Audio turns into extra helpful when purposes can transcribe it, analyze the ensuing textual content, or converse a response. Deepgram supplies speech-to-text, text-to-speech, textual content intelligence, and voice-agent APIs. It seems in G2’s Voice Recognition Software program class.

This information compares Deepgram integration strategies, examines related G2 opinions, and demonstrates a Deepgram-to-Notion transcript workflow utilizing Node.js.

These are implementation examples. The Voice Agent API, text-to-speech API, and textual content evaluation API require totally different requests and output dealing with; the tutorial under covers prerecorded speech-to-text solely.

The API overview paperwork Deepgram’s speech, text-to-speech, voice-agent, and account APIs. The callback information describes callback authentication and retry habits; deal with it as a separate design from the synchronous instance under.

One G2 reviewer says they bought an integration working with out contacting help and located the setup documentation self-explanatory. One other reviewer operating a telephone agent describes utilizing Deepgram’s streaming API to course of dwell telephone audio. These are concrete examples of developer entry and real-time use, not ensures of the identical outcomes for different audio.

The place do customers see alternatives for enchancment?

A reviewer working with an API pipeline asks for clearer steering on edge circumstances involving characteristic combos. The phone-agent reviewer above additionally says switching to Nova-3 required a change from the older key phrases possibility. Deepgram’s present prerecorded API reference confirms that Nova-3 makes use of keyterm as a substitute of key phrases. A separate reviewer studies friction managing staff API keys. These are particular person studies, so groups ought to take a look at their very own mannequin, governance, and audio circumstances.

Deepgram’s prerecorded audio information explains distant and native file requests. Notion’s Create a web page reference describes the guardian web page and content material necessities.

Step 2: Create API credentials

Create a Deepgram mission API key within the console. Create a Notion inside reference to insert-content entry to the guardian web page, then retailer each tokens solely on the server. The instance expects DEEPGRAM_API_KEY, NOTION_API_KEY, NOTION_PARENT_PAGE_ID, and AUDIO_URL as setting variables.

See Deepgram’s API key information and Notion’s inside connections information.

Step 3: Transcribe the recording and create the web page

Use a supported Node.js runtime with built-in fetch(). The next full script checks each HTTP responses, avoids printing secrets and techniques, and stops if Deepgram returns no transcript. It creates a Notion web page underneath the configured guardian; rerunning it creates one other web page.

// save as transcribe-to-notion.mjs
const required = ['DEEPGRAM_API_KEY', 'NOTION_API_KEY',
  'NOTION_PARENT_PAGE_ID', 'AUDIO_URL'];
for (const identify of required) {
  if (!course of.env[name]) throw new Error(`Lacking ${identify}`);
}

const audioUrl = new URL(course of.env.AUDIO_URL);
if (audioUrl.protocol !== 'https:') throw new Error('Use an HTTPS audio URL');

const dg = await fetch(
  'https://api.deepgram.com/v1/hear?mannequin=nova-3&smart_format=true',
  {
    technique: 'POST',
    headers: {
      Authorization: `Token ${course of.env.DEEPGRAM_API_KEY}`,
      'Content material-Sort': 'utility/json'
    },
    physique: JSON.stringify({ url: audioUrl.href })
  }
);
if (!dg.okay) throw new Error(`Deepgram HTTP ${dg.standing}`);
const knowledge = await dg.json();
const transcript = knowledge.outcomes?.channels?.[0]?.options?.[0]?.transcript;
if (typeof transcript !== 'string' || !transcript.trim()) {
  throw new Error('Deepgram returned no transcript');
}

// Notion rich-text objects have measurement limits; this instance is for brief audio.
const chunks = transcript.match(/[sS]{1,1800}/g) || [];
if (chunks.size > 90) throw new Error('Transcript too lengthy for this instance');
const title = `Interview transcript ${new Date().toISOString()}`;
const notion = await fetch('https://api.notion.com/v1/pages', {
  technique: 'POST',
  headers: {
    Authorization: `Bearer ${course of.env.NOTION_API_KEY}`,
    'Notion-Model': '2026-03-11',
    'Content material-Sort': 'utility/json'
  },
  physique: JSON.stringify({
    guardian: { sort: 'page_id', page_id: course of.env.NOTION_PARENT_PAGE_ID },
    properties: {
      title: { sort: 'title', title: [{ type: 'text', text: { content: title } }] }
    },
    kids: chunks.map(half => ({
      object: 'block', sort: 'paragraph',
      paragraph: { rich_text: [{ type: 'text', text: { content: part } }] }
    }))
  })
});
if (!notion.okay) throw new Error(`Notion HTTP ${notion.standing}`);
const web page = await notion.json();
console.log('Created Notion web page:', web page.url);

The Deepgram response path outcomes.channels[0].options[0].transcript is proven in its response information. The Notion request makes use of the documented web page creation endpoint and a web page guardian. Affirm the goal workspace schema and entry earlier than operating it.

