The closed-powered by AI can open new possibilities-and risks

The closed illustrations have become an essential component of watching TV and movies. For some, it’s a way Decoding the dialogue. For others, such as those who fast or difficult to hear, it is very important Access. But the illustrations are not perfect, and technology and studios are increasingly looking for artificial intelligence to change this.

Still by real people, who can help ensure accuracy, maintain accuracy, maintain accuracy, maintain accuracy, maintain accuracy, maintain accuracy, and maintain clarification slides of television shows, still still still still still still still still still does, but there are challenges. Anyone has seen a direct event with closed illustrations that know that the text on the screen often falls backward, and there can be errors in the process of the operation. Textual programming provides more time for accuracy and details, but it can be a thick process of employment-or in the eyes of studios, which is an expensive process.

In September, Warner Bros. announced. Discovery is cooperating with Google Cloud Development of closed illustrations that work in Amnesty International“Besides human supervision to ensure quality.” In a press release, the company said that the use of artificial intelligence in the explanatory designation is up to 50 %, and reduced the time it takes to clarify a file of up to 80 %. Experts say this is a peek at the future.

“Anyone who does not do it is just waiting for displacement,” Joe, Devon, a lawyer for access to the web and a co -founder for him The day of awareness of global accessibilityHe said about the use of artificial intelligence in the illustrations. The quality of manual labels today is “somewhat everywhere, and it definitely needs to be improved.”

As Amnesty International continues to transform our world, it is also Re -reshaps how companies deal with access. Google Expressive comments The feature, for example, is used artificial intelligence to better transfer passion and tone in videos. Apple added Copies For iOS 18, which multiplies as ways to make sound content easier. both of them Google and apple Ownership Real -time label tools To help people deaf or hard hearing Access to sound content On their devices, Amazon added Text features to speech and illustrations To Alexa.

In the entertainment area, Amazon launched an advantage in 2023 called Promoting dialogue In the main video, which uses artificial intelligence to determine and enhance speech that may be difficult to hear the back music and its effects. The company also announced a experimental program in March that uses artificial intelligence to accept films and TV programs, “It was not possible to call it” Blog post. In a sign of the extent of the collective viewers who became an illustrative name, Netflix in April put a The option of translations of dialogue only Anyone simply wants to understand what is said in the conversations, while leaving healthy descriptions.

As development continues, and while we consume more content on both large and small screens, it is just a time before more studios, networks and technology companies that benefit from the capabilities of artificial intelligence – while remembering the reason for the existence of closed illustrations in the first place.

Maintain access in the foreground

The development of the closed illustrations in the United States started The scale of access in the 1970sUltimately, everything from direct TV broadcasts to movies films are made more fair to a wider audience. But many viewers who are not deaf or difficult to hear also prefer to watch movies and TV programs with illustrations – which are also referred to as translations, although this is technically related to linguistic translation – especially in cases where production dialogue is It is difficult to decode.

Half of the Americans Suppose they usually see the content with translations, according to a survey of 2024, which was previously conducted by the language learning site, and 55 % of the total respondents said it became difficult to hear dialogue in movies and shows. These habits are not limited to the older viewers; Yougov 2023 poll found it 63 % of adults under the age of 30 You prefer to watch TV with translations – compared to 30 % of people between the ages of 65 years and over.

Ariel Sims, President and CEO of The deficit belongs. But the illustrations can make it easy for anyone to process and keep information.

By accelerating the illustrations process, artificial intelligence can help make more content available, whether it is a TV show or a clip of the film or social media, or SIMMS notes. But the quality may suffer, especially in the early days.

“We have a name for the explanatory names created from artificial intelligence in the disability community-we call it”prattleTions, “SIMMS laughter.

This is because the automatic illustrations are still struggling with things such as punctuation, rules and appropriate names. Technology may not be able to capture different dialects, dialects or styles of speech in the way a person desires.

Ideally, SIMMS said that companies that use artificial intelligence to generate illustrations will still have a person on board to maintain accuracy and quality. Studios and networks should also work with the disability community to ensure that the possibility of access to this process is not exposed.

“I am not sure that we can completely take people out of the operation,” said Sims. “I think technology will continue to improve and improve. But at the end of the day, if we don’t cooperate with the disability community, we leave an incredibly important perspective on all these access to tools.”

For example, studios such as Warner Bros. confirm. Discovery and Amazon, on the role of humans in ensuring accuracy that is on behalf of artificial intelligence.

“You will lose your reputation if you allow AI SLOP to control your content,” Devon said. “This is where a person will be in the episode.”

But given how quickly the technology development is, human participation may not last forever, as expected.

