How NativePal works
What happens between opening the app and speaking better: generated content at your level, scored speech, spaced repetition, and what the app needs to do it.
What happens between opening the app and speaking better: generated content at your level, scored speech, spaced repetition, and what the app needs to do it.
NativePal is built as one loop rather than a menu of exercises. Each step feeds the next: reading supplies words, the dictionary keeps them, spaced repetition brings them back, speaking makes you produce them, and conversation uses all of it.
Content is generated, not pre-written. Stories, vocabulary lessons, word explanations and feedback are produced for your level by large language models from Google (Gemini) and OpenAI, through their APIs. That is why every reading lesson is new and why the app can teach 57 languages with the same depth of explanation.
Speech is transcribed and analysed by the same class of models. Pronunciation feedback names the sounds, stress and rhythm, fluency and connected-speech issues it hears and gives a score from 0 to 100, so you know what to change on the next attempt. Daily Challenge feedback covers grammar, vocabulary and phrasing, with natural alternatives at three registers.
Flashcard scheduling uses the SM-2 spaced-repetition algorithm and runs on your device. Character voices use Google Cloud and OpenAI text-to-speech.
Most languages use CEFR A1 to C2. Japanese uses JLPT, Chinese uses HSK and Korean uses TOPIK. You choose your level for each language and can change it at any time; there is no placement test. Reading, vocabulary and prompts are generated for the level you set. Proficiency levels explained →
An internet connection for anything AI-generated, which is most of the app. No account is needed: anonymous sign-in lets you start without an email address, and an account only matters if you want progress to follow you to a new device.
Requests to the AI providers carry only what is needed to answer them, and the providers do not use that data to train their models. The privacy policy lists the providers and what is collected.
Not a chatbot with a language skin: chat is one tool inside a loop with levelled content, scoring and spaced repetition. ChatGPT vs NativePal →
Not a course with a fixed syllabus. Learning Scenarios give beginners a guided start, but the app does not march everyone through the same units in the same order.
Not an offline app, and not a human tutor. A teacher will still catch register and nuance that AI feedback misses. italki vs NativePal →
Every feature has a daily free allowance, and premium lifts the limits and unlocks the longer formats. Exact numbers, and where to see the price, are on the pricing page.
Stories, vocabulary lessons, explanations and feedback are generated by large language models from Google (Gemini) and OpenAI, through their APIs. Speech is transcribed and analysed by the same class of models. Character voices use Google Cloud and OpenAI text-to-speech. The privacy policy names the providers and states that they do not use your data to train their models.
No. Anything generated or analysed by AI, which is most of the app, needs an internet connection. Saved words and flashcards are stored for your account, and flashcard scheduling runs on the device, but the app is not designed for offline study.
You read a text aloud on the teleprompter. Your recording is analysed and you get a score from 0 to 100 with feedback on individual vowels and consonants, stress and rhythm, fluency and pacing, and connected speech. You can repeat the same text and compare attempts. Voice replies in chats get similar pronunciation notes.
Each saved word is scheduled with the SM-2 algorithm, the same family of scheduling used by classic flashcard apps. After each review you mark whether you knew the word, and the interval before it comes back grows when you did and shrinks when you did not. Scheduling runs on your device, and flashcard review is unlimited on the free tier.
No. A general chatbot gives you an open conversation and nothing else. NativePal generates reading and vocabulary at your declared level, scores your speech, schedules your review with spaced repetition, keeps your dictionary and progress, and offers chat as one tool among those. The comparison page sets out the trade-offs in both directions.
No, and there is none. You choose your level for each language when you set it up (CEFR A1 to C2, or JLPT, HSK and TOPIK for Japanese, Chinese and Korean), and you can change it at any time. Reading, vocabulary and prompts are generated for that level.
Reading, vocabulary, scored speaking and conversation in 57 languages.
Free to start. No account needed.