What is sayit's conversation practice, and what isn't it?

sayit's conversation practice is a spoken roleplay: the tutor opens a scene and offers you suggested lines to say, you record yourself speaking one of them, and it's scored the same way any read-aloud line is — word by word, against what you were actually saying. It is not an open-ended chat where you can say anything and get a free-form reply; the value is in the speaking and scoring, not in the breadth of what you can say.

30-second version: A conversation session presents a scene, gives you a line or a short set of suggested lines to choose from, and you speak one aloud. Your turn gets scored in reference mode against exactly that line — per-word verdicts and a match score, the same pipeline as any other take — and the dialogue advances. Typing is available as a no-mic fallback, but the exercise is built around speaking. At the end you get a recap: your average match, your best line, and the words worth practicing again.

Why suggested lines instead of a genuinely open conversation?

Because scoring pronunciation requires knowing what you were trying to say. An open-ended conversation, where you could say literally anything, has no fixed target to score your sounds against — that's the same reason freeform mode can measure fluency and accuracy but not completeness. Conversation practice keeps the roleplay format — a scene, a context, a natural back-and-forth — while keeping each of your turns anchored to a specific line, so every turn you actually speak can be scored the same rigorous way a read-aloud passage is.

What does a scored turn actually look like?

The same as any other take, just shorter: your spoken line comes back tinted word by word — clear, unclear, or check — with a match score against the suggested line you chose to say. That's real per-sound scoring, not a simple "did you say approximately the right thing" pass; the phoneme-level comparison behind every other feature in sayit is what's running here too.

What if I don't want to use my microphone for a turn?

Typing is available as a fallback so you can skip past a turn without a mic — useful if you're somewhere you can't record, or just want to move through the scene without stopping to speak every line. It's explicitly a fallback, though: skipping a turn by typing means that turn isn't scored on pronunciation, since there's no recording to score.

What do I get at the end of a session?

A recap covering your average match across the scene, your best-scoring line, and a short list of words worth practicing again — turning the session into something you can act on afterward, not just a transcript of what happened. Words that showed up as weak spots are exactly the kind of material flashcards can pick up automatically for later review.

What conversation practice is, and isn't

Why practice this way instead of just reading more passages?

Because a scene gives spoken practice a shape passages don't: a role, a context, and a sense of a real back-and-forth, even though each individual turn is scripted rather than freely generated. It's a middle ground between reading a flat passage aloud and having a genuinely unscripted conversation — structured enough to score precisely, dynamic enough to feel like more than a list of sentences.

Try it

Open Conversation in sayit and work through a scene — speak a suggested line, see the word-by-word score come back, and follow the dialogue to the recap at the end.