Learning and careers decision
Speak
A capable developer can build a useful spoken-practice workflow (record, transcribe, LLM tutor, simple pronunciation feedback) in about a week using third-party speech and LLM APIs, but reproducing Speak's expert curriculum, personalization, multi-language polish, and large-user trust at scale would be difficult to match.
Visit website↗Open-source builds that already do this
Every project below is open source and already does this job today. Fork one, self-host it, or take the parts you need — the build prompt further down assumes an empty file, and this is the shortcut past that. Licences differ; check the one on each card before you ship.
What a replacement has to do
- Present a short lesson -> record user speaking -> transcribe audio -> score/prioritize pronunciation & phrasing errors -> run an AI conversation turn that responds and gives feedback -> store progress for personalization
What it still won’t have
- Expert-crafted proprietary curriculum and pedagogy
- Polished mobile apps and localization across many languages
- Proprietary Speak Tutor optimizations and personalized review features
- Large user base and associated trust/ratings
- In-app promotions, regional pricing, and partner integrations
What remains hard
- Infrastructure at scale
Speak has been downloaded more than 15 million times worldwide and holds a 4.8-star rating across the App Store and Google Play.
- Brand trust
Speak has a partnership with OpenAI, the company behind ChatGPT, and is backed by the OpenAI Startup Fund.
First-year cost
Keep paying
Paying is—cheaper in year one.
On cash alone, building overtakes the subscription at 11 seats.
Money you would actually spend
Time you would spend
—
What you would spend
What we assumed
The verdict above measures whether you could build it. This one is only about money.
Runnable build prompt
Build a minimal Speak-like web app using Next.js + React frontend, a Node/Express API, Postgres for data, and deploy on Vercel (frontend) + Heroku or Render (API). Integrate Whisper or Google Speech-to-Text for transcription, and call OpenAI GPT-4 (or similar) for conversational tutor turns and feedback. Implement: user signup + one subscription tier (Stripe), lesson storage (JSON lesson templates), audio recording/upload, transcription pipeline, a pronunciation feedback step that compares transcript to target phrases and returns simple error highlights, an LLM-driven tutor response generator, basic progress tracking and a ‘repeat practice’ flow. Out of scope: building custom-trained speech models, iOS/Android native apps, and large-scale localization. Include error handling for failed uploads/transcription, retry logic for API calls, and automated tests for the API endpoints and core lesson flow.
How we checked
How the score was reached
- Partly verdict base52
- An open-source build was found+5
- 3 cited sources+3
- 3/3 assessment runs agreed+4
- Hard moats found in the evidence-3
- Evidence score61
The base comes from the verdict. Everything under it is a check that either happened or did not, and each one is a fact frozen in this record rather than a judgement made at render time — so the same evidence always produces the same number.
How scoring works →Cited sources · 3
Every page the run actually retrieved.
- official productSpeak - The language learning app that gets you speaking
- open sourcemengxi-ream/read-frog
- open sourcehahahumble/speechgpt
Integrity checks
What held up, and what did not.