Step 4: Take a look at the ensuing workflow

• Run the script as soon as with a brief recording and a take a look at Notion guardian web page.
• Evaluate the returned transcript with the recording, particularly names and technical phrases.
• Open the printed Notion web page URL and ensure the title and all paragraph chunks seem.
• Take a look at a lacking or inaccessible audio URL and a Notion token with out parent-page entry; each ought to cease with an HTTP error.
• Earlier than scheduling the job, add persistent request monitoring so retries can not silently create duplicate pages.

For longer audio, use a sturdy job design, extra Notion block batches, and applicable retry dealing with. In case you select Deepgram’s callback mode, authenticate incoming callbacks and account for its documented retries somewhat than treating a callback as a one-time occasion.

Sure. Deepgram paperwork streaming speech-to-text for purposes receiving audio because it arrives. Use its API overview to pick the right streaming endpoint and take a look at your audio format and latency necessities.

Q2. Can Deepgram transcribe a neighborhood audio file?

Sure. Its prerecorded information exhibits a binary add request and a JavaScript SDK transcribeFile instance. The tutorial above makes use of a URL to maintain the pattern targeted.

Q3. Does Deepgram retailer my transcript for later retrieval?

Deepgram’s prerecorded information says it doesn’t retailer transcripts for later retrieval. Save the API response or configure a callback vacation spot you management. The instance writes the returned textual content to Notion.

This fall. Can I take advantage of Nova-3 with key phrase boosting?

Nova-3 helps keyterm prompting. Deepgram’s API reference says the older key phrases possibility isn’t supported with Nova-3. Affirm the precise parameter on the endpoint and mannequin you utilize.

Q5. Can Deepgram ship transcription outcomes to my utility asynchronously?

Sure. Deepgram’s callback information describes supplying a callback URL and receiving a request ID whereas processing continues. Authenticate callbacks and plan for doable retries. That is separate from the synchronous Node.js instance.

Q6. Is Deepgram API integration free?

Deepgram presently gives a beginning credit score and usage-based plans; utilizing the API past any included credit score can incur costs. Mannequin alternative and non-obligatory options have an effect on price. Test Deepgram’s dwell pricing earlier than deployment.

Q7. Can Deepgram generate speech from textual content?

Sure. Deepgram’s text-to-speech API converts a single textual content request into audio, and its streaming API handles steady textual content. Select the transport and mannequin that match the vacation spot’s audio format and latency wants; these APIs are separate from the transcription tutorial.

Q8. Can Deepgram combine with IVR or telephone programs?

Sure. Deepgram’s Voice Agent API helps a conversational WebSocket movement with listening, reasoning, and spoken responses. Its Plivo phone-agent information exhibits how an utility can bridge name audio to that API. An IVR or telephone deployment nonetheless wants a suitable telephony connection, audio transport, and workflow configuration.

Q9. How do I combine the Deepgram API right into a Python app to transcribe audio information?

Set up deepgram-sdk, create a DeepgramClient together with your server-side API key, open the native file in binary mode, and name shopper.hear.v1.media.transcribe_file(request=audio_file.learn(), mannequin="nova-3"). Learn the transcript from response.outcomes.channels[0].options[0].transcript. That is the present Python SDK v5 type; older SDK examples use totally different technique paths. See Deepgram’s official Python SDK migration information and prerecorded audio information.

Q10. How do I combine Deepgram for real-time speech-to-text utilizing WebSockets?

Open Deepgram’s streaming speech-to-text WebSocket with an API key, ship audio frames in a supported format as they arrive, and deal with interim and closing transcript messages individually. The streaming API reference paperwork the connection and parameters; Deepgram’s dwell phone-transcription information exhibits a server forwarding audio into the WebSocket. This streaming path is separate from the prerecorded REST instance above.

Q11. How do I join Deepgram to Twilio to transcribe telephone calls in actual time?

Have Twilio ship name audio to your server with a one-way Media Stream. Ahead these audio frames to Deepgram’s streaming speech-to-text WebSocket and course of its interim or closing transcripts. Deepgram’s Twilio STT information paperwork the bridge and distinguishes transcription from a two-way Voice Agent movement, which makes use of totally different Twilio streaming habits. Confirm consent and audio format for the decision you course of.

Q12. Is Deepgram match for builders constructing voice-first merchandise?

It may be when a product wants documented speech-to-text, text-to-speech, or a conversational Voice Agent API. Select the mix that matches your audio transport and latency necessities, then prototype with your personal languages, noise circumstances, and price profile. Deepgram’s API overview paperwork the obtainable paths; the G2 experiences above are particular person studies, not a efficiency assure.

Q13. Does Deepgram supply speech to textual content and textual content to speech APIs?

Sure. Deepgram paperwork prerecorded speech-to-text, dwell speech-to-text, and text-to-speech APIs. Use speech-to-text to show audio into textual content, and text-to-speech to generate audio from textual content; they’ve totally different request and output codecs.

How are you going to get extra worth from Deepgram integrations?

Begin with one recording sort and a transparent vacation spot for the transcript. Confirm transcription high quality and entry controls earlier than increasing to dwell audio or automated selections.

Evaluating speech instruments? Learn G2’s voice recognition software program information for broader choices and consumer views.