Deepgram SDK Skill

Transcribe audio using the Deepgram SDK by initializing the client with your API key in a Next.js API route, then calling the transcription method with your audio file, URL, or WebSocket stream. The SDK handles server-side key management and returns formatted text transcripts with optional speaker diarization and language detection.

Yes, real-time audio transcription via WebSocket is supported through the Deepgram SDK's streaming API. Use the KeepAlive WebSocket pattern for low-latency responses, enabling live transcription for voice commands, meeting notes, and caption generation without waiting for file upload completion.

Speaker diarization is built into the Deepgram SDK's transcription features. Enable diarization in your transcription request to automatically label and separate different speakers in the output, making meeting notes and multi-participant recordings clearer and more organized.

Deepgram's Nova-3 model includes multilingual support and smart formatting that automatically handles currency symbols, dates, and other formatted text. Configure language detection or specify a language code in your transcription request to get accurate, professionally formatted transcripts.

The Deepgram SDK implements error recovery and KeepAlive WebSocket patterns to handle connection stability. Wrap transcription calls in error handlers and implement reconnection logic in your Next.js API routes to ensure reliable real-time streaming without losing transcript data.

The Deepgram SDK supports transcription from pre-recorded files, remote URLs, and real-time audio streams. Pass a file buffer, audio URL, or WebSocket stream to the same transcription method, giving you flexibility for stored recordings, external sources, and live audio input.