Loading

Video Personalization and Localization pilot

,

Video Personalization and Localization pilot

Google
2023

Video Personalization and Localization pilot

Client: Google

Project date: 2023
Industry: Technology
Services provided: Content strategy, project management, content production, Education

Following the public unveiling of Universal Translator tool at Google I/O 2023, the Google Ads Scaled Content team team collaborated with localization team to explore innovative applications of AI in video dubbing.

This project leverages Google’s most powerful AI models, developed in partnership with Google DeepMind, to enhance the efficiency and quality of language dubbing processes.

Project objectives

  • Investigate and Implement AI in Video Dubbing: Explore and leverage AI capabilities, specifically those offered by Universal Translator, to improve various aspects of the video dubbing workflow.

  • Collaborate with localization team: Foster a strong partnership with the L10N team to combine expertise and insights, ensuring seamless integration of AI solutions into existing processes.

  • Enhance Efficiency and Quality: Identify opportunities to streamline video dubbing tasks, reduce manual effort, and improve the overall quality of dubbed content using AI technologies.

    • Stay at the Forefront of AI Innovation: Contribute to the advancement of AI-powered language solutions within Google by actively participating in the development and application of cutting-edge GenAI tools.

    This project represents a significant step towards harnessing the power of AI to revolutionize the video dubbing landscape, ultimately leading to improved accessibility and reach for multilingual content.

Challenge

The Google Ads Content team is committed to providing the right content to the right audience at the right time. However, current support videos primarily offer English voiceovers, limiting personalization and engagement for non-English speaking users. This presents an opportunity to leverage GenAI dubbing solutions to expand our content reach and impact.

This situation aligns with a broader cross-functional organizational needs:

Content team: Requires more personalized video content in multiple languages to increase coverage, engagement, and user satisfaction.

Location team: Seeks further investment in Hermit technology to achieve faster, more cost-effective, and scalable localization.

Google DeepMind: Benefits from additional partners and diverse video content to enhance and refine its AI models.

By addressing this production gap, we planned create a win-win scenario for all stakeholders, ultimately providing a more inclusive and effective support experience for our global audience.

Task

I conducted a pilot project to localize a set of high-traffic support videos into 10 target languages using Google’s AI dubbing technology.

 

Specifically, this project involved:

  • Selecting content: Identify 10 articles with videos based on high viewership to ensure a representative sample size for the pilot.
  • Localizing video content: Utilize AI dubbing technology to create localized versions of the selected videos in the 10 target languages.
  • Implementing changes: Update the chosen articles to include the newly dubbed videos, while leaving all other content within the articles unchanged.
  • Evaluating Results: Assess the impact of localized videos on user engagement, satisfaction, and other relevant metrics to inform future localization efforts.

Action

Demonstrated a proactive and collaborative approach to exploring the potential of AI dubbing, setting the stage for future expansion and optimization of multilingual video content.

  • Collaborating with localization team: Identifying high-impact videos and CUJs that were suitable for the pilot.
  • Identifying  languages: Chose priority languages to dub, ensuring broad reach and impact. Implemented
  • AI dubbing: Utilized proprietary dubbing technology to create dubbed versions of the selected videos.
  • Established comparative analysis: Set up a measurement framework to compare engagement between the newly dubbed videos and the previously subtitled versions.
  • Explored operationalization: Investigated and documented processes for streamlining and scaling AI dubbing workflows, ensuring quality standards are maintained throughout.
  • Shared Insights: Provided the partner teams with actionable information and recommendations based on the pilot findings, empowering them to operationalize dubbed content at scale.

Results

Significant increase in video engagement: The pilot demonstrated a 14% increase in average watch time across all dubbed languages compared to the previously subtitled-only versions, highlighting the positive impact of localized audio on user engagement.

Specific language success: Portuguese (Brazil) led the way in terms of video views and watch time, followed by Spanish (LATAM) and Japanese, indicating high demand for localized content in these regions.

Cost-effectiveness: The Universal Dubbing solution proved to be remarkably cost-efficient, offering approximately 4x cost savings compared to traditional human dubbing for languages like Spanish (LATAM), demonstrating the potential for scalable and budget-friendly localization.

These results validate the effectiveness of AI dubbing in enhancing user engagement, expanding content reach, and optimizing localization costs, supporting the broader adoption of this technology for future projects.

Back To Top