Create recording summaries utilizing generated AI with Amazon Bedrock and Amazon Transcribe

Assembly notes are an necessary a part of collaboration, however they’re usually forgotten. Whereas main discussions, listening rigorously, and typing notes, necessary info usually goes unrecorded. Even when notes are captured, they might be unorganized or illegible, rendering them ineffective.

This put up exhibits you how one can use Amazon Transcribe and Amazon Bedrock to routinely generate clear, concise summaries of your video or audio recordings. Whether or not it is an inner crew assembly, a convention session, or a monetary outcomes name, this strategy helps distill hours of content material to the purpose.

Describes an answer for transcribing undertaking crew conferences and summarizing key factors utilizing Amazon Bedrock. We’ll additionally present you how one can customise this resolution for different widespread situations resembling course lectures, interviews, and gross sales calls. Learn this text to simplify and automate your note-taking course of.

Resolution overview

Save time, acquire insights, and improve collaboration by combining Amazon Transcribe and Amazon Bedrock. Amazon Transcribe is an computerized speech recognition (ASR) service that makes it straightforward so as to add speech-to-text performance to your functions. Precisely convert speech to textual content utilizing superior deep studying expertise. Amazon Bedrock is a completely managed service that gives high-performance foundational fashions (FM) from main AI firms resembling AI21 Labs, Anthropic, Cohere, Meta, Stability AI, and Amazon in a single API. The capabilities you want to construct generative AI functions. Amazon Bedrock makes it straightforward to experiment with a wide range of high FMs and privately customise them along with your information utilizing methods resembling fine-tuning and acquisition augmentation technology (RAG).

The answer introduced on this put up is orchestrated utilizing an AWS Step Capabilities state machine that’s triggered if you add a recording to a specified Amazon Easy Storage Service (Amazon S3) bucket. Step Capabilities helps you to create serverless workflows to orchestrate and join parts throughout AWS providers. It handles the underlying complexity so you possibly can focus in your software logic. This helps with job coordination, distributed processing, ETL (extract, remodel, load), and enterprise course of automation.

The next diagram exhibits the high-level resolution structure.

The answer workflow consists of the next steps:

1. Customers save recordings to S3 asset buckets.
2. This motion triggers the Step Capabilities transcription and summarization state machine.
3. As a part of the state machine, an AWS Lambda perform is triggered to transcribe the recording utilizing Amazon Transcribe and save the transcription to an asset bucket.
4. The second Lambda perform takes the transcription and generates a abstract utilizing Amazon Bedrock’s Anthropic Claude mannequin.
5. Lastly, the ultimate Lambda perform makes use of Amazon Easy Discover Service (Amazon SNS) to ship the recording abstract to the recipient.

This resolution is supported in areas the place Amazon Bedrock’s Anthropic Claude is out there.

A state machine coordinates the steps to carry out a selected job. The next diagram exhibits the detailed course of.

Conditions

Amazon Bedrock customers should request entry to a mannequin earlier than it may be used. That is his one-time motion. This resolution requires you to allow entry to the Anthropic Claude (not Anthropic Claude Immediate) mannequin in Amazon Bedrock. For extra info, see Mannequin Entry.

Deploying resolution sources

This resolution is deployed utilizing AWS CloudFormation templates. GitHub repository, routinely provisions the required sources into your AWS account. The template requires the next parameters:

• Electronic mail deal with used to ship abstract – The abstract might be despatched to this deal with. It’s essential to settle for the preliminary Amazon SNS affirmation e-mail earlier than receiving further notifications.
• Description abstract – These are the directions given to the Amazon Bedrock mannequin to generate the abstract.

Run the answer

After you deploy your resolution utilizing AWS CloudFormation, carry out the next steps.

1. After you create your CloudFormation stack, you will obtain an Amazon SNS affirmation e-mail a short while later.
2. Within the AWS CloudFormation console, navigate to the stack you simply created.
3. on the stack output Click on on the tab to seek out the worth related to. AssetBucketName; it seems to be like this summary-generator-assetbucket-xxxxxxxxxxxxx.
4. Within the Amazon S3 console, navigate to your asset bucket.

Add your recording right here. Legitimate file codecs are MP3, MP4, WAV, FLAC, AMR, OGG, and WebM.

5. Add your recording recordings folder.

Importing a recording routinely triggers the Step Capabilities state machine. This instance makes use of a pattern crew assembly recording. sample-recording GitHub repository listing.

