the version with the actual working written out.

Your screen also stays awake for the whole call, so it will not lock on you mid-answer.

What it is not good for

Being honest about this saves you frustration:

  • Anything with a diagram, a table, or long working. Maths with several lines of algebra is genuinely worse out loud. Flae will tell you when something is easier to see, and it will be waiting in the chat.
  • Somewhere completely silent. Voice mode is not discreet. In a quiet library, type.
  • Offline. Speaking to Flae needs a connection. Your generated course podcasts do not — those play offline once made.
  • A very noisy room. In heavy background noise she may struggle to tell when you have stopped speaking. Tap the screen to end your turn manually.

There is also a short wait before her first word. We are still working on that one, and we would rather say so than pretend it is not there.

How to get the most out of it

  • Ask one thing at a time. Three questions in one breath gets you a scattered answer.
  • Make her ask you. Say "quiz me on this topic, one question at a time" and answer out loud. That is proper retrieval practice, not passive listening.
  • Explain it back to her. Say "let me try explaining this, tell me what I got wrong." Teaching something is the fastest way to find out whether you actually know it.
  • Interrupt freely. The moment she goes somewhere unhelpful, talk over her. It is designed for that.
  • Use earphones. Everything works better on earphones, especially interrupting her, because the microphone is not fighting the speaker.

The bottom line

Typing is better for anything you need written down. Talking is better for working out whether you actually understand something — because you cannot fake fluency out loud, and the moment you hesitate, you have found the thing to study next.

Use both. Talk to find the gaps, then read the chat to fill them.


Want to try it? Open the AI tutor in Flaevo, tap the wave beside the message box, and ask Flae to quiz you on whatever you studied today.

Flaevo is a free study app for Ghanaian students. Upload your notes, and Flae turns them into sessions, flashcards and quizzes you actually remember.

Get Flaevo free on Google Play

Frequently asked questions

How do I talk to Flae in Flaevo?

Open the AI tutor and tap the wave button beside the message box. That starts a live voice conversation: you speak, Flae answers out loud, and you can interrupt her at any point by talking or tapping the screen.

Why are Flae's spoken answers shorter than her typed ones?

Because you cannot scroll back through speech. A spoken answer has to be short enough to hold in your head first time, so Flae keeps it to one or two sentences and waits for you to ask for more. The full written version, with any working and sources, appears in your chat afterwards.

Can I interrupt Flae while she is talking?

Yes. Start speaking and she stops mid-sentence and listens, or tap the screen to do the same. This works best on earphones, since the microphone is not competing with the phone's speaker.

Does voice mode work offline?

No, speaking to Flae needs an internet connection. Course podcasts are the offline option — once an episode is generated you can replay it as many times as you like without a connection.

Can Flae look things up while we are talking?

Yes. If your question genuinely depends on current information, she tells you she is checking, looks it up, and mentions the source in passing. She does not read out links — the sources appear as tappable links in your chat. You can set this to When needed, Always, or Never under Settings.

Is the voice conversation saved?

Yes. Every question and answer is written into the same AI tutor chat you would have got by typing, with full formatting and any sources, so you can read it back after the call.