VoCaption for Broadcasters

VoCaption is an AI based, live, automated captioning and subtitling solution that provides automated speech-to-text processing of live video to display highly accurate live subtitles in real time without human involvement.

VoCaption uses a speech-to-text processing engine based on AI or artificial intelligence that “listens” to a live audio feed and converts the spoken language into words on the screen so viewers can read the subtitles while they watch the program.

Artificial Intelligence has improved dramatically over the last few years thanks to advancements in computer hardware and software development and along with these improvements accuracy is much improved over just a few years ago and will continue to improve in the future. However, accuracy also depends on the quality of the audio; the presence of background noise or music; the number of speakers; whether the speakers talk over each other and how fast they talk. As an example, a single speaker talking at a normal speed will generate accuracy levels in the high 90’s. As you add in noise or other speakers the accuracy level will decrease and this is also true with human captioners.

99% accuracy depends on a number of things including quality of the audio, the number of speakers, cross talk by those speakers, background noise, music, use of localized or foreign words etc. Given those variables neither machine nor human can maintain 99% accuracy for a full, live broadcast.

The FCC rules for US broadcasters do not provide an accuracy benchmark but instead say:
FCC rules for TV closed captioning ensure that viewers who are deaf and hard of hearing have full access to programming, address captioning quality and provide guidance to video programming distributors and programmers. The rules apply to all television programming with captions, requiring that captions be:

• Accurate: Captions must match the spoken words in the dialogue and convey background noises and other sounds to the fullest extent possible.
• Synchronous: Captions must coincide with their corresponding spoken words and sounds to the greatest extent possible and must be displayed on the screen at a speed that can be read by viewers.
• Complete: Captions must run from the beginning to the end of the program to the fullest extent possible.
• Properly placed: Captions should not block other important visual content on the screen, overlap one another or run off the edge of the video screen. 

Obviously, our goal is to provide the most accurate speech-to-text conversions possible, given the quality of the audio. To that end, we provide two additional ways to improve accuracy:

• Custom Dictionary: A custom dictionary is available to include local, regional and hard to pronounce words, names and geographical places and references, including foreign names and words so you can pre-load these items into the system.

• Alternate Words Dictionary: In addition, we provide an alternate words dictionary that you can populate to swap out words you don’t want going to air, ensure certain words are capitalized or referenced properly to match the needs of your audience.

• Additionally, the entire vocabulary of the speech engine is updated quarterly to ensure continued improvement.

Yes, there is an overlay box that you can select with a mouse to reposition the subtitles manually.

The system will respond to GPI’s, UDP and HTTP triggers from a traditional button box, keyboard hotkeys or from Secondary Events within OASYS Integrated Playout. When integrated with both Polistream and OASYS the system can sense the absence of captions and turn on automatically for emergency captioning and off again when captions return.

Currently the system provides 31 total languages including Global English. The number of languages will grow as the system matures and expands.

VoCaption’s software is available as follows:

• • In a 1RU rackmount unit to work with 3 party subtitle encoders.
  • As software only integrated with Polistream and/or OASYS.
  • As software only with customer provided hardware that must match a specific and supportable platform.
  • Once the software is installed you can purchase minutes in blocks with 25 hours being the minimum purchase and minutes can be re-purchased as needed.

Lead times depend on which installation option you choose. It takes about 4 hours to commission the system including about 90 minutes for training. All commissioning is done remotely. During commissioning we can assist with the creation of the Custom and Alternate Words dictionaries to ensure you get off to a fast and accurate start.