Amazon Transcribe

What it is

Amazon Transcribe is an automatic speech recognition (ASR) service that converts audio to text. Use it on its own or to add speech-to-text to any application. It transcribes media files stored in Amazon S3 (batch) or audio streamed in real time (streaming).

Choose Transcribe when a question asks to turn speech into text: call recordings, meetings, videos that need subtitles, voice notes, or live captions. To turn text into speech, use Amazon Polly. To analyze the resulting text, pass it to Amazon Comprehend.

Features

Batch and streaming: transcribe files in S3, or stream audio and get the transcript back in real time.
Speaker diarization: label who said what, for up to 30 speakers.
Multi-channel audio: transcribe each channel separately, such as the agent and the customer on a call.
Language identification: detect the dominant languages in the media.
Improve accuracy:
- Custom vocabularies for domain terms, brand names, and acronyms.
- Custom language models for a whole domain's language.
Filter content:
- Vocabulary filters delete, mask, or flag unwanted words.
- PII (Personally identifiable information)Information that can identify a person, alone or combined with other data, such as a name, email address, or passport number. Amazon Macie finds PII in Amazon S3.Glossary entry redaction masks personal data in transcripts. You choose to output a redacted transcript only, or both redacted and unredacted transcripts; with redacted-only output, the original media is the only full record. Streaming can also identify PII without redacting it.
- Toxic speech detection flags harmful speech.
Subtitles: create subtitle files for videos.
Call Analytics: built for call center audio. It provides insights per call and per participant, such as sentiment over the course of a call, so you can see whether the agent turned the conversation around.
Amazon Transcribe Medical: speech recognition for medical speech, such as physician dictation and telemedicine. Transcribe is HIPAA (Health Insurance Portability and Accountability Act)A US law that protects health information. AWS customers can process protected health information on HIPAA-eligible services after accepting a Business Associate Addendum (BAA) with AWS, which is available in AWS Artifact.Glossary entry eligible.

Security, resilience, performance, and cost

Security

IAM controls access; batch jobs read media from S3 and can write transcripts to your bucket.
Encrypt output with a KMS key; traffic uses TLS (Transport Layer Security)The protocol that encrypts traffic between a client and a server and lets the server prove its identity with a certificate. It replaced SSL, though the older name is still common; HTTPS is HTTP over TLS.Glossary entry.
PII redaction removes sensitive data from transcripts.
CloudTrail records API (Application programming interface)A defined way for one program to ask another for data or actions, such as a set of HTTPS requests and responses. Every AWS service is controlled through its API, which the console, the CLI, and the SDKs call for you.Glossary entry calls.

Resilience and performance

Transcribe is a managed, Regional service. Use batch jobs for recorded media and streaming for live audio. Handle LimitExceededException on streaming by retrying.

Cost

Pay per second of audio transcribed, with no minimum. Custom vocabularies, vocabulary filtering, speaker diarization, and language identification are included.
Automatic content redaction (PII) and custom language models cost extra, and Call Analytics and Transcribe Medical have their own prices.

Monitoring

EventBridge events when batch jobs change state to COMPLETED or FAILED: start the next step when a transcript is ready, or alert someone (for example through SNS) when a job fails, with the reason in the event. CloudWatch and CloudTrail cover metrics and auditing.

Transcribe or another option?

If the requirement is	Choose
Convert recorded or live speech to text	Amazon Transcribe
Insights from contact center calls	Transcribe Call Analytics
Transcribe medical dictation	Amazon Transcribe Medical
Convert text to speech	Amazon Polly
A chatbot that understands spoken requests	Amazon Lex
Translate the transcript	Amazon Translate

Reverse mindmap: when the question points here

Is it Amazon Transcribe?

Yes, when the question says
- Convert speech or recordings to text leads to Transcribe
- Generate subtitles for videos leads to Transcribe subtitles
- Live captions leads to Transcribe streaming
- Analyze call center calls and sentiment leads to Call Analytics
- Physician dictation leads to Transcribe Medical
Tune the output
- Who said what leads to Speaker diarization
- Agent and customer on separate channels leads to Multi-channel audio
- Brand names and jargon leads to Custom vocabulary
- Hide personal data in transcripts leads to PII redaction
- Keep an unredacted copy as well leads to Redacted and unredacted output
- Flag PII in live audio without masking leads to Streaming PII identification
- Alert when a transcription job fails leads to EventBridge rule on FAILED
- Remove offensive words leads to Vocabulary filter
No, when the question says
- Text to speech leads to Polly
- Chatbot conversation leads to Lex

Key takeaways

Transcribe converts speech to text, from files in S3 (batch) or real-time streams.
Diarization labels up to 30 speakers; multi-channel transcription separates channels; custom vocabularies improve accuracy.
PII redaction, vocabulary filters, and toxic speech detection protect the output.
Call Analytics serves contact centers; Transcribe Medical serves healthcare.
Chain Transcribe with Translate and Comprehend for multilingual analysis of calls.