Captions beta

Q: Which platforms show CTA-708 closed captions?

Closed captions travel inside the video as CTA-708 data, so what viewers see depends on the platform's player. YouTube and Twitch can display embedded captions and let viewers toggle them. On YouTube, embedded 608/708 captions also have to be enabled in the stream settings. Support on other platforms varies, and some strip caption data on ingest. If you need captions to appear everywhere without exception, use one of the burned-in modes instead.

The Captions element turns speech into live subtitles in real time. Captions can travel with the stream as closed captions that viewers switch on and off, or be drawn directly into the video as burned-in captions, also called open captions, so they show everywhere.

How do I add live captions to my stream?

Place the Captions element between your input and output in the Streamrun Editor. Speech recognition runs in the cloud on the live audio, so nothing is needed from your streaming software or encoder, and no subtitle file needs to be prepared in advance. Every output downstream of the element receives the captions.

Recognition happens as you speak, so captions appear a moment after the words are said. Speaking clearly with a good microphone and low background noise gives the best results.

Settings

Element title

A name for this element in the editor. Useful when a configuration has more than one Captions element, for example one for closed captions and one for burned-in captions on a different output.

Primary language

The language you mainly speak. It is used as a hint for the speech recognition, and other languages are still recognized. Leave it on Auto-detect if you are unsure. Bilingual options are available for Malay, Mandarin, Tamil, and Tagalog mixed with English.

Custom terms

Comma-separated names and words that should be spelled exactly as written, for example your channel name, guest names, product names, or in-game terms that the recognition would otherwise get wrong.

Caption output

How the captions reach your viewers. See the comparison below.

Sync video with captions

Off by default. Speech recognition takes a moment, so captions naturally appear slightly after the words are spoken. Turn this on to delay the video so captions line up with the speech. The stream reaches viewers a little later as a result, so leave it off when low latency matters more than caption timing, for example when you interact with chat.

Caption output modes

CTA-708 closed captions

Captions are embedded in the video stream as standard closed captions. Viewers turn them on or off in the player, and the video itself stays untouched. They appear on platforms and players that support embedded captions and support varies by platform.

Burned in, multiple lines

Captions are burned into the picture as a block of text, two lines at a time. They work on every platform and player because they are part of the video, but viewers cannot turn them off.

Burned in, word by word

Captions are drawn into the video one word at a time as they are recognized, in the style popular on short-form video. Like the multi-line mode they show everywhere and cannot be turned off.

Ideas and use cases

Accessibility for deaf and hard of hearing viewers

Viewers watching with the sound off on mobile

Live subtitles without a separate captioning service

Talks, panels, and events with several speakers

Reaching viewers who read your language better than they hear it

Subtitled clips and replays cut from the recorded stream

Checking captions before going live

Each output element has a Preview captions toggle, on by default, that shows the closed captions generated by the Captions element in the editor's preview window. Use it to check recognition quality and the effect of Custom terms before the output goes live. Burned-in captions are part of the picture, so they are visible in the preview regardless of this toggle.

Pricing

Live captions are a paid add-on. Each Captions element costs $0.40 per hour on top of the configuration's base price, billed per second while the configuration is running in live mode.

See add-on pricing for the full list of add-ons and billing and credits for how usage is charged.

Beta

Captions are in beta. Recognition quality depends on audio quality, accent, and vocabulary, and the feature is still being tuned. If captions come out wrong for your content, add the problem words to Custom terms and let us know.

Frequently asked questions

Which platforms show CTA-708 closed captions?

Closed captions travel inside the video as CTA-708 data, formerly known as CEA-708, so what viewers see depends on the platform's player. YouTube and Twitch can display embedded captions and let viewers toggle them. On YouTube, embedded 608/708 captions also have to be enabled in the stream settings. Support on other platforms varies, and some strip caption data on ingest. If you need captions to appear everywhere without exception, use one of the burned-in modes instead.

Do captions work with OBS, a phone app, or a hardware encoder?

Yes. Recognition runs on the audio Streamrun receives, so it does not matter what produced the stream. OBS, Moblin, IRL Pro, Larix, a BELABOX device, or any RTMP or SRT encoder works without plugins or extra software.

How much delay do captions add?

Captions appear a moment after the words are spoken because recognition happens live. By default the video is not delayed to wait for them. Turn on Sync video with captions to delay the video so captions line up with the speech, at the cost of slightly higher stream latency.

Which languages are supported?

The Primary language list has 55 languages plus four bilingual modes, and other languages are still recognized when they come up. Set Primary language to the one you mainly speak so recognition has a hint, or leave it on Auto-detect. Bilingual options exist for Malay, Mandarin, Tamil, and Tagalog mixed with English.

Can I fix words the recognition keeps getting wrong?

Add them to Custom terms as a comma-separated list. Channel names, guest names, product names, and in-game terms are the usual candidates. Custom terms are spelled exactly as written whenever they are recognized.

Can I use captions on only some destinations?

Yes. Every output downstream of the Captions element receives captions. Place the element on the path to the outputs that should have captions and leave other outputs connected directly to the input, or use a burned-in mode on one path and closed captions on another.

How are captions billed?

Each Captions element costs $0.40 per hour as an add-on on top of the configuration's base price, billed per second while the configuration is running in live mode. See also add-on pricing.

Do captions end up in the recording or VOD?

Burned-in captions are part of the picture, so they are in every recording and VOD. Closed captions are kept only where the platform preserves CTA-708 data in its recordings, which varies by platform.