Meetings and dictation

Record conversations on your users’ own devices, with no bot joining the call. Opengeni uploads the audio as it is recorded, transcribes it in the background, tells the speakers apart, and keeps versioned transcripts your product reads through the API. Composer dictation uses the same recordings.

Record a meeting in React

import { useRecorder } from "@opengeni/react/recordings";

function RecordButton({ workspaceId }: { workspaceId: string }) {
  const recorder = useRecorder({
    workspaceId,
    title: "Site walk",
    externalRef: "project:42", // your own id, returned in events
  });
  return recorder.state === "recording" ? (
    <button onClick={() => recorder.stop()}>Stop</button>
  ) : (
    <button onClick={() => recorder.start()}>Record</button>
  );
}

Each track is saved on the device every few seconds before it is uploaded straight to storage, so a reload, a crash or a lost connection loses nothing; the next useRecorder finishes uploading what was left. Pass systemAudio: true to also record a tab’s or the computer’s audio as its own track, and liveTranscript: true for a read-along preview while recording. useRecordings, useRecording and useTranscript list recordings, follow processing, and read and edit a transcript: name speakers, move a segment to another speaker, correct text, re-run with another model, or switch the shown version. useRecordingPeople reads the people directory speakers are named from.

On phones and other hosts

In React Native, pass the phone’s capture to the same hooks:

import { useRecorder } from "@opengeni/react/recordings";
import { useNativeRecorderPlatform } from "@opengeni/react-native/recordings";

const platform = useNativeRecorderPlatform({ storeScope: accountId });
const recorder = useRecorder({ workspaceId, platform });

It records AAC into the app’s files and keeps recording in the background. Without React, @opengeni/sdk/recordings has the headless pieces: OpenGeniRecordingsClient for every route, RecordingUploader for local-first upload with retries, and createBrowserRecorder for capture.

Read the transcript

import { OpenGeniRecordingsClient } from "@opengeni/sdk/recordings";

const recordings = new OpenGeniRecordingsClient(og);
const { recordings: latest } = await recordings.listRecordings(workspaceId, {
  externalRef: "project:42",
});
const page = await recordings.getRecordingTranscript(workspaceId, latest[0].id);
for (const segment of page.segments) console.log(segment.speakerLabel, segment.text);

Pages carry nextAfter until the last one; each segment.speakerLabel points at one of page.speakers, which carries the name people gave that speaker. A recording is private to the person who recorded it unless they share it with the workspace.

React from your backend

Recording events go out through your workspace webhooks: recording.started, recording.finished, recording.transcript.ready, recording.transcript.failed, recording.transcript.current_changed, recording.speakers.updated and recording.deleted. They carry ids, states and your externalRef, never titles, audio or transcript text, so your backend reads the recording through the API with its own authority.

Dictation

The composer’s microphone records each dictation as a recording with purpose: "dictation": one microphone track, always private, transcribed with the workspace’s voice input, billed as voice input, never listed with meetings and deleted a day after it is heard. Its text goes into the draft and the dictation is deleted; if the tab closes or transcription fails first, the next composer in the workspace offers the text. ChatComposer does this for you. For your own composer, use useVoiceInput from @opengeni/react/recordings, or useNativeDictation from @opengeni/react-native/recordings on phones.

Moving from transcription recordings

The /v1/workspaces/:workspaceId/transcription-recordings routes and the SDK’s createTranscriptionRecording, uploadTranscriptionRecordingChunk, finalizeTranscriptionRecording, processNextTranscriptionRecordingSegment, getTranscriptionRecording, listTranscriptionRecordings and discardTranscriptionRecording are deprecated and answer with Deprecation and Sunset headers. Record a dictation instead:

createRecording with purpose: "dictation" and one microphone track, then upload and finalize it like any recording, or let useVoiceInput do it.
Poll listRecordingTranscriptVersions until the version is complete, then read its text with getRecordingTranscript.
deleteRecording once the text is in the draft.

One-shot transcribeAudio is unchanged.

Get started

Build

Concepts

Guides

Reference

Examples

Further reading

Meetings and dictation

Record a meeting in React

On phones and other hosts

Read the transcript

React from your backend

Dictation

Moving from transcription recordings

​Record a meeting in React

​On phones and other hosts

​Read the transcript

​React from your backend

​Dictation

​Moving from transcription recordings

Record a meeting in React

On phones and other hosts

Read the transcript

React from your backend

Dictation

Moving from transcription recordings