6. Within the Step Capabilities console, go to the summary-generator state machine.
7. Choose the title of the state machine that can run within the state operating.

Right here you possibly can monitor the progress of the state machine because it processes the information.

8. After reaching success If that’s the case, it is best to obtain an e-mail abstract of your recording.

Alternatively, you possibly can go to your S3 asset bucket and look at the transcripts within the transcripts folder there.

Examine the overview

A report abstract is emailed to the deal with you specified when creating the CloudFormation stack. Should you do not obtain the e-mail after some time, ensure you consent to the Amazon SNS affirmation e-mail it is best to have obtained after creating your stack, after which attempt importing your recording once more. This can set off the abstract course of.

This resolution consists of recordings of mock crew conferences that you should use to check the answer. The abstract ought to appear to be the next instance: Nevertheless, because of the nature of the technology AI, the output might be barely completely different, however the content material needs to be shut.

The important thing factors of stand-up are:

• Joe has completed reviewing the present state of job EDU1 and created a brand new job to develop the longer term state. That new job is within the backlog and might be prioritized. He’s at the moment beginning EDU2, however is blocked from choosing sources.
• Rob created SLG1’s tagging technique based mostly on finest practices, however might have to coordinate with different groups who’ve created their very own methods to align with a unified strategy. A brand new job has been created to regulate your tagging technique.
• Rob has made progress in debugging SLG2, however may have further assist. This job might be moved to dash 2 to permit time to acquire further sources.

Subsequent steps:

• Joe will proceed to work on EDU2 so long as doable till useful resource choice is decided.
• New duties must be prioritized to coordinate tagging methods throughout groups
• SLG2 has moved to Dash 2
• Stand-up will transfer to Mondays beginning subsequent week

Lengthen your resolution

Now that we’ve got a working resolution, listed below are some doable concepts for customizing the answer on your particular use case.

• Be at liberty to change the method to fit your accessible supply content material and desired output.
  • In conditions the place a transcript is out there, create an alternate Step Capabilities workflow to herald an present text-based or PDF-based transcript.
  • As a substitute of utilizing Amazon SNS to inform recipients by e-mail, you should use Amazon SNS to ship the output to a different endpoint, resembling a crew collaboration web site or your crew’s chat channel.
• Strive modifying the high-level directions offered by Amazon Bedrock CloudFormation stack parameters to generate output particular to your use case (this can be a generated AI immediate).
  • When summarizing an organization’s monetary outcomes, the mannequin can deal with potential promising alternatives, areas of concern, and issues that needs to be repeatedly monitored.
  • Should you’re utilizing it to summarize course lectures, the mannequin can establish upcoming assignments, summarize necessary ideas, listing details, and filter small speak from recordings.
• Create completely different summaries for various audiences for a similar recording.
  • Engineer summaries deal with design selections, technical challenges, and upcoming deliverables.
  • The undertaking supervisor’s abstract focuses on timelines, prices, deliverables, and motion objects.
  • Challenge sponsors obtain temporary updates on undertaking standing and escalations.
  • For longer recordings, attempt producing summaries for various ranges of curiosity and time. For instance, create a single sentence, a single paragraph, a single web page, or an in depth abstract. Along with the immediate, max_tokens_to_sample Use parameters to accommodate completely different content material lengths.

cleansing

To scrub up your resolution, delete the CloudFormation stack you created earlier. Observe that deleting a stack doesn’t delete the asset bucket. Should you not want the recordings or transcripts, you possibly can delete this bucket individually. Amazon Transcribe routinely deletes transcription jobs after 90 days, however you possibly can manually delete them earlier.

conclusion

On this put up, we explored how one can use Amazon Transcribe and Amazon Bedrock to routinely generate clear, concise summaries of video or audio recordings. We advocate that you simply proceed to guage Amazon Bedrock, Amazon Transcribe, and different AWS AI providers resembling Amazon Textract, Amazon Translate, and Amazon Rekognition to see how they may also help you obtain what you are promoting targets.

Concerning the writer

rob burns is a principal marketing consultant for AWS Skilled Providers. He works with prospects to handle safety and compliance necessities at scale in his complicated, multi-account AWS environments by means of automation.

Jason Stael is a Senior Options Architect at AWS based mostly within the New England area. He works with prospects to align the capabilities of AWS to their greatest enterprise challenges. Outdoors of labor, he spends his time constructing issues and watching comedian ebook motion pictures along with his household.