Add Amazon Transcribe for Audio Input

When to Use This Step

Use this step after transcript input works. Audio adds more variables, so it should not be the first path tested.

Create a Transcription Job

For audio input, the workflow should:

1. Read the uploaded audio object from S3.
2. Start an Amazon Transcribe job.
3. Write the transcript output to transcripts/<jobId>/.
4. Continue to Bedrock analysis after the transcript is available.

Example AWS CLI command:

aws transcribe start-transcription-job `
  --transcription-job-name cognitive-coach-job-test `
  --language-code en-US `
  --media MediaFileUri=s3://<bucket-name>/uploads/job-test/sample.mp3 `
  --output-bucket-name <bucket-name> `
  --output-key transcripts/job-test/

Check job status:

aws transcribe get-transcription-job --transcription-job-name cognitive-coach-job-test

Language Choices

Initial testing can use:

• en-US for English audio.
• vi-VN for Vietnamese audio if supported in your selected region.

Validation

Confirm:

• The Transcribe job status becomes COMPLETED.
• A transcript JSON file appears in S3.
• The transcript text is readable enough for Bedrock analysis.

If audio quality is poor, use the transcript upload path for the final demo.