Analysed one million video records within two weeks with the Gemini API in Vertex AI
Scaled to process 500K video segments concurrently with the Gemini API in Vertex AI
Creates robust and consistent metadata to enable intuitive searches in seconds with the Gemini API in Vertex AI
The ABC modernised its 90-year-old archive using the Gemini API in Vertex AI for multimodal analysis to create rich, descriptive metadata.
The ABC modernised its 90-year-old archive using the Gemini API in Vertex AI for multimodal analysis to create rich, descriptive metadata.
Bringing 90 years of footage history to the present with AI-powered archiving
The Australian Broadcasting Corporation (ABC) holds priceless national treasure within its archives: decades of audio and video from newscasts, interviews, and documentaries that chronicle the history of the continent. Expertly captured, curated, described, and preserved over 90 years, the archive is a wonderful resource used each day by the ABC’s content making teams, to create and tell new stories. For this content to be easily discoverable by content makers, it needs high-quality metadata — a time-consuming challenge given the inconsistent archival formats used over the past 90 years. While published content has been carefully catalogued, the ABC Archive also holds thousands of hours of raw, unpublished audio-visual content.
Once writing a label and date on a tin of film was enough. Today, modern metadata standards require consistent data points for all records to ensure content is searchable, exchangeable between systems, and usable for generations. This metadata, captured daily, includes detailed descriptions of what is shown, names of talent, dates of capture, detailed transcripts, and more. Understanding the breadth of the content and surfacing the best of it is a priority. So, how did the ABC bridge the gap for older content, between yesterday’s “lean” metadata into today’s “rich” format?
The answer lies in an open-access digital platform called CoDA (Content Digital Archives), launched in 2018. Content makers, such as journalists, producers, and editors can use CoDA on millions of videos, audio, and photos on their desktops, without having to visit the vault to view source tapes on analog equipment.
This work strengthens one of the ABC’s greatest assets, our archive, by making its depth instantly accessible to our journalists. By generating rich, consistent metadata at scale, we’ve opened up decades of footage in a way that supports faster discovery, sharper storytelling, and better use of this valued national resource.
Damian Cronan
Chief Digital and Information Officer, ABC
More recently, the CoDA team launched an AI-enhanced archive project to make it even easier for content makers to find and retrieve footage. With the Gemini API in Vertex AI, and Gemini 2.0 Flash models at the core of this AI initiative, the project streamlines the creation of rich, AI-generated metadata for CoDA’s ever-expanding library of multimedia assets.
The project requires AI that can effectively handle long-form video streams, such as newscasts and documentaries. At the project’s peak, Gemini showcased scalability by processing half a million video segments concurrently without missing a beat.
Making content searches a breeze with rich, semantic metadata
Now, the CoDA team uses Gemini 2.0 Flash to enable multimodal analysis beyond human-generated metadata. The model automatically catalogues a segment of video and flags incidental footage, such as “dolphins swimming around Sydney Harbour” or “people lining up waiting at a COVID station.” Content makers can then leverage this as valuable, accessible stock footage.
High-quality metadata is critical to powering CoDA's search and helping content makers find the right footage to enrich their stories.
We are proud of the decades of care and expertise that has gone into capturing and preserving important content, ensuring the ABC maintains a rich Archive for future generations — and now with these new tools we are excited to see this history being made even more discoverable to ABC content makers.
Janelle Mikkelsen
ABC Archives Manager, ABC
Previously, a video transcript might only contain vague terms, such as “cricket game,” meaning it could have taken hours to sift through archival material just to find the necessary footage. Gemini addresses this by generating rich, structured metadata that powers natural language search and makes archival footage more accessible for all. This semantic approach goes beyond matching keywords by describing the actual meaning and visual context of the content, making it searchable even if the user asks for very specific things, such as “Find clips of a cricketer with zinc on their nose.”
Additionally, a new “assist” tooling uses natural language semantics to make search even more intuitive and powerful by surfacing similar relevant footage (say, of a cricketer "with zinc on their cheeks"), dramatically improving content retrieval and options.
In practice, this means that instead of taking an hour to find the right video and then scrubbing through hours of tape to find the best possible footage to tell a news story, a content maker can find the specific clip they need in seconds – freeing up content makers to focus on higher storytelling value.
Unlocking a new era of content accessibility with Gemini in Vertex AI
The editability of the AI-generated metadata will be a key advantage, with crucial human oversight, such as reviewing and refining output to ensure consistency and to meet the ABC's high quality standards. A human-in-the-loop correction process is vital, as the feedback fine-tunes the AI models over time, resulting in continually improving search accuracy.
Leveraging this established quality and consistency, the ABC has integrated AI-tagging directly into the content pipeline to proactively assist with creating descriptive metadata from the start. Now when a new video, such as an interview or raw footage, is uploaded to Archives, Gemini in Vertex AI automatically applies the metadata through daily workflows.
The technology is now driving a new era of rapid content access. Next steps will be to unlock even more content and to turn attention from video to audio with a similar process. Next time you see some historical footage illustrating a story on the ABC, it might be that AI was used to help surface it for re-use like this.
The ABC Archives is a national treasure, and it’s both a privilege and a responsibility to ensure this unique resource is readily available to ABC content makers and journalists. Using CoDA — now enhanced thanks to the Gemini API in Vertex AI — ensures more of the Archives are accessible — and to a remarkably granular level.
Rebecca Matthews
Head of Content Management and Distribution, ABC
The Australian Broadcasting Corporation is an Australian public service broadcaster that provides radio, television, online, and mobile services throughout metropolitan and regional Australia, and the Asia Pacific region.
Industry: Media and Entertainment
Location: Australia
Products: Gemini API in Vertex AI, Gemini 2.0 Flash
