Automated Subtitling Solutions
Andovar’s AI-based machine translation solutions are fully trained for television and media applications, providing you with accurate translated subtitles in hours rather than weeks.Contact Sales
Putting Artificial Intelligence to Work for You
When you’ve got thousands of hours of content to localize, automation is the only answer. Andovar’s intelligent automated subtitle solutions ensure that you can go to market quickly with accurate, culturally aligned translations for all the shows you want to broadcast. Reach TV audiences around the world in record time — no matter how high your volume.
Automated Subtitle Solutions Designed for OTT Media
Media Studio offers a comprehensive set of enterprise-class tools for managing, creating, translating, editing, and processing subtitles, captions, and videos. Built on a foundation of high-quality subtitle-optimized machine translation engines, Media Studio has already delivered significant productivity and translation quality gains over hundreds of thousands of subtitles processed to companies all across the world.
- Extraction of clean dialogue in minutes
- Removal of unwanted noise by artificial intelligence
- AI-driven sentence splitting and structuring
- Supports Microsoft Word/Excel, RTF & PDF
- Speech recognition and artificial intelligence matches dialogue text in timed-sourced templates
- Minimal edits required for finalization
- REST API & connectors
- Custom workflow capabilities and advanced automation tasks
Translate More for Less
Media Studio features a suite of tools and technology with the sole purpose of making translators' lives easier. Media Studio's state-of-the-art Hybrid Deep Neural/Statistical Machine Translation optimizes machine translation specifically for subtitles helps deliver high-quality translations, typically with 50-90% of subtitle sentences either perfect or requiring minimal edits.
- In-browser video players
- Visual subtitle editing and creation
- Simultaneous second-screen video review
- User configurable hotkeys
- Hybrid Neural/Statistical Machine Translation engines
- Fully customizable with client subtitle data
- Progressive quality improvements via machine learning AI
- Glossary/terminology analysis
- Bilingual glossary creation
- Proven to cut tasks from days to hours
- 200 - 300% productivity gains as standard
Technology that Works Overtime for You
However large your project, Andovar has the expertise – and the technology – to handle it efficiently. We will meet your high-volume and high-quality requirements on a global scale. Take your high-volume content to market in a matter of days and hit all your targets. Andovar’s intelligent, automated translation technology works overtime for you.
Comprehensive project management
- Real-time progress tracking in Gantt charts
- Productivity metrics reporting
Team collaboration tools
- Multiple editors can work on files simultaneously
- Context-linked messaging
- Chat & video calls
- Manage and track subtitles, video and audio
- Maintain terminology and associated lists
- The client retains full control over data movement
- Data remains in a designated AWS or private server
Challenges of Automated Subtitling
While machine translation has been used in localization for a long time, a few new challenges are present in subtitling projects:
For both humans and machines alike, not knowing when one sentence ends and the next sentence starts provides a challenge. Most subtitles lack end-of-sentence markers, and most have sentences that are split between frames; these issues need to be fixed before translation.
Stylistic requirements would normally be specified in a style guide for human translators, but they can also be applied through rules and machine learning for an automated process. As an example, profanity may be quite acceptable in the US market, but the vocabulary must be toned down considerably for a Thai-speaking audience by instructing the MT engine to soften any profanity found.
Translation can result in shorter or longer text strings compared to the source language. Timing cues need to be adjusted in each frame, and sometimes frames must be inserted or deleted to accommodate these length changes.
As the input to MT is complete sentences, the output from MT is also complete sentences. It must be reformatted into a subtitle format such as SRT or TTML.
Like what you see?
Solutions for Automated Subtitling
In addition to the actual translation, there are several processes surrounding the creation and localization of subtitles that typically require significant human efforts. Many of these processes can be automated by Media Studio:
Before a subtitle can be translated, a source language subtitle is usually made available. The layout and structure for each screenplay is wildly different, and even a human can take considerable time to extract dialog from the surrounding noise of video directives, mood indicators, general time stamps and other supporting information. Fortunately, these tasks can be automated – saving time and cost, and minimizing human error.
The second means of creating subtitles is using the audio-visual content as a source. This can be used with or without the dialog and metadata from the screenplay but is more accurate when the screenplay data is available. The accuracy of speech recognition, especially in videos where there is music, loud background noise or other factors, is not perfect, but accuracy has improved in recent years.
The final challenge is the addition of timing information. Timing relates to when the actual text is spoken in a movie, as well as the timing of the frame changes. The time of the frame changes is important to understand; subtitles can be split across different frames, and if necessary can be adjusted and paraphrased to match the desired reading speed. Timing is adjusted by analyzing the audio-visual content, considering when people start and stop speaking as well as what scene changes and other country- and language-specific variables may be relevant.
Typically, creating original source-language subtitles for a 90-minute video as an SRT or TTML file with the correct timing cues can take between 36 and 48 hours. By combining the above approaches, the time needed can be reduced from days down to a few hours.
Talk to an expert
Please provide a little bit of information in the form below and we will be in touch soon. If you are interested in working for Andovar (as translator/ SLV or full-time employee), please visit our careers site.
Audiences across Europe and Asia can now access Agoda’s extensive travel services in their own languages.
Reaching out to music lovers in Thailand, Spotify translated its user interface into Thai, expanding into a new market.