Community
Bring Speech Lab to your community.
A free, privacy-first way to practise speaking — and a real example of a student building their own AI project. Here is everything you need to run a session at a library, club, scout troop or school.
0
accounts needed, and recordings stored on our server. Everything is measured in the browser unless a participant opts in.
4
numbers you can check — pace, filler words, pauses, pitch variation — each with the method shown.
8
parts in the student build kit, from “count the words” to “your own app”.
A 45-minute session
- 5 min · The question. “What makes a short talk easy to follow?” Collect three answers on a board.
- 10 min · The demo. Open the Studio, play “Sample: first try”, press Measure. Ask: which number surprises you? Then play “Sample: second try” and compare.
- 15 min · Practice in pairs. Each person records 30–60 seconds on their own device, reads the transcript, and puts back every “um” the machine dropped. Partners count fillers by hand: does the app agree?
- 10 min · Change one thing. Choose one hint, record again with the same task label, compare. Nobody is graded.
- 5 min · Build your own? Show the build kit's flight plan. Anyone curious can start part 1 at home.
Demo script (2 minutes)
- “This is a practice tool, not a judge. It never rates you as a person.”
- Play Sample: first try → Measure. “Look: every filler word is counted, and here is each one, highlighted.”
- “Every number says how it was made. Missing data shows a dash, never a zero.”
- Play Sample: second try → Measure → Keep both → compare. “Same task, same way of measuring — only then is a comparison fair.”
- “The AI steps are optional. When the model quotes your talk, the app checks the quote is really in your words.”
Safety checklist for hosts
- Participants use their own devices or a shared device with the history switch left off.
- Under-13s take part with a parent or guardian present.
- AI steps stay locked unless an adult host has a demo code and explains what is sent where.
- Nobody records anyone else without asking. No recordings are collected by the host.
- Phrase feedback as observations (“3 pauses”), never labels (“bad speaker”).
Questions people ask
Is it free?
The Studio's on-device mode is free. AI steps for classrooms and clubs are available through a pilot — see Schools.
Where does my recording go?
Nowhere, unless you tick the box to send it for transcription. Even then it is processed and discarded — this app stores no audio, transcript or result on its server. See Privacy.
Does it work for non-native speakers?
The measurements describe what was said, not how “good” an accent is — we never rate accents. Speech recognition can be less accurate for some voices; that is why you always check and correct the transcript before counting.
How do you know the numbers are right?
Each metric is a small, published rule with automated tests, including tricky cases (“umbrella” is not “um”; silence before the first word is not a pause). Speech recognition is checked against hand counts. What we have not tested yet, we say so.
Who made this?
LIGHT HOPE built the Studio. The build kit is designed so that a student can build their own version from scratch, as their own original project.
Host a session
Write to info@x1000.ai with your organisation, the age group, and a date. We'll send a host pack and, for adult hosts, a demo code for the AI steps.