[FEAT]: Automate Transcription Pipeline

[HoukasaurusRex]

Infrastructure

Sample Architecture

JavaScript SDK

Create CDK repo for infrastructure

DynamoDB

Store podcast episode lists and transcription meta info.

Columns

• id (guid from rss)
• title
• audio_url
• transcription_status
• transcription_url

Lambdas

• Download RSS information to DynamoDB from RSS
• Start jobs for AWS Transcribe
• Extract transcription (depends what happens when transcription finished. If added to s3, can initiate site build process instead)

AWS Transcribe

• Find out where transcriptions are stored
• Custom voice model training?

Pricing

Bulk of cost comes from AWS Transcribe, with cost potential in DynamoDB depending on storage, reads, and writes per month.

Estimate

[HoukasaurusRex]

Example API request

{
    "TranscriptionJobName": "raw-ep1-what-is-dd-test",
    "LanguageCode": "en-US",
    "MediaSampleRateHertz": 44100,
    "MediaFormat": "mp3",
    "Media": {
        "MediaFileUri": "s3://raw-transcription-test/raw-ep1-what-is-dd.mp3"
    }
}

Example API response

{
    "TranscriptionJob": {
        "TranscriptionJobName": "raw-ep1-what-is-dd-test",
        "TranscriptionJobStatus": "IN_PROGRESS",
        "LanguageCode": "en-US",
        "MediaSampleRateHertz": 44100,
        "MediaFormat": "mp3",
        "Media": {
            "MediaFileUri": "s3://raw-transcription-test/raw-ep1-what-is-dd.mp3"
        },
        "Transcript": {},
        "StartTime": "2021-07-10T10:57:51.298Z",
        "CreationTime": "2021-07-10T10:57:51.222Z",
        "Settings": {
            "ChannelIdentification": false,
            "ShowAlternatives": false
        }
    }
}

[HoukasaurusRex]

Convert aws transcribe json to srt package / code