Deepgram ASR (Nova-3) for Openclaw

A lightning-fast, cost-effective speech-to-text skill leveraging Deepgram Nova-3 for highly accurate multi-language transcription.

vahnxu
v0.3.0
Feb 25, 2026
0
561
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install deepgram-asr

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install deepgram-asr using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is Deepgram ASR (Nova-3)?

The Deepgram ASR skill is a high-performance audio transcription tool designed for the Openclaw Skills ecosystem. By utilizing the state-of-the-art Nova-3 model, it delivers sub-300ms latency and exceptional accuracy for over 50 languages at a competitive price point of $0.0043/min. This skill bridges the gap between raw audio data and actionable text, making it an essential component for developers building voice-enabled AI workflows.

Built with privacy and efficiency in mind, this Openclaw Skills integration communicates directly with Deepgram's API via HTTPS. It requires no heavy third-party libraries, relying only on standard system utilities like curl and python3, ensuring a lightweight footprint for any deployment environment.

Deepgram ASR (Nova-3) Use Cases

  • Transcribing voice notes sent via Discord or WhatsApp integrations.
  • Converting recorded meetings or interviews into searchable text files.
  • Generating AI-powered summaries of long-form audio content.
  • Automated language detection and speaker diarization for multi-party recordings.
  • Integrating cost-effective cloud transcription into automated DevOps or productivity pipelines.

How Deepgram ASR (Nova-3) Works

  1. The skill identifies and reads a local audio file provided via CLI or a messaging platform integration.
  2. It initiates a secure POST request to the Deepgram API endpoint with specified parameters like model and language.
  3. The audio is processed by the Nova-3 engine, which returns a detailed JSON response.
  4. A lightweight Python script parses the JSON to extract the transcript or metadata.
  5. The final output is saved to a local file or returned to the user in the requested format.

Deepgram ASR (Nova-3) Setup

To get started with this addition to your Openclaw Skills library, follow these steps:

  1. Obtain an API key by signing up at the Deepgram Console.
  2. Set your environment variable:
export DEEPGRAM_API_KEY="your_api_key"
  1. Run a test transcription:
./scripts/transcribe.sh /path/to/audio.m4a

Deepgram ASR (Nova-3) Data Schema & Taxonomy

The skill organizes transcription data based on user-defined flags and standard defaults:

Data Type Format Default Location
Standard Transcript .txt Same directory as input
Detailed Metadata .json Specified via --json flag
Formatted Text Markdown/Paragraphs Managed via --paragraphs

It supports various input formats including WAV, MP3, MP4, M4A, OGG, FLAC, and WebM.

Deepgram ASR (Nova-3) Advanced Features

  • Speaker Diarization: Use --speakers to identify and separate different voices in a single recording.
  • Smart Paragraphing: Automatically format transcripts into readable sections using --paragraphs.
  • AI Summarization: Generate concise summaries of audio content directly through the API using --summarize.
  • Multi-Language Support: Force a specific language with --language <code> or use --detect-language for automatic identification.
  • Custom Output Routing: Redirect results to specific file paths using the --out flag for better workflow integration.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*