Speak vs Duolingo: price, languages, and who each one is for

The short answer

If you can't yet produce two unscripted sentences about your day, pick Duolingo, stay on the free tier, and come back in a few months. Conversation practice you can't hold up your end of is an expensive way to feel bad.

If you have vocabulary but freeze when a real person waits for you to talk, pick Speak. That gap is what it's built for, and Duolingo's free speaking exercises can't fix it because they put the sentence on screen before you say it.

And if you're learning Russian, German, Portuguese or most other languages, check the coverage lists below first. One of these two may not teach your language at all, and the other may not do the speaking part in it.

Now the detail, starting with the thing that decides this for more people than price does.

Check the language list before the price

Three pages currently rank for this comparison. They give Speak's language count as 5, 5 and 7. All three are wrong.

Speak's own homepage says it plainly: "English speakers can use Speak to learn Spanish, French, Korean, Japanese, Italian, and Simplified Chinese." That's 6, and Speak sells the reverse direction too, English for speakers of other languages. Read 21 September 2026.

So if you want German, Portuguese, Russian, Arabic or Hindi, Speak is not a candidate and the rest of this page is academic for you.

Duolingo's side has a subtler version of the same trap, and one ranking page gets close to it before landing wrong. Duolingo teaches 40-odd courses. Its open-ended AI conversation feature, Video Call, covers far fewer.

Duolingo's own announcement lists Video Call with Lily for "English, Spanish, French, German, Italian, and Portuguese, and iOS learners also have access to Video Call in Japanese and Korean." The newer Falstaff beginner version, announced 14 January 2026, covers "English, Spanish, Portuguese, French, Japanese, German, Italian, Korean, and Chinese on iOS."

So the real figure is 8 or 9 depending on the feature and your platform. A ranking page claiming Duolingo's AI speaking covers about 5 languages is underselling it by nearly half, and naming the wrong five.

Two more things about Video Call that change the price maths below. It's a Duolingo Max feature, not a Super one. And Duolingo's own wording puts several of those languages on iOS only.

What each one actually costs

Speak's website publishes no price anywhere. I looked, including at /pricing, which 404s. What it does publish is a 7-day free trial and a line that pricing "varies by region and promotion."

The US App Store listing does carry figures, so that's where these come from, read on 21 September 2026.

That last row is the honest and awkward one. Every page ranking for this query quotes a Duolingo Max price, usually "about $29.99/mo or $167.99/yr." I can't confirm either figure from a Duolingo-owned page, and duolingo.com/max carries no dollar amount at all. So I'm leaving it blank rather than laundering a third-party number into a comparison table.

The Super row deserves a flag too. Apple lists five price points and labels none of them with a billing period. A page that tells you Super Duolingo is "about $12.99/mo" has picked one of five numbers and attached a term the store never stated.

Prices move, regions differ, and Apple converts by store. Check your own before you buy, and treat every figure here as dated 21 September 2026 rather than permanent.

The stacking bill nobody adds up

Reviewers love recommending you run both. The Android Authority writer kept his 700-day Duolingo streak alive while praising Speak, and closed by asking whether to ditch the others without answering.

Fine as a personal choice. Worth pricing before you copy it.

Speak Premium at $83.99 a year plus Super Duolingo at its $83.99 App Store point is about $168 a year. Go Premium Plus instead and it's about $249. Neither figure includes Max, because Max has no published price to include.

The cheaper version of the same stack is Duolingo's free tier next to one paid speaking app. Duolingo's free tier is genuinely full-course, gated by energy rather than by content, so you're mostly paying Super to remove ads and energy limits. If your goal is speaking, that's the subscription to question first.

For scale: the State Department's own figures put a Category I language like Spanish or French at 552 to 690 class hours, on a week of 23 class hours plus 17 hours of homework. Against that, $168 a year is not the expensive part of learning a language. Your hours are. Which is the argument for spending the money on whatever you'll actually open daily, and not on a second app you'll check twice.

What the research reports actually say

One of the ranking pages links a Duolingo research report as evidence that Video Call produces measurable speaking gains, then never states the gain. I went and read it, and now I know why.

