AI overview
Subtitles are crucial for ensuring that video content can be accessible to all audiences. As AI speech recognition tools become more sophisticated, it is possible to generate transcripts and subtitles by simply uploading a video to a website.
It is likely that the role of a subtitler will change from one of transcribing, to one of proofreading and checking the synchronisation of generated subtitles, and potentially introducing some creative interpretation for the tone and style of the content into their work.
Additional work may also be required to provide notes on the delivery and to highlight other audio that may require a subtitle that is missed or misunderstood by automated tools, ultimately configuring the AI to deliver consistent and high quality results.
In the future AI may also be used as a tool to enable a subtitler to enhance the creative aspects of subtitling, allowing enhanced timing to lip sync, or context aware creative styling and animation.
This could be applied to the subtitles by constantly adjusting, in real time, for optimal contrast or readability, or adjusting scale and location dependant on the user's needs, or moving locations during important aspects of onscreen action.
This can also be combined with eye tracking and other reactions so that an AI can respond subtly to a viewer's input, for example fading the subtitles when they are no longer needed. All of these elements will need to be tested and checked for accuracy and a subtitler will need to make sure they fit with the tone and style of the content.
Download a PDF version of this page.
How can I prepare for the future?
AI can currently provide good results under the right conditions and can be used as a first draft for a human to correct any errors.
People are still typically relied upon where they are to provide field-specific expertise or governmental, legal and medical transcriptions where precision and accuracy are critical and a personal review is considered essential in the pipeline.
Captioning events with high background noise is also challenging for AI, often reducing transcription accuracy. Stenography, using specialised machines that allow fast "chorded typing" by pressing multiple keys for syllables or words, remains a reliable solution.
Unlike current generations of AI, stenographers can distinguish speech from background noise through contextual understanding. While AI struggles with sound pattern recognition in noisy environments, human stenographers excel.
Improvements are being made to the AI solution by using many different techniques. Future AI systems (where multiple AI systems are layered together), such as noise reduction and directional microphone arrays, contextual text understanding, real-time video (lip reading) and machine learning-based speech to text solutions, will likely resolve many of these issues in the coming years.
Learning how to manage these new layered AI tools within the existing captioning workflow, preparing the input content for future AI tools and developing the creative potential of subtitling will be a valuable skill for future subtitling work.
What AI tools can I use right now?
AI tools allow users to upload video content and generate captions in many different languages, automatically translating the content.
Editing tools are now adopting these features as well, which can allow editors to directly generate subtitles at the editing stage.
Some tools can generate subtitles in real time and automatically overlay them on top of live streams for most popular platforms, in multiple languages.
Visual speech recognition (VSR): VSR technologies are able to generate a transcription of a video based on the movement of lips from a video.
In context understanding (visual and language): AI systems will be able to recognise an environment and context within a sentence and paragraph and use that to adjust their ability to predict hard to hear words in a noisy environment.
Real-time collaboration platforms: Cloud-based platforms that allow subtitlers to share subtitles and collaborate with multiple editors, handle approvals and implement version control easily.
Creative subtitles: Subtitling that goes beyond simply conveying words to the viewer, but enhances the immersive experience by adding tone and contextual elements to the work.
ScreenSkills offers a variety of training opportunities for people at all stages of their career. Explore all training, events and opportunities.
ScreenSkills resources:
- Subtitler job profile
- Subtitler post production job profile
Other resources:
- Papers and implementations of visual speech recognition can be found online, with testable we implementations also available
- Frame.io and other cloud-based systems provide tutorials on real-time collaboration