Cost Per Speaking Hour: The Real Language School Math
Vlad Podoliako
Founder & CEO, LinguaLive
Vlad Podoliako is the founder of LinguaLive, an AI-powered language learning platform. With a background in data science and artificial intelligence, Vlad is passionate about using technology to make language learning accessible and effective for everyone.
Follow on LinkedInA private language school sells a 60-minute group class like it's selling 60 minutes of speaking practice. It isn't. Between instructions, modeling, listening tasks, and twelve students sharing one clock, the average student in that room talks for a few minutes, not sixty — and the real price of a course isn't the tuition line, it's the cost per minute of speech a student actually produces.
In a commonly modeled classroom breakdown, a typical 60-minute group class of 12 students hands each student roughly 3 to 6 minutes of individual speaking time — after teacher talk time, instructions, and receptive tasks — call it 5 minutes as a working number. That turns a $300, 10-week course into about $6.00 per minute, or roughly $360 per hour, of speech a student actually produces, versus roughly $0.22-$0.89 per minute (about $13-$53 per hour) for a private tutor, and a few cents per minute (about $2-3 per hour) for capped AI voice practice used daily.
This is about private language schools and the P&L an owner or director of studies actually has to run. If you're on the university side instead, the university speaking-practice math covers the accreditation and procurement angles this piece doesn't. And this isn't a debate about gamified vocabulary apps — we've written separately about whether Duolingo builds real speaking ability — it's about time economics for tools built specifically for spoken conversation.
The Arithmetic Nobody Writes Down
Course brochures talk about contact hours. They never talk about speaking share — the fraction of the class clock that any one student spends with their own mouth open, producing language instead of receiving it. That's the number that actually predicts whether a student gets fluent, and it's nowhere on the invoice.
Here's a transparent model. It won't match your exact lesson plan, but the structure holds across most communicative group formats — swap in your own minutes.
| Segment of a 60-minute class | Minutes | Is the student speaking? |
|---|---|---|
| Warm-up, admin, attendance | 5 | No |
| Teacher instruction & modeling of new language | 12 | No — teacher talk |
| Listening / reading / writing tasks | 10 | No — receptive |
| Whole-class feedback & correction | 8 | Mostly teacher talk |
| Speaking practice (pair or group work) | 20 | Yes — split across 12 students |
| Wrap-up & homework assignment | 5 | No |
| Total | 60 |
Pair work vs. teacher-fronted: the ratio that actually matters
Of the 20 minutes earmarked for speaking practice, format changes everything. Run it as pair work — six pairs of two — and each pair owns the full 20 minutes to split between two speakers; assume real-world friction (task-switching, one partner dominating, instructions between activities) eats about a third of that, and each student nets roughly 5 to 7 minutes of actual speech. Run the same 20 minutes as teacher-fronted turn-taking instead — the teacher asks, one student answers, the class moves on — and 20 minutes divided across 12 students, minus the teacher's own framing time, nets closer to 60-90 seconds per student.
That's the real range: 1.5 to 7 minutes of individual speaking time in a 60-minute class of 12, depending entirely on how the room is run. Five minutes is a reasonable middle figure for a well-run class that uses pair work deliberately, and it's the number we'll use below.
Cost Per Student-Speaking-Minute: The Worked Comparison
Take that 5-minutes-per-class figure and run it against three ways students get speaking practice: the group class itself, a marketplace private tutor, and AI voice practice between classes. Same question each time — what does one minute of a student's own speech cost?
Start with the class. Say a 10-week evening course, one 60-minute session a week, priced at $300 — a common structure at private language schools (adjust to your own tuition). Ten classes at 5 minutes of individual speaking each is 50 minutes of actual speech across the whole term.
Marketplace platforms — we've compared them honestly in our iTalki alternative and Preply alternative breakdowns — list tutor rates that typically run $10 to $40 an hour depending on language and tutor experience, per each platform's published pricing as of mid-2026 (iTalki's professional teachers list $10-$40 per lesson; Preply's tutors average $10-$15 an hour, up to $40+ for experienced specialists). A focused 1:1 lesson runs closer to 75% actual student speaking time (the tutor still frames questions, corrects, and explains), so that's roughly 45 minutes of speech per hour.
