AWS Speech to Text: The Ultimate 2025 Guide to Amazon Transcribe

1import boto3
2
3def start_transcription_job(job_name, s3_uri, output_bucket, language_code='en-US'):
4    transcribe = boto3.client('transcribe')
5    response = transcribe.start_transcription_job(
6        TranscriptionJobName=job_name,
7        Media={"MediaFileUri": s3_uri},
8        MediaFormat='mp3',
9        LanguageCode=language_code,
10        OutputBucketName=output_bucket
11    )
12    return response
13
14# Example usage
15start_transcription_job(
16    job_name="my-first-job",
17    s3_uri="s3://your-bucket/audio.mp3",
18    output_bucket="your-output-bucket"
19)
20

1import asyncio
2from amazon_transcribe.client import TranscribeStreamingClient
3from amazon_transcribe.handlers import TranscriptResultStreamHandler
4from amazon_transcribe.model import AudioEvent
5
6class MyEventHandler(TranscriptResultStreamHandler):
7    async def handle_transcript_event(self, transcript_event):
8        results = transcript_event.transcript.results
9        for result in results:
10            if result.is_partial:
11                print(f"Partial: {result.alternatives[0].transcript}")
12            else:
13                print(f"Final: {result.alternatives[0].transcript}")
14
15async def stream_audio():
16    client = TranscribeStreamingClient(region="us-east-1")
17    stream = await client.start_stream_transcription(
18        language_code="en-US",
19        media_sample_rate_hz=16000,
20        media_encoding="pcm"
21    )
22    handler = MyEventHandler(stream.output_stream)
23    # Simulate sending audio
24    with open("audio.raw", "rb") as f:
25        while chunk := f.read(1024):
26            await stream.input_stream.send_audio_event(AudioEvent(audio_chunk=chunk))
27    await stream.input_stream.end_stream()
28    await handler.handle_events()
29
30asyncio.run(stream_audio())
31