Not 'what it means' but 'how to say it' — a pocket pronunciation coach for 60+ languages

Here's the idea in one line: you type any word — a dish on a menu, a coworker's name, a brand — and the app breaks it into syllables, writes out how it sounds with the stress marked, and reads it aloud, slowly. And you stop freezing before saying it in front of people.

And here's what's new. The model could translate a word for ages — but that's "what it means." In late June, OpenAI added pronunciation to ChatGPT: ask how to say a word and you get a written explanation plus audio, across 60+ languages, free, right in the chat. So the model now reliably has the thing this project needed: "how it sounds." That's what the whole coach rests on.

Why this one

A translator covers "what it means." The awkward moment is the other one: saying it out loud. Ordering "gnocchi" without melting. Saying a new coworker's name without mangling it. Reading a French label instead of staying silent out of fear.

Picture the café. You point at the menu because you're scared to say the word wrong. But you could glance for three seconds — "NYOH-kee, stress on NYOH" — listen twice, and say it out loud, calmly. You'll use this yourself, regularly.

And there's less "magic" than it looks. The app is a pipe: it hands the word to a model to break down the pronunciation, then hands the result to a voice. All the difficulty lives in two clean requests.

What you'll learn

• A structured answer from the model. You'll ask not for "text" but for exact fields: syllables, the sounded-out spelling, a memory hook. That's structured output — the model answers in a shape the app can show easily.
• A text → sound chain. First the model breaks the word down, then a voice reads it aloud. You'll build a pipeline where one step's output is the next step's input.
• "The prompt is the feature." "How to pronounce" isn't a separate technology. It's an instruction: "break it into syllables, mark the stress, give a hook." A good prompt is your main feature.

A ready starter prompt

Don't ask the agent to "make a pronunciation app" — it'll guess the format. Give it the fields and two steps:

A strong prompt leaves no room for guessing: you can see exactly the fields you need, the two steps, and the ban on inventing words that don't exist. The first result lands closer to what you wanted.

What you end up with

You're at a café, the menu says "bruschetta." You open the app, type the word. On screen, large: "broo-SKEH-tah", a button under it — you tap, hear it slowly, again. You look up at the waiter and say it out loud, calmly. He nods. A small win that, a minute ago, was blocked only by the fear of saying it wrong.

One bit of honesty to close: the model isn't a strict examiner — it won't grade the fine details of your accent. But as a confidence trainer, so you don't go silent out of fear, it works great.

Source: OpenAI — ChatGPT release notes: text and audio pronunciation in 60+ languages