Speak
AI conversation tutor for language learners — talk first, drill later
Overview
Speak pairs AI Tutor lessons, Live Roleplay scenarios, and open-ended Free Talk with real-time pronunciation feedback across six languages, built on OpenAI's Realtime API; there's a 7-day trial but no permanent free tier.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 31 Aug 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- 6 languages
- Live Roleplay & Free Talk
- 3 custom lessons/day
- Everything in Premium
- Unlimited custom lessons
- Personalized study plan
Use cases
What you can produce with Speak
- Spoken fluency improvement through daily AI conversation practice
- Real-time pronunciation feedback identifying specific sound errors
- Completed roleplay scenarios modelling common real-world interactions
- Personalised lesson plans generated from tracked errors (Premium Plus)
- Grammar explanations delivered via Speak Tutor in-context chat
- Progress reports on pronunciation accuracy and vocabulary coverage
- Enterprise-grade English training deployment via Speak for Business
ASEAN Perspective
Speak in Southeast Asia
Speak's earliest and deepest market penetration was in South Korea, where roughly 6% of the national population used the app and it ranked as the No. 1 Education app on Apple's Korean App Store — a credibility signal that resonates across ASEAN where K-pop and Korean drama exposure has also raised demand for Korean lessons. Taiwan is cited as another breakout market. ASEAN learners seeking to improve workplace English — a high-priority credential across Singapore, Malaysia, the Philippines, Vietnam, and Indonesia — will find Speak's conversation-first model and Speak for Business enterprise tier directly relevant. The primary friction point for price-sensitive ASEAN markets is the lack of a free permanent tier; regional pricing varies and promotional discounts are common, but the standard rate is above what many competing apps charge in the region.
Speak's pitch holds up in the roleplay mode: dropping a learner into a coffee-shop or job-interview scenario and reacting in real time via OpenAI's Realtime API beats static flashcard apps for anyone who actually needs to speak a language, not just read it. Six languages are covered, curriculum quality is strong for beginners, and Speak for Business claims 85% employee adoption across 200+ corporate clients.
The cracks show at the edges. Speech recognition is lenient enough that garbled word order can still score full marks, feedback thins out past intermediate level, and there's no spaced repetition for vocabulary. Premium ($20/mo) versus Premium Plus ($40/mo) is a confusing split — mostly usage caps, not features. No free tier, no public API. ELSA Speak beats it on pronunciation precision; Duolingo beats it on price.
What people say
Mispronounce a word or scramble the sentence order entirely in a Speak drill, and the app will often still mark it correct — a complaint that shows up again and again in user reviews of the speech-recognition engine. That leniency undercuts what is otherwise Speak's strongest feature: dropping learners straight into spoken scenarios (a coffee order, a doctor's visit, a job interview) instead of matching-pairs exercises.
The app has real scale behind it. 15 million-plus downloads, a 4.8 App Store rating, and a December 2024 Series C that valued the OpenAI Startup Fund-backed company at $1 billion. Six languages are supported — English, Spanish, French, Korean, Japanese, Italian — with AI Tutor video lessons, Live Roleplay, and Free Talk conversation practice built on OpenAI's Realtime API for low-latency voice.
Where it thins out: feedback past intermediate level turns generic, there's no spaced-repetition vocabulary system, and the Premium vs. Premium Plus split (mostly usage caps on the same features) confuses more than it clarifies. There's no permanent free tier, just a 7-day trial, and no public developer API — anyone wanting LMS integration has to go through the separate Speak for Business sales process. For a beginner who wants to start talking fast, it delivers. For an advanced learner wanting nuance, ELSA Speak's pronunciation drilling or a human tutor will serve better.
Summary of public user & expert reviews, compiled by RECATOOLS.
Notable facts
- Speak's co-founders Connor Zwick and Andrew Hsu are both Thiel Fellows — Peter Thiel's programme that pays young people $100,000 to skip college and build companies instead.
- Users collectively spoke over one billion sentences on the platform in 2024 alone, generating a vast proprietary dataset for model training.
- Speak reached approximately 6% of South Korea's entire population before expanding globally — making it one of the fastest-adopted consumer apps in Korean App Store history.
- The company's Series C in December 2024 valued it at $1 billion — unicorn status — with the valuation doubling from $500 million in under six months.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including Speak's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Speak unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Speak directly →
Spotted something out of date? Suggest an update →
More in Education
