28 language pairs. Audio you record gets transcribed, broken down word by word, and explained with examples, conjugations, IPA pronunciation and the cultural notes a textbook would never bother to include.
Every query expands into a small dossier: translation, examples in context, conjugation grid when relevant, IPA, and a cultural note. Built for learners who already tried the flashcards and got bored.
Paste a sentence or speak it. Each word gets parsed into root, form, role and meaning in context — not the dictionary entry, the meaning here.
Whisper-style models handle the audio. You can ask a question by talking — useful when you're not sure how to spell what you just heard on a podcast.
Three examples per query, rotated. Real sentences, not the textbook "Where is the library?" type. Tells you when a word is formal, slang, regional or dated.
Verbs come with the tenses you actually need first, marked irregular where they are. No drowning in a wall of forms you'll never use.
Pronunciation in IPA so you can hear it in your head, and a one-paragraph cultural note where it matters — false friends, register, regional usage.
Each pair is a directional learner experience, tuned to the source-target combination. Pick yours.
A deliberately small stack so I can iterate fast and self-host it cheap.
The full app, ported to this site with a real LLM behind it. 8 queries per browser session. History pre-seeded across 11 language pairs so you land in a lived-in state.
Open the demo →For your school, your app, your kid, your product. I'll adapt the model, the language pairs, the UI. Tell me what you need.
WhatsApp