deepgram-stt

Name: deepgram-stt
Availability: InStock
Author: shakudo

Convert spoken language into searchable text via the Nova-3 model.

Updated Feb 14, 2026

Search Tags:#real-time #speech-to-text #deepgram #diarization #stt #audio-processing #endpointing

Ready to run

npx skills add https://github.com/Shakudo-io/opencode-skills --skill deepgram-stt

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: deepgram-stt
Source: https://github.com/Shakudo-io/opencode-skills/tree/main/deepgram-stt
Command: npx skills add https://github.com/Shakudo-io/opencode-skills --skill deepgram-stt

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Deepgram STT speeds up and scales the conversion of spoken language into accurate, searchable text for real-time and batch audio workflows.

Core Features & Use Cases

Streaming transcription with sub-300ms latency for live conversations, calls, and media analytics.
Pre-recorded transcription for batch processing, meeting notes, and archive tagging.
Diarization, smart formatting, and language support across 36 languages to improve readability and insights.

Quick Start

Transcribe an audio stream in real time using the Nova-3 model and return the text with basic formatting.

Frequently Asked Questions about deepgram-stt

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I convert real-time audio streams to text with sub-300ms latency?▼

Real-time speech-to-text conversion uses the Nova-3 model via WebSocket streaming to transcribe live audio streams with sub-300ms latency. This approach supports live conversations, calls, and media analytics by returning text dynamically with smart formatting.

Can I use speech-to-text diarization for batch processing pre-recorded audio?▼

Speech-to-text diarization is fully supported for batch processing pre-recorded audio. The Nova-3 model identifies and separates speakers in audio files, which improves readability and generates structured meeting notes or archive tags.

Does this speech-to-text workflow support multi-language processing?▼

Multi-language processing is supported across 36 languages for both real-time streaming and batch transcription. The Nova-3 model applies smart formatting and endpointing to deliver accurate, searchable text regardless of the spoken language.

What do I need to authenticate a WebSocket streaming transcription request?▼

Authenticating a WebSocket streaming transcription request requires a valid Deepgram API key. This key validates your access to the Nova-3 model for real-time audio processing and endpointing features.

What is the best way to add smart formatting and endpointing to live transcriptions?▼

The best way to apply smart formatting and endpointing to live transcriptions is using the Nova-3 model. It automatically punctuates and structures spoken language during real-time streaming, ensuring accurate and readable text output for live conversations.

When should I use batch transcription instead of real-time speech-to-text?▼

Batch transcription should be used instead of real-time speech-to-text when processing pre-recorded audio files for meeting notes or archive tagging. Real-time streaming is better suited for live conversations requiring sub-300ms latency.