Media and content / Sports video content
Formula 1: Finding archive footage through natural-language search combined with race records
- Company
- Formula 1
- Country
- Multiple countries
- Adoption stage
- In operation
- Source published
- In use from
- Date basis
- The date the source was published. It can differ from the date adoption started.
- How the source was checked
- Read the full source text
The work problem
Formula 1 production teams and broadcasters needed to find footage described as a driver's first win or an emotional moment, which simple tags could not readily capture. AWS describes a 12-petabyte archive and explains that some questions require historical results from a separate database before precise search conditions can be constructed.
Technology and data
The company introduced Media Archive Discovery in April 2026. Amazon Nova on Amazon Bedrock sits in front of existing OpenSearch search, converting natural-language requests into actual archive tags and search conditions. For factual race questions, the model returns a lookup structure rather than writing SQL; a validator converts it into read-only queries over approved tables and columns. The system uses a DynamoDB tag catalog and race records in RDS, falling back to plain-text search on failure. Producers inspect results on a timeline and choose footage to use.
Results
AWS says footage research that previously took hours can now take minutes. It reports that prompt caching alone reduced model cost by 88%, with compact JSON for the tag catalog trimming another 12%. These figures concern the described model-processing cost, not total archive operating expenditure or production labour costs.
Limits and open questions
The source provides no search-accuracy evaluation sample, measurement period or before-and-after total-system cost comparison. The two cost figures have not been combined into a total cost-reduction percentage.
Sources
- Finding the perfect moment: How Formula 1® uses AWS AI to search archives in secondsAWS for M&E Blog, Accessed
Compiled from public sources. These are not results from ATF Works customers.