“The studios and broadcasters will do everything that costs less, and this is certain,” Devon said. But he added: “If technology is able to help do this task better, then who is anyone standing on the way?”

The separation line between a joint and a overwhelming

It is not only TV and movies where artificial intelligence is clarified. Like social media platforms Tikhak and Instagram Automatic planning features to help make more content are available.

These original illustrations often appear as an ordinary text, but sometimes, creators choose shiny display screens in the editing process. The common “karaoke” style includes highlighting each individual word where it is spoken, with different colors of the text. But this most dynamic approach, while resorting to the eye, can harm the ability to read. People cannot read in their pace, and all colors and movement can be distracted.

“There is no way to make 100 % of the users happy with the illustrations, but only a small percentage that you benefit from and prefer the karaoke style,” he said. Merrill K. EvansAccessibility Marketing Adviser, Deaf. She says she should watch videos with dynamic illustrations several times to get the message. “Empting labels are boring. They allow the video to be the star.”

But there are ways to maintain simplicity with the addition of a useful context. Google AI’s expressive explanatory features are used to emphasize some sounds and give viewers a better idea of ​​what is happening on their phones. Enthusiastic “Happy Birthday!” It may appear in all hats, for example, or the enthusiasm of the sports broadcaster may be transferred by adding additional messages on the screen to say, “Amaaazing Shot!” Expressive labels also look like applause, plow and whistling. All text appears on the screen in black and white, so it does not pay attention.

Access was a basic axis when developing the feature, but Angana Ghosh, Director of Android Product Management, said that the team realizes that users who are not deaf or difficult to hear will benefit from its use as well. (Think about all the times I was in public places without headphones, but you still want to follow what was happening in a video, for example.)

“When we develop for access, we are already building a much better product for everyone,” says Gush.

However, some people may prefer more vibrant comments. In April, the FCB Chicago ads agency appeared for the first time on a platform that works with a business called An explanatory name with the intentionWhich uses animation, color, and changing printing to transmit passion, tone and speed. Distinguished text colors represent different characters lines, and the words are distinguished and synchronized with the actor’s speech. The conversion of the sizes of the type and the weight helps in transferring the voice of a person, as well as improving them. The open source platform for studios, production companies and broadcasting platforms is available for their implementation.

FCB has a partnership with the Chicago Hearing Association to develop and test the differences suspension with deaf and hearing people. Bruno Mazoti, CEO of FCB Chicago, said his experience that he created by deaf religion also helped form the platform.

Mazoti said: “The closed illustrations were part of my life. It was a decisive factor for what we will see as a family.” “After hearing privilege, I could always notice when things did not succeed well,” he pointed out, as it was when the illustrations were lagging behind the dialogue or when the text stumbled when many people were speaking once. “The main goal was to bring more feelings, speed, tone and the identity of the speaker to people.”

In the end, Mazoti said, the goal is to provide more customization options so that viewers can adjust the intensity of the illustration. However, this approach may be the most enthusiastic fee for some viewers, and it may make it difficult for them to follow what is happening on the screen. In the end, it is summarized in personal preference.

“This does not mean that we must categorically reject these methods,” said Christian Voeller, director of the Technology Access Program at the University of Galoudette. “But we need to study carefully with deaf and difficult to hear viewers to ensure that it is a clear benefit.”

There is no easy solution

Despite its current faults, artificial intelligence can eventually help expand the scope of the availability of an illustration and provide greater customization.

YouTube Automatic cutting An example of how, Despite the approximate startAmnesty International can make more accessible video content, especially since technology improves over time. There can be a future in which the explanatory designations are designed for different levels of reading and speeds. Non -speech information can become more descriptive as well, so instead of public stickers such as “Scary Music”, you will get more details that transmit mood.

But the learning curve is sharp.

“The illustrations of artificial intelligence are still worse than the best in human nomenclature, especially if the quality of the sound is at risk, which is very common on both TV and films,” said Vojler. Halosa can also serve inaccurate illustrations that end the isolation of deaf and luxurious viewers. For this reason, humans must remain part of the illustrative designation process.

Dibora Fils, director of the Information and Comprehensive Design Center at the University of Toronto Metropolitan, said that what will happen most likely is that Jobs will adapt. She predicts that she will supervise the work that was before artificial intelligence that will be involved in artificial intelligence.

“Now, we have a different type of function needed in the illustrative naming,” said Fils. “Humans are much better to find errors and determine how to correct them.”

Although the artificial intelligence of the designation is still an emerging technology limited to a handful of companies, this is likely not for a long time.

“They all go in this direction,” Fels said. “It is a matter of time – not a lot of time.”