After Duolingo: A 90-Day Plan to Start Speaking

Finished your Duolingo course, or finished with it? Either way, Duolingo did its job, and its job was never fluency. You now have somewhere between A1 and B1 worth of recognition: you can read sentences, match words, and survive multiple-choice grammar. What you almost certainly cannot do is hold a ten-minute conversation, because the app never made you do what conversation requires: produce language out loud, under mild time pressure, about something you didn't choose.

That's not a scandal. It's a hand-off. This post is the bridge plan for the 90 days after the green owl.

What did Duolingo give you?

More than the internet's cynics admit: a daily habit (the hardest part of language learning, genuinely solved), a few thousand words of passive vocabulary, core grammar pattern-recognition, and proof you can stick with something for months. Keep all of it.

What it didn't give you is active vocabulary. Research on the passive-active gap (Laufer 1998) shows productive vocabulary lags far behind recognition. Intermediate learners typically understand roughly twice as many words as they can deploy in speech. App-only learners sit at the extreme end of that gap: hundreds of hours of recognition reps, near zero production reps. You're not "bad at speaking." You've literally never trained it.

What are your options now? (Comparison)

The four realistic next steps, with real trade-offs. Prices are typical ranges as of June 2026; all of them move you forward. The question is cost, friction, and whether speaking happens.

Option	Typical cost	What it's great at	The catch
Human tutor (italki, Preply…)	~$10–30/hour, community tutors from ~$5	Real human reactions, cultural nuance, accountability	$200–600 for a 90-day daily habit; scheduling friction kills most streaks
Language exchange (Tandem, local meetups)	Free	Real people, real stakes, possible friendships	You pay in time teaching your language; partners flake; beginners get little patience
AI conversation partner (Dia and similar)	Free tier; ~$8/month for unlimited chat	Speaking happens daily with zero scheduling, zero ego cost, infinite patience	No social stakes, no real-world noise; an AI is not a culture
Another app / more Duolingo	$0–10/month	Comfort, streak preservation	More recognition practice cannot fix a production gap. It's the same rep you've already mastered

The trap is the last row. The most common post-Duolingo move is more input (a new app, a podcast, "one more course") because input feels safe and speaking feels embarrassing. Input got you here; only output gets you out (Swain's output hypothesis is the research version of this point; we've written about why speaking beats listening).

The 90-day bridge: from recognition to conversation

One rule drives all three phases: every day includes out-loud production. Twenty minutes is enough.

Days 1–30: activate what you already know

This month, you're not learning new material. You're converting passive vocabulary to active.

• Daily 10-minute conversation (tutor, exchange partner, or AI, whatever you'll actually do daily). Topic: your real life. Yesterday, your job, your annoying neighbor.
• After each conversation, write down the three things you couldn't say. Look them up. Say each one aloud five times.
• Expect to feel like you've regressed. You haven't. Recognition skills just don't transfer one-to-one, and feeling that gap is the diagnosis working.

Days 31–60: stretch the turns

• Target 30-second answers instead of one-liners. If your partner asks "how was your weekend," the answer is a story, not "good."
• Add one domain you genuinely care about (your work, your hobby) and load its vocabulary. The plateau post explains why one deep domain beats ten shallow topics.
• Keep a 5-minute Duolingo session if you like the streak, as dessert, not dinner.

Days 61–90: make it unpredictable

• Ask your partner to interrogate you: follow-up questions, disagreements, "why?" three times in a row.
• Once a week, do something at full speed: a TV episode without subtitles, a voice call at native pace. (Why native speed matters)
• Measure: record a 2-minute monologue on day 61 and day 90. The difference will be audible. That's your before/after, no app gamification required.

By day 90 you won't be fluent; nobody promises that (the hour counts look like this). You will be conversational on familiar ground, which is the inflection point where the language starts feeding itself: conversations create vocabulary needs, needs get filled, fills get used.

Where Dia fits (and where it doesn't)

We built Dia for exactly this hand-off: daily speaking practice with an AI partner that adapts to your level, recycles the words you've used (the word bank auto-captures them from your conversations), and does voice calls at native speed in the dialect you choose. The free tier gives you daily messages to test the habit; the 90-Day Fluency Challenge is a one-time purchase built around exactly the arc above, with no subscription or auto-renew.

Where it doesn't fit: if what you need is a human, for cultural texture, exam prep, or the accountability of a person expecting you, book the tutor. The comparison table above is real; for some learners the $20/hour is the right spend. The only wrong move is the comfortable one: another six months of silent recognition practice.