2023年2月17日
By 詹姆斯Broberg StreamShark创始人兼首席执行官
博客

视频字幕难题

今天, we are consuming more video than ever before:

In 2022, 全球82%的互联网流量将来自视频流媒体或视频下载, 根据思科.
每天84分钟 is the global average for video consumption, 根据 MediaSpot
据估计，到2027年，全球直播市场将超过2470亿美元美通社.

有了这样的统计数据，很明显视频是我们世界的重要组成部分，从个人到专业.

不幸的是, while most of us engage with video quite frequently in our everyday lives, from live-streamed conferences to internal meetings at work, 以下数据证实，有几个常见的问题会干扰观众对录制视频的理解:

80%的观众 aged 18-24 and 64% of viewers aged 26-35 绝大多数人更喜欢看带字幕的视频内容(StageText UK的2021年研究结果)
80%的观众 in the UK using captions were not d/Deaf or hard of hearing (Ofcom study).

第一个, it’s often difficult to understand a video when there are multiple presenters, 或者是技术故障干扰了录音. While many attempt to resolve this with captioners, human captioners are costly and must sink hours into captioning. 这就是我们所说的字幕难题:视频字幕对于可访问性和理解性至关重要, 但它对时间很敏感，而且价格昂贵. 如果需要多种语言的字幕，转换字幕的时间和成本就会增加.

为什么我们需要视频字幕?

首先，回答这个问题很重要:为什么你需要视频字幕? 今天, 对视频的字幕和转录的需求至关重要，主要有两个原因:视频内容的可访问性和提高观众的理解力.

美国残疾人法案于1990年通过，以保护残疾公民不受歧视. Many associate 《欧洲杯在线投注赔率》 landmark legal requirements like accessible parking spots, 建筑的入口通道, 还有洗手间和饮水机. 然而, 《欧洲杯在线投注赔率》也需要“辅助设备”,，比如字幕或音频描述, to be made available to anyone with a disability for two groups:

公共实体:国家和地方政府，在内部和外部视频通信.
公共场所:公共或私人企业供公众使用的场所. Private clubs and religious organizations are exempt.

Adding captions to pre-recorded video conferences, 演讲, and meetings is one of the best ways to ensure 每一个人 can enjoy, learn from, and engage with great video content. As an organization using VOD for internal and external meetings or events, adding captioning is adding accessibility to your workplace. 另外, 在组织的视频点播中添加字幕对于提高观众的理解力至关重要.

正如我之前提到的, 在流中出现故障是很常见的, 演讲者的网络有问题, 或者口音或沟通风格的差异(想想:在与不同利益相关者的会议上使用许多行业特定的形容词)会干扰视频观众对演示或会议的理解.

在视频点播中添加字幕可以让观众更容易理解正在发生的事情, especially when there are multiple speakers or presenters.

There are other key non-accessible focused reasons to add captions. They can improve the organic reach of your video content, 对内对外, more than just the standard video meta-data of title, 描述和标签. 它还可以提高观众在安静的办公室等对声音敏感的环境中观看视频内容的参与度, 或者公共交通.

标题难题

不幸的是, 许多组织都非常熟悉在视频中添加字幕所涉及的三大问题:高成本, 长, 工人的人工工时, and the likelihood of human error as the quality of captioners can vary.

从历史上看, 视频会议或会议结束后, 公司将雇佣一个人来听录音，并手动转录字幕. 后来，字幕被添加到录制的视频中，并作为书面转录提供.

While human captioning is very helpful, it is far from efficient or cost-effective. Human transcribing captions are quite expensive, especially when hiring a high-quality transcriber, 这是我们很多人想要做的吗. 另外, manually transcribing is a time-consuming, tedious process. This manual process can delay the release of a recorded video, 并使重要信息过时. 然后，还有人为失误的问题. 人类转录员容易出现不可避免的错误或打字错误，因为，他们是人!

So, captioning on videos is critical for user accessibility and comprehension. 但, 雇佣一名转录员来制作字幕是昂贵的，而且可能会拖慢视频的发布, thanks to 长 hours the transcriptionist must put in.

What’s the solution to the captioning conundrum then?

以交付高品质, 快速准确的字幕, 我们必须利用人工智能和机器学习技术的力量，为视频自动创建字幕. 当今世界对视频的需求太大了，无法继续以其他方式解决字幕问题.

机器学习字幕功能

机器学习是自动为录制视频创建字幕的关键 分钟.

然而, not all AI-powered captioning services are the same. In order for captioning to be the most efficient and effective, 字幕功能必须具备三个关键特性:可编辑的字幕, the ability to be tailored to different dialects, 还有一个可定制的词汇库, so the ML can be trained to recognize key terms like industry slang, 常用缩写词, 以及演讲者的名字.

可编辑的标题

第一个, 所有优秀的机器学习字幕程序都应该提供自动添加字幕的选项. 这使得最快的周转，并确保视频内容可以交付，同时仍然相关和及时.

然而, a great ML captioning program will 也 allow a user to 编辑发布视频前的字幕. 当视频是高优先级并且需要在字幕中没有任何错误或错别字时，编辑字幕变得至关重要. 当敏感信息需要在发布前从视频中删除时，编辑字幕也很有用.

带有可编辑的说明文字, 用户可以下载字幕, 修复错误, 根据需要删除敏感信息, then upload and distribute the most polished, 全字幕视频.

根据方言定制

Another great feature your ML captioning program needs? The ability to tailor the captioning to understand different dialects

This feature is best for helping viewers gain full comprehension of a speaker, and is especially useful when speakers are speaking the same language, 但也有地区性的调整.

例如, think about all of the nuances and differences between American English, 英式英语, 澳大利亚英语, 印式英语, 爱尔兰的英语, 苏格兰英语, 和威尔士英语. While a Brit saying “knackered” in a meeting, while an American English speaker would use the term “tired.带有可编程ML字幕服务, different dialects can be programmed into the captioning, to most accurately reflect differences in word choice.

可定制的图书馆

最后, 在您的ML字幕平台中寻找的另一个关键功能是创建可定制的词汇库的能力.

This is especially helpful for industry abbreviations, 俚语术语, or even presenter names that 每一个人 watching a video might not understand. This customizable library functions helps to improve comprehension for VOD viewers, and is particularly useful for internal meetings with a lot of stakeholders.

另外, 特别定制的单词库将创建最准确的标题, which cuts down on the time needed to 编辑 captions and ultimately release a video.

**********

字幕是当今视频的核心部分. When used correctly, it increases accessibility and comprehension for our video. 创造性技术可以帮助组织以高效的方式生成清晰、正确的标题.

There’s no need for a conundrum when it comes to captioning.

詹姆斯Broberg是 StreamShark，端到端直播和点播服务 企业视频流媒体平台. Prior to StreamShark, James was a research fellow in content delivery networks and cloud computing at the University of Melbourne.

[Editor's note: This is a contributed article from StreamShark. 流媒体 accepts vendor bylines based solely on their value to our readers.]