Note: This blog post was originally written in Japanese for our Japanese website. We used our machine translation platforms to translate it and post-edit the content in English. The original Japanese post can be found here.
From video-on-demand streaming services like Netflix and Amazon Prime to social media ads, product manuals, and eLearning, video content is now used across virtually every aspect of business.
When delivering this content to audiences around the world, the enormous costs and time involved have long been major barriers. However, thanks to dramatic advances in AI, there are now more options for achieving multilingual delivery with high quality, low cost, and fast turnaround.
In this post, we introduce a range of multilingual solutions tailored to your needs, from traditional methods to the latest ways to leverage generative AI.
Until now, subtitle translation in professional settings has required all of the following steps to be done manually.
• Transcription: Listening to the audio in a video and converting it into text.
• Translation: Professional translators interpret the context and translate it into another language.
• Timecode adjustment (spotting): Timing the subtitle display to match the video down to the second.
• Dubbing (optional): Hiring voice actors and recording and editing the audio in a studio.
When all of these tasks were done by humans, it was not uncommon for even a 10-minute video to require several days and cost thousands of dollars. However, with the introduction of AI, it is now possible to automate more than 80% of these processes.
Today’s AI-driven efficiency gains go far beyond simple “translation.” Let’s look at three key ways AI is dramatically transforming the entire workflow.
Using high-performance speech recognition AI, even videos containing multiple languages can be instantly transcribed with over 95% accuracy. Technical terms and proper nouns can also be handled through advance training and custom dictionaries.
*If you already have subtitle files in a format such as SRT, you can skip this step and reduce costs even further.
Using the latest tools, such as ElevenLabs and HeyGen, it is possible to generate speech in another language while preserving the speaker’s own voice quality.
Furthermore, lip-sync technology, which automatically adjusts lip movements to match the translated language, has also helped eliminate the unnatural feel often associated with dubbing.
Unlike conventional machine translation (which tends to be literal), the latest generative AI can now understand context to some extent, including who is speaking, who the message is intended for, and what tone is being used. As a result, it is increasingly possible to produce natural, fluent translations.
While generative AI excels in fluency, research has shown that machine translation can be more accurate. (If you’re interested, read more in our post: The Surprising Relationship Between ChatGPT and Machine Translation (Part 3).) In addition, there is a risk of hallucinations, in which AI generates factually incorrect information, so using AI output as-is still raises serious quality concerns.
If you’re thinking, "We want to improve efficiency with AI, but accuracy is absolutely non-negotiable," that’s exactly the challenge our patented technology is designed to solve.
Our AI post-editing process is a new translation service that combines the strengths of machine translation, generative AI, and human translators.
• Machine translation: First, produce an accurate initial translation as a solid foundation.
• Generative AI: Automatically refines grammar, terminology, and style to improve fluency.
• Professional translators: Perform a final human review to ensure accuracy.
These three steps make it possible to maintain the highest level of quality while achieving both speed and cost savings.
The process of having a translator revise machine translation output or AI-generated translations is called post-editing. By adjusting the level of revision and review, we can develop the optimal plan for each use case.
Rigorously reviewing if any of the content touches on cultural taboos in the target region and whether it aligns with the brand image.
Even for high-volume content, implementing a quick review for mistranslations and omissions while keeping costs to a minimum.
Video translation is no longer expensive or out of reach. By using AI wisely and adding a human touch, we can help deliver your message to audiences around the world.
If you have any questions, such as the cost of translating a video or how to strike the right balance between AI and human translation, please feel free to reach out.
Kawamura International provides high-quality localization services for video content. We support both traditional subtitle translation processes and new AI-powered workflows, optimizing everything from transcription and translation to subtitle generation, helping reduce costs and shorten turnaround times.
We propose the optimal translation approach based on your content’s intended use and quality requirements. Contact us today to get started on your project.