Delivering live streams with AI closed captions and speech translations
This section is part of the AI Live Streaming Manual. It explains the basic workflow for delivering a live stream with multilingual AI-generated closed captions and speech translations.
The captions and speech are generated using AI speech-to-text conversion. By default, this happens automatically, without any human intervention. To increase accuracy, you can let a human make corrections in real-time to the text resulting from speech-to-text conversion (see the speech-to-text correction manual).
1 Creating an event
Use the AI Multilingual Event wizard to create a new event. This will let you configure your broadcast, select the AI closed captions and speech translations and add initial vocabulary terms.
When the wizard has finished, you are redirected to the event page. On this page, you can still change all aspects of the live stream configuration.
To embed the video player on your website or third-party platform, go to the event's Management tab and copy the embed code.
2 Test broadcast
To start testing, go to the event's Management tab and press the 'Set to preview' button. The event status must be set to Preview for your broadcast to be received.
Next, go to the event's Broadcast tab to configure your encoder or web application. Start your broadcast, in accordance with our broadcast guidelines.
Use the 'Preview player' on the event's Management tab to test the closed captions and speech translations. The live stream in the preview player has a delay of 1-2 minutes. As soon as you start broadcasting, the preview player will indicate the number of seconds before the live stream is available.
Optional:
- If you are using real-time correction, corrector(s) will be able to connect to their rooms when the event is set to
Preview. While the live stream is being processed, correctors are able to see it and correct the captions. - You can also use the Real-time Management page when the event is set to
Preview.
3 Starting the live stream
When the event is about to start, go the Management tab and press the Start event button. This will make the live stream visible in the embedded player. Because of the latency, we recommend to start the event at least four minutes before the live action begins. This allows the player to start buffering and ensures that nobody misses the start.
When you start the event, Clevercast also starts recording the live stream. So, starting the event in time also ensures that the cloud recording is complete.
The embedded player automatically detects the status change and makes the live stream visible to your viewers. You can also see the live stream in the 'Public Player' on the Management tab. If you want to display a message in the embedded player, use the 'Service Message' on the Management tab.
4 Stopping the live stream
After the live action has ended, wait at least 4 minutes before setting the event status to Ended, by pressing the Change Event Status button and selecting End event. The embedded player automatically detects this and shows a poster image or message to your viewers (see the Appearance tab) instead of the live stream.
Clevercast automatically completes the cloud recording and converts it to an MP4 file. This is another reason to wait at least two minutes (ideally four) after the live action ends before pressing End event. Otherwise, the final closed captions and speech translations may be missing (since processing is delayed).
You can download the cloud recording via the Events > Recording menu. If your plan includes VoD, you can publish the recording to Video on-Demand (this also allows you to turn the multilingual recording into single-language MP4's).
5 Pausing the live stream (optional)
You can set the event status to Paused during breaks. The embedded player automatically detects this, and shows a poster image or message (see the Appearance tab). While the event status is Paused, you can continue to broadcast and AI captions and speech will continue to be generated (unless you also pause them).
Of course, you can also leave the status to Started during breaks. The advantage of using Paused is that Clevercast stops recording when the event is paused, and starts a new recording when the event is restarted. This way, you have separate recordings for each session.
Therefore, you should also wait four minutes after/before the action stops when changing the event status to/from Paused (see Starting & Stopping the live stream).
Notes
An AI event has a maximum duration of 24 consecutive hours. If your event spans multiple days, you can overcome this by setting the event to Inactive or Ended during breaks and reset it again. If this isn't possible, contact us.
The maximum duration of a cloud recording is also 24 hours. Again, you can simply reset the event status to start a new recording.
An AI event can be combined with local and remote simultaneous interpretation and subtitling by humans (for specific languages). These workflows are largely similar.