LinguaLive's Pro plan costs $7.99 a month for up to 20 live voice minutes a day. A live session is a genuine back-and-forth — the AI is talking too, not just listening — so assume conservatively that half of every live minute is the AI's own turn. At daily use, that's up to 300 minutes a month of a student's own speech; at a more realistic 15 days a month, closer to 150 minutes.
| Delivery mode | Price | Individual speaking minutes | Cost per speaking-minute | Cost per hour of actual speech |
|---|---|---|---|---|
| Group class (12 students, 10-week term, 1×60-min class/week) | $300 example tuition | ~5 min/class × 10 classes = 50 min/term | ~$6.00/min | ~$360/hr |
| Marketplace private tutor (1:1) | $10-$40/hr | ~45 min/hr (75% speaking share) | $0.22-$0.89/min | $13-$53/hr |
| LinguaLive Pro (AI voice practice) | $7.99/month | 150-300 min/month (15-30 days used, 50% speaking share) | $0.03-$0.05/min | $2-$3/hr |
The gap isn't small. Group-class seat-math runs roughly 7 to 27 times more expensive per speaking-minute than a private tutor, and roughly 120 to 200 times more expensive than capped AI practice used daily. None of that means the group class is a bad product — it means the group class is selling something other than raw speaking volume: curriculum, structure, a room, a deadline, other humans. Keep reading — that part matters just as much.
What the Cheap Minutes Can't Do
Cheap minutes are still just minutes. Here's what they don't include, no matter how good the underlying model gets:
| What AI conversation practice can't do | Why it matters |
|---|---|
| Grade or certify | It has no accreditation or exam-board authority. It can rehearse a student for IELTS or TOEFL speaking sections, but it can't issue a credential a university or employer recognizes. |
| Make a pedagogical judgment call | Whether a student's confusion is really a grammar gap or a motivation problem is a human read, not a scripted branch. |
| Build peer dynamics | Classmates, gentle competition, and shared deadlines are a big part of why people enroll in a school instead of just buying an app. |
| Create external accountability | A fixed class time and a human instructor who notices when you stop showing up is a commitment device solo practice doesn't have. |
| Handle fully open-ended, culturally loaded conversation | A trained tutor improvises around a student's actual life and history; a voice-mode AI practices scenarios, not relationships. |
| Replace the institution itself | Enrollment, visas, in-person community, and the school's own brand are what a school actually sells. Speaking minutes are one input, not the whole product. |
None of this knocks the technology — it's an honest accounting of what a product like ours is actually for. We built LinguaLive to be cheap, available, judgment-free repetition, not to replace the reasons people walk into a language school in the first place.
The Blended Model: AI for Volume, Instructors for Judgment
The honest read of the math above isn't "replace the tutor" or "replace the class." It's "stop paying class-level prices for practice-level volume." Use the cheap channel for the thing it's actually good at: repetition. Use the expensive channel for the thing only humans do well: judgment.
Run the numbers on a single student. Five minutes of speaking time in class this week, plus 10 minutes a day of AI practice five days a week, adds roughly 50 minutes of additional speech — about 10 times their in-class total — without the school adding a single teaching hour or the student booking a single extra tutor session.
Match difficulty to the student, not the other way around. LinguaLive's four voice modes — Beginner, Practice, Pro, and Strict — roughly track a class's own leveling, so homework practice reinforces the syllabus instead of fighting it.
Packaging Speaking Practice Without Hiring More Conversation Staff
For a private school owner, this is a margin story as much as a pedagogy story. Conversation-heavy staffing is the hardest cost to scale — you either cap class size, hire more instructor hours, or accept that most students in a group format don't get enough airtime. None of those are great options alone.
| Approach | What it solves | Trade-off |
|---|---|---|
| Bundle it as an included "conversation lab" | Answers the #1 complaint — "I don't get to talk enough" — without adding class time | Needs framing so it reads as an addition, not a downgrade |
| Sell it as a premium-tier perk | Differentiates a higher-priced enrollment tier | Only works if positioned around outcomes, not the app itself |
| Assign it as leveled homework tied to the syllabus | Reinforces this week's vocabulary and grammar instead of competing with it | Requires instructors to actually reference it in class |
| Put the price gap in your own marketing | "$300 covers structured class time; a few dollars a month covers practice between sessions" | Only credible if the school is honest about what each layer actually does |
Before You Add Any AI Tool to a Paid Course
This applies to any AI vendor a school considers, including us. Do the compliance homework before the pilot, not after a parent or an auditor asks.
