How to use Real-time Subtitles

Real-time subtitles add AI subtitles to a live stream, translate them into the languages you choose, and can also add an AI voice-over that reads the translations aloud as extra audio tracks. The feature is easy to use but requires contacting support before the first use. Just ask us to enable the feature and we will!

If you want to see a demo, you can see one here .

×

How the stream works

On a live subtitles channel, the stream your viewers watch doesn’t depend on the subtitle service:

• The stream plays with or without subtitles. You can stream before you start the subtitles, for example for a stream test. Starting, stopping or restarting the subtitles doesn’t interrupt the picture or the original audio. If the subtitle service has a problem, viewers lose the subtitles and the voice-over, and the stream carries on.
• Viewers are about 15 seconds behind your encoder, plus the usual streaming delay. The delay is fixed so that the subtitles and the voice-over can be placed on the frames they belong to.
• Subtitles end up in the recording. See Will I be able to download the subtitles or use them in the VOD?

Language settings

1. Make sure that you have Real-time Subtitles compatible live channel selected. If support has enabled Real-time subtitles but you can’t see Real-time subtitles in the menu under the Live section, you most probably don’t have the Real-time subtitles compatible live channel selected. You pick the channel under Live → External encoder, and the right one usually has real time subtitles in its name.

×

2. When you have the right live channel selected, you can find Real-time subtitles under the Live section of the Event Admin menu. Click it to open the settings. If the page tells you the chosen live channel is not real-time subtitles capable, go back to External encoder and pick another channel, or contact support to have one enabled.

×

3. Turn on Show subtitles at the top of the page. This is what makes the subtitles appear in the player on your event page. Save the event afterwards. It is the same setting as Enable subtitles under Edit → Subtitles and transcripts, so you only need to turn it on in one of the two places.

4. Choose the Source language. This is the language that is spoken in the stream. If more than one language is spoken, for example a panel discussion that switches between Finnish and English, choose Multilingual (auto-detect) instead of a single language.

When you choose Multilingual (auto-detect), an Expected languages field appears. It is optional: leave it empty and any supported language is detected. Listing the languages you expect to hear in the stream improves the accuracy of the transcription, so fill it in when you know what will be spoken. If you list a language the automatic detection does not support, it is skipped and the page tells you which one.

5. Choose the Translation engine, the AI engine that translates the subtitles. We recommend Gemini. If one of your chosen languages is not supported by the engine you pick, a supported engine is used automatically for that language, and the language row shows which one.

6. Pick your translations. The source language is always shown to viewers as subtitles, so it is not in the translation list. If you pick no translations, viewers get subtitles in the spoken language only.

7. Optionally turn on Voice-over for individual languages in the Selected translations list. Voice-over produces a spoken AI translation as an extra audio track on the stream, not a subtitle track. See AI voice-over below.

8. Check the panel next to the settings. It sums up your selection and the estimated duration from your event settings, and on accounts that show pricing, an estimated cost.

9. Click Save Language Settings. If a language you picked turns out not to be supported, the settings are still saved and a notification tells you which languages were skipped.

Billing

Real-time subtitles are billed by active minutes, meaning the minutes when your encoder is streaming and the subtitle service is running. The estimate in the admin is based on the event duration in your event settings, and the final bill reflects the actual active time. Set-up fees may apply, so check with your Videosync contact before use.

Turning the service on and off

1. You turn on the service before the event by clicking Start subtitled stream in the Live control room. Seeing this button requires having the Real-time subtitles capable live channel selected.

×

2. Turn the service on 5–10 minutes before you need the subtitles. Starting it can take up to 5 minutes depending on the load on the servers. While it starts, the button reads Initializing subtitled stream. The stream itself keeps playing in the meantime, and the subtitles and the voice-over appear once the service is running.

×

3. Once the Real-time Subtitles service is up and running, you will see a notification and there will be Stop subtitled stream text in the button.

×

4. Start streaming to the live channel from your encoder with the RTMPS URL and stream key of the live channel. You can do this before or after starting the service. Bitrate in shows when Videosync is receiving your stream. Viewers see the stream about 15 seconds after your encoder sends it, plus the usual streaming delay.

×

5. Ending the stream. When you are ready to end the stream, set the event to End live first, then click Stop subtitled stream, and after that stop the encoder.

Changing languages during a live event

You can change the languages while the event is live. After Save Language Settings, the page shows Changes saved, not live yet: the running subtitles keep the previous settings until you apply the change.

Click Apply now to apply it. The confirmation, Apply the change now?, tells you what applying the change does to the stream before anything happens:

• Most changes restart only the subtitle service. The subtitles and the voice-over pause for about 1–5 minutes. The picture and the original audio carry on.
• Some changes also restart the stream on our side. The stream cuts for a moment and your encoder has to reconnect, which most encoders do by themselves. Viewers also miss about 15 seconds of the program.

If you can, make language changes before going live.

AI voice-over

Voice-over turns a translation into speech. For every language you tick Voice-over for, the stream gets an additional audio track in that language, on top of the original audio. Viewers switch between them from the audio track selector in the player, the same way they would switch between interpreted feeds. Voice-over does not replace the subtitles. The subtitle tracks stay as they are, and a viewer can listen in one language and read in another.

Voice-over is not available for every language. The page checks the current availability for each language you have selected, and greys out the tick box with the note Voice-over is not available for this language where it cannot be produced. If the check cannot be completed for a language, voice-over is left off for that language and a notification after saving asks you to save again. Nothing unverified is turned on, so you are not billed for a voice-over that would not have been produced.

A language with voice-over costs the same as one more translation language, so a language you both translate and voice over counts twice in the estimate.

The voice-over follows the speaker by a few seconds, because a sentence is translated and spoken once it is finished. Until the first voice-over arrives, for example before the subtitles are started, the voice-over track is silent. A viewer who picks it hears silence instead of an error.

Default audio track

As soon as one language has voice-over, the stream has more than one audio track, and a Default audio track setting appears on the page. It decides which track the player starts with:

• Automatic lets the player choose. In browsers that means the viewer’s own language. A viewer whose browser is set to a language you produced a voice-over for hears that voice-over, and everyone else hears the original audio.
• The option marked (source), which is your source language, starts everyone on the original audio, whatever their browser language is. This is pre-selected for you when you turn on the first voice-over.
• Any voice-over language starts everyone on that AI voice-over.

Keywords

Names and acronyms the AI would otherwise hear wrong (is the CEO Kathy or Kati?) are managed in the Keyword bank section on this page. The bank is account-wide and shared with automatic transcripts. See Keyword bank for how to set it up.

FAQ

/* BLACK BACKGROUND FOR THE SUBTITLES */
.bmpui-ui-uicontainer .bmpui-ui-subtitle-overlay {
    display: inline !important;
}
.bmpui-ui-subtitle-label {
  display: inline !important; /* Ensures it doesn’t force full width */
}
.bmpui-ui-subtitle-label span {
  display: inline !important;
  background-color: rgba(0, 0, 0, .75) !important;
  padding: 7px 5px 2px 7px !important;
  line-height: 1.6 !important;
  font-family: Arial, Helvetica, sans-serif !important;
}