Lingo Pals — Children's Language Learning App
Learning Spanish feels like play, not school
Project Overview
An edtech publisher wanted to move beyond flashcard apps and build something children would genuinely want to open every day. Lingo Pals is built around an AI companion character (Mochi the fox) who speaks the target language, reacts to correct and incorrect answers with genuine emotion, and guides children through branching story episodes where language choices affect the plot. The result is an app children ask to play rather than one parents force.
The Challenges
- 1
Speech recognition for children is notoriously inaccurate — standard ASR models are trained on adult voices and fail badly with childlike pronunciation, accents, and uneven pacing.
- 2
Keeping a 5-year-old engaged requires immediate positive feedback, short sessions, and a sense of narrative progress — standard drill formats fail in under 2 minutes.
- 3
Parents needed visibility into learning progress without the app feeling like a school report card.
- 4
Content needed to scale to 6 target languages without 6× the content production cost.
Our Approach
Speech recognition uses a custom model fine-tuned on a 120-hour dataset of children aged 4–12 recorded in 8 English-speaking countries — dramatically improving accuracy for young voices. Mochi is a character driven by a state machine: 12 emotional states with corresponding animations, triggered by the learner's answer pattern, streak length, and session energy. Stories are authored in a branching-narrative engine: each episode has 3–5 vocabulary targets woven into plot-relevant dialogue, and correct pronunciation unlocks new story branches. Parental dashboards show vocabulary mastered, session length, and a weekly 'language moment' highlight — designed to share, not just monitor. The content engine supports multi-language rendering so new languages add vocabulary and audio without rebuilding story logic.
Key Features & Metrics
Custom child-voice ASR model fine-tuned on 120 hours of children aged 4–12
Mochi AI companion: 12 emotional states reacting to learner performance in real time
Branching story engine: pronunciation unlocks new plot branches and character reactions
Vocabulary spaced-repetition layer woven invisibly into story dialogue
Parent dashboard: vocab mastered, session time, weekly highlight clip
6-language support (Spanish, French, Mandarin, Arabic, Japanese, German) from single content engine
Results & Business Outcome
30-day retention hit 89% — 3.9× the mobile language-learning category average of 23%. Average daily session length is 14 minutes, sustained over 6 weeks. Children completed an average of 240 vocabulary items in their first month. The app received a Common Sense Media Editor's Choice Award in its launch quarter.
Children do not learn by being taught — they learn by playing. Build the play, and the learning follows without anyone noticing it is happening.
Let's Build Something Intelligent Together
Tell us about your project. We'll respond within 24 hours with a custom plan and transparent pricing.