Summarize Video Content Using Bedrock Claude Instant and Transcribe

Architecture

Through the SST framework, we built a Serverless application to achieve overall functionality. A lambda is triggered through the S3 file upload success event to execute the Amazon Transcribe Job. This job will convert the voice content in the video into text content and store it in another S3 bucket. Through the creation event of the video content text in S3, it will automatically trigger a lambda function to read the file content and forward it to Bedrock's Claude model for text summary tasks. The final summary text obtained will be stored in a new S3 bucket.

Deploy

Before running, you need to install node.js. If you have not installed it yet, please download the installation package from node.js official website

git clone https://github.com/yourlin/summarize-video-with-bedrock.git
npm install
sst build
sst deploy

Usage

introduce

Open S3 in the AWS console, search for the summarize-video keyword, and you can find 3 related buckets. They are:

• -summarize-video-summarizeinputbucket
• -summarize-video-summarizeoutputbucket
• -summarize-video-transcribeoutputbucket

Or you can find these three buckets in the stack corresponding to Cloudformation and in the output tab page.

• summarizeinputbucket: used to upload video files
• transcribeoutputbucket: used for Transcribe output results
• summarizeoutputbucket: used by Bedrock to summarize the text and output the results

Test

Now you can upload a video/audio file to the summarizeinputbucket bucket. After waiting for tens of seconds, we can see that the summarized text has been generated in the S3 bucket of summarizeoutputbucket.

Reference

• SST Document
• Amazon Bedrock API runtime InvokeModel