CONFIDENTIAL CLIENT
How one of Brazil's largest TV networks automated content transcription with GenAI
Transcribing and summarizing large volumes of news coverage with editorial neutrality
Advanced automation for transcribing news coverageLower operationalcosts and higher efficiencyPrecision and impartiality in summary generation
One of Brazil’s largest TV networks faced challenges transcribing and summarizing its news coverage, which hurt operational efficiency and the multiplatform distribution of content. Working with BlueMetrics, the broadcaster rolled out a solution built on generative AI, using AWS Transcribe and Bedrock to automate the conversion of audio into text and generate impartial summaries.
Overview
The client is one of Brazil’s largest TV networks, with a significant national presence through more than a hundred stations, both owned and affiliated. Known for its investigative journalism and crime-focused programming, the broadcaster stands out for its intensive coverage of criminal cases and investigative stories.
In a market where digital news is consumed at a fast pace, the client wanted a solution that would raise operational efficiency, guarantee impartial content, and improve the multiplatform distribution of its coverage.
Market context:
- The need to turn audiovisual content into text
- Growing demand for accessible, structured digital content
- Speed in making journalistic information available
- The evolution of digital journalism and the need for editorial standardization
The problem: how to make news coverage processing more efficient?
The manual process of transcribing and summarizing audiovisual content required considerable resources and a great deal of time, which hurt the speed of multiplatform distribution and the later reuse of that content in other journalistic productions.
“This is a very common pain in organizations that have to handle large volumes of unstructured data, in this case, content. But it also applies to a wide variety of companies, in segments as diverse as education, healthcare, and finance.”
Key challenges:
- Operational limitations:
- Business limitations:
- Technology limitations:
The solution: generative AI for automated transcription and summarization
BlueMetrics built an automated solution for transcribing and summarizing audiovisual content, using AWS cloud services to process crime coverage and generate neutral, impartial summaries while preserving journalistic integrity.
According to Diórgenes Eugênio, Head of GenAI at BlueMetrics, “Beyond the complexity of transcribing audiovisual content with high accuracy, we had to make sure the summaries carried no bias or opinion. That was the real challenge, because even small details could compromise the neutrality required. We used AWS Transcribe to convert the audio into text. After transcription, we moved to the second component: generating impartial summaries. Here the challenge was making sure the model did not infer subjective interpretations or draw conclusions about the facts. We used AWS Bedrock to bring in fact-based summarization techniques. The combination of Transcribe, Bedrock, and our custom validation layer was essential to deliver a pipeline aligned with the broadcaster’s requirements.”
Key components:
- An automated system for processing audio into text
- A summary generation engine with neutrality controls
- A scalable asynchronous processing pipeline
- A structured database for storage and lookup
- An automated flow for data input and output
Technology differentiators:
- Use of AI specialized in Portuguese audio transcription
- Use of generative AI to summarize the extracted text
- A serverless architecture with on-demand processing
- Native integration with existing infrastructure
Immediate benefits:
- A significant reduction in content processing time
- Standardized quality across generated summaries
- Guaranteed neutrality and impartiality
- Automatic scaling with demand
- Better use of operational resources
- Faster availability of content in text format
Results:
Rolling out the automated solution had a direct impact on the productivity of the client’s editorial team, significantly cutting the time needed to transcribe and summarize crime coverage.
As a result, the broadcaster began making its content available in multiple formats faster and more efficiently, meeting the growing demand of the digital market. On top of that, adopting AI ensured a higher level of standardization and impartiality, both essential to the credibility of investigative journalism.
By streamlining the workflow, the solution let journalists focus on strategic work instead of repetitive tasks, raising their capacity to produce new content. Structuring and organizing the extracted data also made it easier to reuse journalistic material, adding value to the broadcaster’s archive.
Operational efficiency:
- Full automation of the transcription process
- A significant reduction in content processing time
- Freeing teams up for strategic work
- Standardized documentation of news coverage
Editorial quality:
- Guaranteed neutrality and impartiality in summaries
- Consistency in the language and format of generated text
- Greater precision in documenting criminal cases
- Adherence to the broadcaster’s journalistic standards
Content management:
- Structured documentation of the journalistic archive
- Easy retrieval of historical information
- Better organization of produced material
- Speed in multiplatform distribution
Technologies used
The solution was designed using a range of AWS technologies, including:
AWS services
- Transcribe
- Bedrock
- S3
- Lambda
- DynamoDB
- Cognito
- Amplify
Languages, libraries, and frameworks
- Python
- Javascript
- Node
- React
Conclusion:
Rolling out this solution significantly transformed the client’s ability to process and publish its journalistic content, setting new standards of efficiency and quality in the Brazilian television market.
The project not only streamlined operational processes but also opened new possibilities to monetize and make the most of the broadcaster’s rich journalistic archive.
This project shows once again our company’s ability to build solutions that deliver real business impact.* “This is a solution with enormous potential for companies that generate large volumes of information, such as newspapers, radio stations, media networks in general, and content or streaming platforms, to name just a few examples.”*