Your first captions¶
Who this is for: admins (event producers) — operating the capture page.
This page is for reading captions when your captioner or admin already has the speech service running. You only need a browser and the page address they gave you. Unfamiliar words are explained in the glossary.
The speech service is the program that listens to audio and writes words down. The capture page is the browser tab where you choose a microphone and read the captions it produces.
Before you start¶
You need:
- The capture page address from your captioner, for example
https://caption.example.org/. - A desktop browser: Chrome, Edge or Firefox.
- A working audio input, such as a built-in microphone, a USB microphone or an audio interface.
Service access token
If the speech service is protected, the page may ask for a service access
token (CAPTION_API_KEY). Enter the token your captioner gave you. It is
not a Google, OpenAI or Gemini key. The page forgets it when you reload, so
keep it handy.
Open the capture page¶
- Open the address your captioner gave you in Chrome, Edge or Firefox.
- If the page asks for a service access token, type it in and connect.
- Wait until the page confirms it can reach the speech service.
Allow the microphone¶
The browser will ask for microphone permission. Choose the allow option so the page can hear you. Device names often stay hidden until you allow access, so the input list may fill in only after you grant permission.
If the page is not served over https:// or from localhost, the browser may
refuse the microphone entirely. Use the address your captioner provided rather
than a plain http:// one.
Pick the right audio input¶
In the Microphone selector, choose the input you actually want captioned:
| Input | Use it for |
|---|---|
| Built-in or USB microphone | One person speaking near the computer. |
| Audio interface | A mixer or venue feed you have plugged in. |
| Virtual audio device | Another app's sound that you set up separately. |
The app does not capture other apps' audio on its own. If you want system or conference audio captioned, route it into an input the browser can see.
Set the language¶
In the Spoken language selector, choose the language people will speak.
Choosing it explicitly is more reliable than letting the page guess, especially
for short phrases. The default small model handles many languages.
For output, you can choose one of these modes:
| Output mode | What you get |
|---|---|
| Transcription in the spoken language | Captions in the language being spoken. |
| English translation | Speech translated into English. |
| Transcription and English translation | Both together. |
Translation here means into English only. English-only models whose names end
in .en cannot translate.
Start and read captions¶
- Choose Start captions.
- Allow the microphone if you have not already.
- Speak a few sentences, then pause.
A microphone meter shows that audio is arriving. Visible text is the real check; if the meter moves but no words appear, see Troubleshooting.
Captions are not instant. Speech is usually collected in about six-second windows before it is processed, so text arrives a few seconds behind the speaker. Leave the sensitivity at Normal and the caption interval at 6 seconds unless you have rehearsed another setting.
First test
Read a short passage that you know well, pause, and compare the captions with what you said. This quickly shows whether the right microphone and language are selected.
Stop and keep the transcript¶
- Choose Stop and wait for the last speech to finish processing.
- Choose Download transcript.
- Open the downloaded file to confirm it contains what you need.
The transcript is the text collected in that tab. Nothing is archived automatically, so download it before you close or reload the page. A browser crash can lose unsaved text. See Save and share transcripts for more detail.
Nothing happens?¶
If the page will not load, the microphone is refused, or no captions appear, read Troubleshooting for the common fixes.