Services · Demo
Live captions, in your browser
This demo transcribes what you say in real time, right here. No servers of ours are involved: recognition is done by your own browser. That is exactly why there is something you should know before turning on the microphone — we tell you below, not in fine print.
The demo
This demo is interactive
Seeing it live requires JavaScript and a browser with speech recognition (Chrome or Edge). Below you can still read a real example of what it produces.
Sample transcript
This is what the demo produces with a clear voice and a close microphone:
- Hello, and welcome to Instituto AIA.
- This is a live captioning demonstration.
- What I say appears as text, almost instantly.
- It works in a classroom, a talk or a community meeting.
- The full service will also translate into other languages.
- Translation into Indigenous languages will be built hand in hand with their communities.
- Your audio is not stored on our servers.
- When you finish, you can download the transcript.
- Thanks for trying it.
This is a demo: it works best with a clear voice close to the microphone, and it does not fix mistakes. Sessions last up to 2 minutes; the full service will allow 60.
Before turning on the microphone
To recognize your voice, your browser sends the audio to its maker's servers (Google for Chrome, Microsoft for Edge). Instituto AIA does not receive, store or see your audio or your transcript: everything happens between your browser and its maker. If you prefer not to use the microphone, watch the pre-recorded example.
The full service
What the demo does not do yet
- Live translation: it already works right here on recent Chrome and Edge, using the browser's built-in translator — the model runs on your device. On other browsers it arrives with the full service, and Indigenous-language translation will be built hand in hand with their communities.
- Long sessions: the limit here is 2 minutes; the full service will allow 60.
- A precision about privacy: our general commitment says your audio is never stored on our servers, and that holds here too — but in this demo the recognition is done by your browser with its maker's servers. The full service will process audio on our own infrastructure.