parakeet-stt

Transcribe audio locally to text with NVIDIA Parakeet TDT 0.6B.

Updated Feb 6, 2026

Search Tags:#offline #transcription #CPU #multilingual #speech-to-text #OpenAI-compatible #ONNX

Ready to run

npx skills add https://github.com/dcarmitage/claws --skill parakeet-stt

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: parakeet-stt
Source: https://github.com/dcarmitage/claws/tree/main/skills/parakeet-stt
Command: npx skills add https://github.com/dcarmitage/claws --skill parakeet-stt

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Transcribes audio locally into text without cloud services, preserving privacy and giving you full data control.

Core Features & Use Cases

Offline, CPU-based transcription with NVIDIA Parakeet TDT 0.6B v3.
Multilingual support across 25 languages with automatic language detection.
OpenAI-compatible API for seamless integration into local apps and workflows.

Quick Start

Start the Parakeet TDT service using Docker compose (recommended) or run locally with Python. Then transcribe an audio file by posting to the API at $PARAKEET_URL/v1/audio/transcriptions with the file and desired response_format (e.g., text, json, verbose_json, srt).

Frequently Asked Questions about parakeet-stt

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

FAQPage Schema

How do I run speech-to-text transcription locally on CPU without cloud services?▼

Local speech-to-text transcription on CPU is handled by the NVIDIA Parakeet TDT 0.6B model, which processes audio entirely offline to preserve privacy and data control. It can be deployed via Docker Compose or a local Python FastAPI server.

Can I use an OpenAI-compatible API for offline audio transcription?▼

Yes, an OpenAI-compatible API is exposed for offline audio transcription, allowing seamless integration into local applications and workflows. You post audio files to the /v1/audio/transcriptions endpoint and select text, json, verbose_json, or srt response formats.

Does offline speech-to-text support automatic language detection for multiple languages?▼

Automatic language detection supports offline speech-to-text across 25 languages. The Parakeet TDT model identifies the spoken language from the audio input and transcribes it into text without requiring manual language selection.

What is the best way to integrate local transcription into existing applications?▼

The best way to integrate local transcription is through its OpenAI-compatible API, which drops into existing OpenAI client configurations. Start the service via Docker Compose or Python, then post audio files to get text or subtitle formats back.

Do I need a GPU to run multilingual speech-to-text transcription?▼

No, a GPU is not required because the speech-to-text transcription runs entirely on CPU using the ONNX-optimized Parakeet TDT 0.6B v3 model. This allows local deployments on standard hardware without GPU dependencies.