Hoot

• 

  Thumbnail

• 

• 

• 

Inspiration

We recently met a kid who launched into a confident, curious monologue about Wormholes and the Kailasa Temple — topics most adults couldn't explain on the spot. And then we thought about every other curious kid we'd seen at a dinner table, asking questions an overworked parent didn't have the energy to answer, eventually turning to a phone for stimulation that doesn't stimulate. Hoot started from that gap: kids today are far sharper and more curious than apps give them credit for, and the alternatives they're left with — endless reels, dumbed-down "kids" content — are actively shrinking their attention and confidence in speaking. We wanted to build something that takes curious kids seriously, talks back to them, and grows their voice instead of replacing it with a scroll.

What it does

Hoot is a voice-first curiosity companion for kids and pre-teens, built around a friendly owl mascot. Children speak with Hoot the way they'd speak with a patient, knowledgeable friend: they ask anything ("why was Pluto removed from the planets?"), follow up, go on tangents, and get answers that are real and honest — pitched to their age, never dumbed down.

The app has three parts:

• Ask Hoot — open-ended voice Q&A with full conversation context.
• Explore by Subject — browse AI-generated "wonders" across Space, History, Nature, How Things Work, Body & Mind, and more, and tap one to start a conversation.
• Story with Hoot — interactive or continuous narrated stories the child can interrupt or co-author by speaking their choices.

Hoot adapts to the child's age band, lets them steer the depth of any answer ("explain it simpler," "tell me more," "why?"), supports multiple languages, and lives inside a custom doodle-illustrated design with multiple themes so kids can make it feel like theirs.

How we built it

Hoot is a Next.js (App Router) + TypeScript app styled with Tailwind, mobile-first and responsive. The entire visual identity — the Hoot owl mascot, the icon set, the doodle illustrations — is custom inline SVG so the look stays consistent and crisp at any scale, with a CSS-variable theme system that re-skins the entire app instantly.

The agent layer is server-side: Hoot's persona, the age-band adaptation, the conversational guardrails, the multilingual handling, and the depth-control intents all live in our own LLM logic, with conversation state maintained per session so follow-ups and tangents work naturally. The real-time voice layer is powered by the ElevenLabs Speech Engine SDK, which gave us speech-to-text, text-to-speech, turn-taking, and barge-in interruption. Profile data (name, age band, language, theme) is persisted in localStorage so each child picks up where they left off.

Challenges we ran into

The hardest challenges weren't technical — they were product-ethical. We initially designed a feature that would listen to a child's emotional chats and silently send summaries to their parents. It sounded protective, but on closer look it failed the basic test of trust (kids confide because it feels private), and in the worst case it could route sensitive information to the wrong adult. Cutting it took discipline, and it made the product genuinely safer.

Designing age adaptation that respects kids was another tension. Spoken explanations need to be shorter and simpler than written ones at the same age — there's no scrollback in audio — and the line between "pitched to a six-year-old" and "patronizing a six-year-old" is thin. On the engineering side, the interruption-and-resume behavior (Hoot stops mid-sentence to handle a tangential question, and can pick up the original thread on request) needed careful conversation-state and request-cancellation handling to feel seamless. And then there was the timeless solo-builder challenge: ruthlessly cutting scope so we could actually ship.

Accomplishments that we're proud of

• It feels natural to talk to. The interruption-and-resume in particular makes Hoot feel like a thoughtful friend, not a button-mashed chatbot.
• It actually doesn't talk down. A six-year-old gets a real answer about Pluto's reclassification — just with different words than a twelve-year-old gets.
• A fully custom visual identity. Every icon, the mascot, every illustration is our own SVG — no stock emojis, no generic UI.
• Multilingual from day one, so children who don't think in English aren't left out.
• A clear ethical line. Hoot is a companion that nudges kids toward the people in their life — friends, family, teachers — never a substitute for them, and never a secret confidant.
• We shipped. End-to-end, working, deployed.

What we learned

The hardest part of building a voice agent isn't the voice — it's the behavior: when to be brief, when to dig, when to admit uncertainty, when to gently defer to a trusted adult. We learned that kids are routinely under-served by software that treats them as fragile or unintelligent, and that respect is itself a design principle, not a tone of voice. We learned that features which feel kind can quietly violate trust — and that the discipline of ruthlessly evaluating product-ethics, including cutting things that sound caring but aren't, is one of the most important skills in building for children. And we relearned the oldest lesson in shipping: a focused, finished build always beats a sprawling, half-finished one.

What's next for Hoot

We want Hoot to grow alongside the children who use it.

• Deeper subject collections with curated wonders co-designed with educators.
• A transparent, opt-in parent and educator view — a positive summary of what a child has been curious about, never an emotional surveillance feed.
• Offline and low-bandwidth modes so kids without reliable internet aren't shut out.
• More languages and regional voices, especially for under-served languages.
• Accessibility modes — low-vision support, dyslexia-friendly typography, simplified navigation.
• An on-device option so privacy is structural rather than promised.
• Partnerships with schools, libraries, and educators to put Hoot in the places where curious kids already gather.

The goal stays simple: more kids growing up confidently asking and answering questions out loud, instead of silently scrolling.

Built With

• elevenlabs
• next.js
• speechengine