The practice side is quick to build: a conversation partner that speaks your target language at your level, explains each mistake and feeds those mistakes into a review deck is a sitting's work with a language-model API and the browser's speech tools. What Busuu sells is a structured course across 14 languages, certificates, and a community of fluent speakers who answer you. A model's lessons contain errors, and a chat covers only what you ask it. Treat the build as practice for someone past the first lessons, not as a course.
Before you buildBusuu's free account gives a limited course, and Anki plus Language Transfer cover a lot at no cost. Try them before paying for either Busuu or a build.
Build me a personal language conversation partner to practise alongside Busuu's AI conversations. Requirements: - Node.js 22 + TypeScript: a Fastify server with one plain HTML page and a little vanilla JS on http://127.0.0.1:4940. `npm start` serves it. - Setup page: choose the target language, my level (CEFR A1 to C1) and one of 8 scenarios kept in scenarios.json (booking a hotel, renting a flat, a job interview, small talk and so on). - Conversation: the `openai` package pointed at LLM_BASE_URL with LLM_MODEL and LLM_API_KEY from .env plays the role, answers only in the target language at my level in 1 to 3 sentences, and streams to the page. - After each of my turns a second call returns JSON validated with `zod`: the corrected sentence and up to 3 mistakes (type: grammar, word choice, spelling or register; the wrong span; the fix; a one-line explanation in English). Show corrections under my message, never inside the reply. - Mistake repair: save every mistake in better-sqlite3 at ./data/partner.db and schedule it as a card with `ts-fsrs`. The Review page shows the original sentence, asks me to rewrite it correctly and shows the fix after I answer. - Speaking: replies are read aloud with the browser's `speechSynthesis` in the language's voice; a mic button fills the text box with `SpeechRecognition` where supported and says that Chrome sends the audio to its speech service. - Progress page: sessions, minutes, mistakes by type and the 10 most repeated mistakes; export all mistakes as CSV for Anki. - Single user. No accounts, no telemetry, local-first. Outbound calls: the model endpoint and the browser's speech services. - Out of scope: a curriculum, pronunciation scoring, certificates, other learners, a mobile app. - README: where to get an API key or run a local model (any OpenAI-compatible server works), and a plain warning that the model makes mistakes, more in less common languages, so check what you memorise.
Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.
The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Busuu Premium.
Add a reading mode to the conversation partner in this folder. Requirements: - A Read page that asks the configured model for a 150-word story at my CEFR level about a topic I type, in the target language, and shows it with a glossary of the 10 hardest words. - Clicking a word shows its glossary entry; a Save button adds the word to a vocabulary table and schedules it as an `ts-fsrs` card with the sentence it came from. - Add a quiz of 5 questions generated from the story, validated with `zod`. - Mention in the README that stories can contain errors.
Add a vocabulary importer to the conversation partner in this folder. Requirements: - `npm run import -- words.csv` reads rows of word, translation and an optional example, and creates cards scheduled with `ts-fsrs`, skipping exact duplicates. - Print counts of added, duplicate and invalid rows before writing, and ask for confirmation. - The Review page mixes vocabulary cards with mistake cards, capped at 20 new cards a day. - Update the README with how to paste your lists into a CSV by hand.
Add a daily reminder to the conversation partner in this folder. Requirements: - `npm run nudge` sends a message to an ntfy topic from .env with the number of cards due and the current streak, only if cards are due or no session happened today. - A launchd plist and a systemd timer that run it at 19:00 local time. - Count a day as practised after 5 minutes of conversation or 10 reviewed cards. - Update the README.
A course tells you what to learn next, in order, and that is what people lack when they try to study with a chat window alone. Busuu adds real speakers' feedback, certificates for each level and apps that work on a commute, which is why learners pay for it even though conversation practice is now cheap.
| Plan | Monthly | Annual, per month | What you get |
|---|---|---|---|
| Free | $0 | $0 | A limited course with ads; the plans page lists complete courses, grammar and vocabulary review, offline mode and certificates as paid features. |
| Premium | custom | — | Grammar Review, Vocabulary Review, speaking assessments in selected languages, no ads and certificates. |
| Premium Plus | custom | — | Everything in Premium plus AI conversations, pronunciation practice, exclusive content, streak protection and communication courses, in selected languages and on selected platforms. |
A free account gives a limited course; Premium adds grammar and vocabulary review, no ads and certificates.
Not really. A course tells you what to learn next, in order, and that is what people lack when they try to study with a chat window alone. Busuu adds real speakers' feedback, certificates for each level and apps that work on a commute, which is why learners pay for it even though conversation practice is now cheap. The prompt on this page builds the closest honest substitute.
Busuu Premium's pricing is usage-based or varies by plan.
The course: written lessons across all levels in 14 languages, with a structure to follow; Certificates when you pass a level; The community of learners and fluent speakers who reply to what you write and say; Grammar and vocabulary review built on the lessons you actually did; Offline lessons and a polished mobile app with streak repair. If any of those matter to you, keep paying.
Yes: Anki (open source), Tatoeba (open source), Language Transfer (free).