Setting up the IAM Role.
→ The first step in the process is to create an IAM Role using “Create role” action that would allow Lambda function to access other AWS Services securely.
→ Since an AWS Service called Lambda would be using this role, the trusted entity type had to be an AWS Service and Use case being Lambda
→ Once done, I attached the required permissions AmazonPollyFullAccess(to let function interact with Polly service) and AWSLambdaBasicExecutionRole(required for basic Lambda functionality).
→ After reviewing the permissions and settings, I gave the role a meaningful name to easily identify it later and created it successfully.
Creating the Lambda function.
→ Once the IAM role was ready, I moved on to building the core component of the project: the AWS Lambda function.
→ Since this project involved writing the code from scratch, I chose the “Author from scratch” which gives us the same Hello World program boiler plate. I then filled the basic information like function name and runtime which is Python 3.9 and chose existing IAM role which we created in the previous step.
→ Finally I added this Python code with Polly integration
import boto3
import base64
import json
polly_client = boto3.client('polly');
def lambda_handler(event, context):
text = event['queryStringParameters']['text'];
response = polly_client.synthesize_speech(
Text = text,
OutputFormat = 'mp3',
# The VoiceId is a required parameter in synthesize_speech API call of AWS Polly.
# It tells which voice to use when converting text to speech.
# For Example, Raveena correspond's to female's voice with Indian English accent.
VoiceId = 'Raveena'
)
audio = response['AudioStream'].read()
# This converts binary audio data from AWS Polly to text based data in Base64 format.
# Base64 is a way to represent binary data using only text characters.
# It is needed because browser can't directly process raw binary in JSON responses.
encoded_audio = base64.b64encode(audio).decode('utf-8')
return {
'statusCode': 200,
'headers': {
'Content-type' : 'audio/mpeg',
'Access-Control-Allow-Origin' : '*'
},
'body': encoded_audio,
'isBase64Encoded': True
}
Testing the Code.
→ Before proceeding towards API Gateway, it was needed to test the Python code to make sure our Polly integration was working correctly. I clicked on the “Test” inside Lambda console and configured a new Test Event with a name and API Gateway AWS Proxy template.
→ In the event JSON, I passed a sample query string parameter with the text I wanted to convert.
{
"resource": "/speech",
"path": "/speech",
"httpMethod": "GET",
"queryStringParameters": {
"text": "Testing AWS Lambda and Polly!"
},
"headers": {},
"multiValueHeaders": {},
"requestContext": {},
"body": null,
"isBase64Encoded": false
}
→ When I ran the test, the function executed successfully and returned a long base64-encoded string.
→ To confirm that this encoded string was a valid audio, I copied the response and used Base64 Guru to decode and play the audio. The result? It played exactly as expected.
Creating the API Gateway Endpoint.
→ After confirming that the Lambda function worked as expected, I proceeded to expose it via API Gateway so that it could be triggered through a browser-based frontend.
→ I created a HTTP API and configured it with appropriate name and IP Address type.
→ I then added a new route with the GET method and defined the path as /speech. For the integration target, I selected the Lambda function I created earlier. This essentially allowed the frontend to send a GET request with text and receive audio in return.
→ Finally, I created a stage named dev, enabled Auto Deploy so I wouldn’t need to manually push changes, and noted the invoke URL, which I’d use in the frontend code.