Duolingo's report, published 30 March 2026 by Audrey Kittredge and Xiangying Jiang of its Efficacy Research Lab, describes three studies. Japanese speakers learning intermediate English who used Video Call "at least twice a day" for a month "improved more on a standardized speaking test than those who didn't." English speakers learning early intermediate Spanish "showed greater improvement in speaking and reported higher confidence."

Direction, and nothing else. No sample size, no effect size, no test name, no confidence interval, no control-group detail beyond "those who stuck to regular lessons." The competing page couldn't quantify the claim because the source doesn't.

That isn't proof the feature fails. It's a first-party marketing report doing what those do, and the "at least twice a day" condition is worth noticing: that's a heavy usage bar, and heavy users of anything tend to improve more.

Speak's side is thinner still. Its homepage leads on an OpenAI partnership, backing from the OpenAI Startup Fund, and 15M+ downloads. The method is named, Learn then Practice then Apply, credited to "language education experts." There's no trial and no outcome figure. I found no published study of the shipping product with a proficiency outcome and a comparison group, which is the same hole most of this category has.

So when you choose here, you're choosing on mechanism and on price, and not on evidence. Anyone telling you otherwise is quoting a number their source didn't publish.

When conversation practice is worth paying for

This is the question all three ranking pages skip, and it's where the money actually gets wasted.

Nobody has published a threshold, so what follows is a rule of thumb rather than a finding. The anchor I'd use is Nation's 2006 coverage work: about 3,000 word families plus proper nouns gives over 95% coverage of speech, and 6,000 to 7,000 families gets you to 98%.

What 95% feels like is the part worth sitting with. Nation puts it at roughly 7 unknown words in every minute of speech at 150 words a minute. That's a conversation with a hole in it every 8 seconds, and that's the good end.

Below a few thousand words, an open-ended AI conversation is mostly you apologising. The app will be patient about it, and you'll still be paying $17.99 a month to be patient at.

The free test: try saying two sentences about your day out loud, right now, with nothing on screen. Slow and ugly counts as a pass. Nothing at all coming out means spend this month on vocabulary, on the free tier, and revisit in eight weeks. I wrote up what the coverage numbers mean in practice if you want the longer version.

Where each one genuinely wins

Duolingo wins on getting you to show up. The streak mechanics are the most effective habit design anyone has shipped in this category, and the free tier is a real full course rather than a demo. Its Spanish course now claims content through CEFR B2, up from A2, across nine languages as of 22 April 2026. For building the vocabulary base, it's hard to beat at $0.

Duolingo also wins on breadth. 40-odd courses, plus Chess, Math and Music. Web, iOS and Android. Speak is six languages and a phone.

Speak wins on spontaneous speech. Its whole product is you talking without a script, which Duolingo's free tier structurally doesn't do. Duolingo's speaking exercises show you the target sentence before you say it, so the hard part, pulling the words out of your own head, is done for you. Duolingo has the unscripted version, and it's behind Max.

Speak wins on pronunciation feedback for the same reason, though neither company publishes reliability figures for its speech recognition, so I can't tell you how accurate either is.

What neither does is keep the words you personally fumbled. You have a conversation with Speak, you stumble on acordarse, the call ends, and the word goes wherever those go. Duolingo teaches its own curriculum on its own schedule, which is why a long streak and a blank mouth coexist so often.

That gap is the job Verbamor does, and it's a third job rather than a third contender. It records a tutor lesson or takes a translation, pulls out the words and phrases you actually got wrong, and builds fill-in-the-blank cards scheduled by FSRS so they come back before you forget. Five languages: Spanish, French, Italian, Portuguese and Russian. iOS only, $19.99 a month or $129.99 a year with a 7-day trial on the annual plan, and no free tier. Both are US prices, read off the app's own plan definitions on 21 September 2026; the App Store sets a different number in other countries. It teaches you nothing new, which is the point and also the limit: it holds what somebody else taught you, so it won't get a beginner their first thousand words. Duolingo is better at that, and free.

If you'd rather not pay anyone for the holding part, Anki is free on desktop and Android and its scheduler is excellent. You type the cards yourself. Plenty of people prefer that, and they're not wrong.

Questions people actually ask

---

Sources

Every load-bearing claim Verbamor makes is traced to its paper on the research page.