AWS Transcribe

Convert speech in audio and video into text with Amazon Transcribe. Start and manage asynchronous S3-based transcription jobs and custom vocabularies from inside a flow.

8 actions available

New sales call recording lands in S3

Agent starts a transcription job with the product custom vocabulary

Agent reads the completed transcript when the job finishes

Agent scans the transcript for commitments and follow-up items

Agent files the transcript and action items against the deal record

Rep gets the summary and the follow-ups in Slack

A transcript mentioning a contract change or complaint routes to a manager

What This Integration Enables

Transcribe makes spoken conversations usable. It runs a transcription job against audio or video in S3, applies a custom vocabulary so your specific terms come back right, and returns text you can search, summarize, and route. Managing jobs and vocabularies from inside a flow means transcription happens the moment a recording lands, not when someone gets around to it. A transcript on its own is just more text to read. The point is what the flow does next: extract the follow-ups, file them against the right record, and surface the moments that need attention. An orchestration layer turns a transcript into action items and escalations, and FlowRunner is built for that layer.

Without FlowRunner

Calls not searchable Recorded calls sit as audio nobody can find later

Notes taken by hand Someone listens back and types out what was said

Terms transcribed wrong Product and domain names come back garbled

With FlowRunner

Audio made text Every recording becomes a searchable transcript

Action items extracted Follow-ups pulled from the transcript automatically

Vocabulary honored Custom terms and product names transcribe correctly

Use Case Scenarios

Sales Call Follow-Up

A sales call recording lands in S3. The agent starts a transcription job with the product custom vocabulary so model names come back correct, reads the transcript when the job finishes, and extracts the commitments and follow-up items. It files everything against the deal record and sends the rep a summary. The rep gets accurate follow-ups without listening back to the call.

Meeting Record Keeping

Recorded meetings need a durable, searchable record. The agent transcribes each recording, produces a summary with decisions and owners, and stores it where the team can find it. Meetings that used to vanish into a recording nobody replays become a searchable archive, and decisions are captured without a dedicated note-taker.

Voicemail Triage

Customer voicemails accumulate in a shared line. The agent transcribes each one and reads the intent. Routine messages are logged and routed by topic. A voicemail that mentions a cancellation, a complaint, or an urgent request is flagged and sent to a person immediately, so the messages that matter do not wait in a queue.

Human-in-Loop Highlight

A transcript can surface something that needs a real response: a complaint, a contract change, a commitment someone made on a call. When the agent detects that in the text, FlowRunner routes it through a [human-in-the-loop](/concepts/human-in-the-loop/) step rather than filing it silently. The agent pauses, assembles the relevant passage and the surrounding context, and sends it to the right manager via Slack. A person reads what was actually said and decides. Transcription handles the volume; a human handles what the words mean.

Agent processes routinely

Detects that in the text

Clear match Continues automatically

Ambiguous Routes to human via Slack

Human decides

Agent resumes with decision

Agent Capabilities

8 actions

Transcription Jobs

Start Transcription Job Starts an asynchronous transcription job for an audio or video file already stored in Amazon S3.
Get Transcription Job Retrieves the current status and details of a transcription job. When the job status is COMPLETED, the result includes the TranscriptFileUri where the transcript JSON is stored.
List Transcription Jobs Lists transcription jobs in your account, most recent first. Optionally filter by status or by a substring of the job name, and page through results using the returned cursor.
Delete Transcription Job Permanently deletes a transcription job and its metadata. This does not delete any transcript file already written to your own S3 output bucket.

Custom Vocabularies

Create Vocabulary Creates a custom vocabulary to improve transcription accuracy for domain-specific terms, names, and acronyms.
Get Vocabulary Retrieves information about a custom vocabulary, including its processing state (PENDING, READY, or FAILED) and a temporary download URI for the vocabulary contents when available.
List Vocabularies Lists the custom vocabularies in your account. Optionally filter by processing state or by a substring of the vocabulary name, and page through results using the returned cursor.
Delete Vocabulary Permanently deletes a custom vocabulary. Any in-progress transcription jobs already using it are unaffected, but it can no longer be applied to new jobs.

Start building with AWS Transcribe

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us