Group classes leave little talk time
One teacher, many children: in a typical class each child gets only a few chances to speak, and shy kids often get fewer.
Pre-launch · Speaking tutor for ages 5–12
An AI tutor that gets kids speaking English.
We're building a 1-on-1 English speaking tutor for children learning English as a second language. It's a real-time voice conversation, not a chat window: our AI teacher talks with your child and runs a real lesson on an interactive screen — picture cards, model sentences, and plenty of turns for your child to talk.
A voice lesson: just speak and tap — the demo runs in your browser.
The problem
Kids learn to speak by speaking. Most rarely get the chance.
Speaking improves with many turns, quick feedback, and another try. Today's options make that hard to get enough of.
1-on-1 tutors are hard to access
Human tutors are wonderful, but they're expensive, hard to schedule, and hard to find for every family that wants one.
Chatbots aren't built for kids
A blank text box and long replies don't work for a six-year-old. Young learners need pictures, short turns, clear structure, and lots of encouragement.
Our approach
AI teacher + interactive courseware
Our AI teacher doesn't just chat. It runs a structured lesson on a screen it controls: it shows picture cards, models a sentence, opens the microphone when it's the child's turn, and responds to every answer or tap.
The goal is a lesson that feels like a patient human tutor — one where the child does most of the talking.
| A typical chatbot | Our lessons |
|---|---|
| A text box to type into | Picture cards, audio, and a microphone |
| Long replies to read | Short teacher turns; the child talks most |
| Open-ended conversation | Follows a lesson plan with clear goals |
| Waits for the user to lead | Leads, encourages, and adapts the pace |
How it works
A two-way lesson protocol
The AI teacher and the courseware talk in small, structured messages. The teacher drives the screen; everything the child does flows back to the teacher, who decides what happens next.
AI reasoning layer
AI Teacher
Follows the lesson plan, reads each student event, and decides the next teaching move.
The lesson screen
Interactive Courseware
Picture cards, model audio, hints, the microphone, and stars — all driven by the teacher's commands.
Ages 5–12
Student
Speaks, taps a card, or asks to hear it again. Every action goes back as a structured event.
One question, message by message
A simplified look at a single question in a weather lesson. The screen never decides what to teach, and the teacher never has to guess what the child did.
-
⬇ Downlink
Teacher → courseware. Commands likeshow_cards,highlight_card,play_model,open_mic,show_hint,award_star,next_activity. -
⬆ Uplink
Student → teacher. Events likespeech_result,card_tapped,request_repeat.
Messages are illustrative and simplified. Our protocol is still evolving.
-
⬇ Downlink
{"cmd":"show_cards","cards":["sunny", "rainy", "snowy"]} Teacher: “What's the weather like today?”
-
⬆ Uplink
{"event":"card_tapped","card":"rainy"} Child taps the rainy card.
-
⬇ Downlink
{"cmd":"highlight_card","card":"rainy"} {"cmd":"play_model","text":"It's rainy."} {"cmd":"open_mic"}Teacher: “Yes! Listen: It's rainy. Now you say it.”
-
⬆ Uplink
{"event":"speech_result","transcript":"it rainy"} Child: “It rainy.”
-
⬇ Downlink
{"cmd":"show_hint","text":"It's …"} {"cmd":"open_mic"}Teacher: “So close! It's rainy. Try once more.”
-
⬆ Uplink
{"event":"speech_result","transcript":"it's rainy"} Child: “It's rainy!”
-
⬇ Downlink
{"cmd":"award_star"}{"cmd":"next_activity"}Teacher: “Great speaking! Here's a star. Let's look at the next picture.”
What we're building
Designed for how young children learn to speak
Here's what we're building into the product. We're pre-launch, so each item is labeled with where it stands today.
Building now
Speaking-first lessons
Short questions, quick turns, and many chances to try again, so the child — not the teacher — does most of the talking.
Building now
Courseware the teacher controls
Picture cards, highlights, model audio, and the microphone respond to the teacher's commands, keeping each lesson visual and on track.
Building now
Kid-friendly feedback
Warm, specific encouragement in simple words. A mistake gets a clear model to copy and another try — never a red X.
Building now
Adaptive pacing
Stuck? The teacher slows down, gives a hint, or offers a choice. Flying? It moves on to something more challenging.
Early development
Pronunciation practice
Listen-and-repeat practice with clear models of sounds, words, and sentences, plus gentle tips on what to try next.
Roadmap
Progress for parents
Simple summaries of what your child practiced and how they're growing, with parents in control of their child's data.
Technology
An AI reasoning layer behind every lesson
Behind our AI teacher is a reasoning layer — the part that decides what to do next in the lesson. The teaching ability comes from large language models such as Claude or GPT; our lesson protocol and interactive courseware turn that ability into a structured, kid-friendly lesson.
Every spoken turn runs through a real-time voice pipeline: speech recognition (ASR) → AI reasoning → text-to-speech (TTS). The child talks, the AI teacher decides, and the teacher answers out loud while driving the courseware.
-
Follow the lesson planStay on the lesson's goals, vocabulary, and sequence.
-
Decide the next teaching movePraise, model, hint, repeat, or move on — after every answer.
-
Speak at the child's levelGenerate short, warm, age-appropriate language and feedback.
-
Drive the coursewareEmit screen commands as structured tool calls.
Whichever model runs underneath, the job is the same: follow the plan carefully, use the courseware tools precisely, and stay patient and consistent, turn after turn.
Claude is a trademark of Anthropic and GPT is a trademark of OpenAI. Hyperlink Dynamics is an independent company and is not affiliated with either.
Every turn of a lesson
In
- 📘 Lesson plan & goals
- 🧭 Where we are in the lesson
- 🙋 The child's latest event
AI reasoning layer
An LLM such as Claude or GPT
Out
- 💬 What to say next
- 🧩 Courseware commands as tool calls
{"cmd":"highlight_card","card":"rainy"}
Safety & privacy
Built for kids. Parents stay in control.
These are the commitments we're designing the product around from day one.
Designed with COPPA in mind
We're building around the U.S. Children's Online Privacy Protection Act, including verifiable parental consent before a child uses the product.
No ads
No advertising in lessons or anywhere in the product. A child's attention belongs to learning.
We don't sell data
We don't sell personal information about children or their families.
Minimal data collection
We aim to collect only what's needed to run lessons and show progress, and to keep it no longer than necessary.
Parents in control
Parents will be able to see what's stored about their child and have it deleted.
Age-appropriate by design
An encouraging tone, simple language, and content kept within the lesson. It's a teacher, not an open-ended chatbot.
These are design commitments for the product we're building. For how this website and the demo handle data today, read our privacy notice.
Try a 3-minute interactive lesson
Have a short voice conversation with the AI teacher: listen, look at the picture cards, and answer out loud. Best in Chrome or Edge with a microphone and speakers on.
- 🎧 Listen
- 🎙️ Speak
- 👆 Tap
Start the demo
Runs in your browser. No sign-up.
Contact
We'd love to hear from you
We're an early-stage startup based in the US, building the product and piloting lesson design. If you're a parent, teacher, school, or researcher interested in speaking practice for young learners — or you simply have feedback — get in touch.
tony.xu@hyperlinkdynamics.comWhere we are
-
Now
Building the AI teacher and interactive courseware, and piloting lesson design.
-
Next
More lessons for ages 5–12 and early pronunciation practice.
-
Later
Progress summaries for parents and lessons for teens.