| Item | Why it matters | Action |
|---|---|---|
| Vendor's minimum-age terms | Many consumer AI apps, LinguaLive included, set a minimum age in their terms of service | Confirm your cohort's ages against those terms before you enroll a class — don't assume approval |
| Data processing agreement (GDPR Article 28) | If you have EU students, your school is the data controller and the vendor is the processor | Get a signed DPA specifying what's processed, for how long, and where it's stored |
| Children's consent (GDPR Article 8) | Under GDPR Article 8, a child's own consent for an information-society service is valid from age 16 by default, but member states may set a lower threshold no younger than 13 (e.g., the UK, Spain, and Ireland use 13; Germany and the Netherlands keep 16) | Confirm the exact age threshold your cohort's member state has adopted within that 13-16 band before enrolling under-16 students |
| FERPA (U.S. institutions) | Sharing student data with a vendor can implicate FERPA's education-records rules | Confirm the vendor meets the "school official" exception under 34 CFR § 99.31(a)(1)(i)(B) — which permits disclosure without consent to a contractor performing an outsourced institutional function under the school's direct control — or get direct parental/eligible-student consent instead |
| Where voice audio is processed and stored | Real-time speech is sent to a third-party AI provider, not kept on-device | Ask the vendor directly — don't assume |
| Model training use of student data | Some AI vendors train on user content by default | Get written confirmation this is opted out for your cohort |
Once compliance is settled, design the pilot deliberately: small and bounded, not a school-wide rollout on faith.
| Step | Action |
|---|---|
| 1. Scope | Pick one cohort or course, not the whole school. |
| 2. Baseline | Measure your own version of the seat-math above — how many speaking-minutes does a student actually get today? |
| 3. Define success upfront | Usage frequency, self-reported confidence, instructor-observed change, term-over-term renewal — decided before day one, not after. |
| 4. Run one full term | Four to eight weeks: long enough for habit formation, short enough to cut losses if it doesn't work. |
| 5. Sign compliance off first | DPA, minimum-age terms, and data location settled before go-live, not during it. |
| 6. Debrief with instructors | Did "not enough talking time" complaints drop? Did class time free up for higher-value work? |
FAQs
How much speaking time do students get in a group language class?
In a commonly modeled classroom breakdown, a typical 60-minute group class of 12 students hands each student roughly 3 to 6 minutes of individual speaking time after teacher instruction, listening tasks, and whole-class feedback — often closer to 1-2 minutes in teacher-fronted classes and up to 7 minutes in well-run pair-work formats. These are modeled estimates, not a universal measured average, but the gap they point to is real: course fees don't reflect it.
How much does a private language tutor cost per hour?
Marketplace private tutors (iTalki, Preply, and similar platforms) list rates that typically run $10 to $40 an hour depending on language and tutor experience — per each platform's published pricing as of mid-2026 (iTalki's professional teachers list $10-$40 per lesson; Preply's tutors average $10-$15 an hour, up to $40+ for experienced specialists). Since a focused 1:1 lesson runs closer to 75% actual student speaking time, that works out to roughly $0.22 to $0.89 per minute — about $13 to $53 per hour — of speech the student actually produces.
What does conversation practice cost per student?
It depends entirely on delivery mode, not the sticker price. A $300, 10-week group course delivering about 5 minutes of speaking per class works out to roughly $6.00 per speaking-minute (about $360 per hour of actual speech); a private tutor runs $0.22-$0.89 per minute (about $13-$53 per hour); capped AI voice practice can run a few cents per minute (about $2-3 per hour) at daily use.
Can AI reduce the cost of conversation classes?
Yes, for volume, not for judgment. AI voice practice like LinguaLive's costs a few cents per minute of a student's own speech at typical usage, letting schools multiply practice reps between classes without adding teaching hours. It doesn't replace the instructor time, correction, and motivation a group class or tutor provides.
Is AI conversation practice as good as a human tutor?
No, and it isn't trying to be. AI conversation practice is honest, low-stakes repetition: available anytime, judgment-free, and cheap per minute. A human tutor brings pedagogical judgment, nuance, accountability, and relationship an AI can't replicate. The strongest setup uses AI for daily reps and humans for everything else.
If you run a private language school and want to see this math against your own numbers before committing budget, talk to our education team about a scoped, time-boxed pilot — we'll help you set the baseline and the success metrics before a single student logs in. And if a human tutor or a different tool is genuinely the better fit for your cohort, we'll tell you that too.
Related Topics
Share this article
Ready to Start Learning?
Try LinguaLive's AI-powered conversation practice free. 10 minutes a day can transform your fluency.
Start Free - 10 Min DailyMore Articles
Language Learning Software Procurement Guide (Schools)
The four numbers to demand before buying AI language software for a school: per-seat price, usage cap, overage terms, and deletion terms.
AI Tutor Pilot Program for Schools: What to Demand
What a credible AI tutor pilot must include: fixed cohort, hard cost caps, human-scored baseline/endline, and exit criteria agreed before day one.