joiemoie/parakeet-runpod

0

stars

7

commits

Python

primary language

Jan 19, 2026

updated

README

Parakeet-TDT-1.1B RunPod Serverless

This is a RunPod serverless worker for the NVIDIA Parakeet TDT 1.1B ASR model.

Prerequisites

  • RunPod Network Volume: This template expects the model file to be present in a Network Volume.
    • Mount path: /workspace
    • Model file path: /workspace/models/parakeet-tdt-1.1b.nemo

You must download the .nemo file and place it in your network volume before starting the endpoint.

Usage

1. Build and Deploy

Build the Docker image:

docker build -t <your-repo>/parakeet-runpod:latest .
docker push <your-repo>/parakeet-runpod:latest

Deploy on RunPod Serverless using this image. Ensure you attach your Network Volume to /workspace.

2. Inference

Send a POST request to your endpoint.

Input Format:

You can provide either audio_url (publicly accessible URL) or audio_b64 (base64 encoded audio file).

{
  "input": {
    "audio_url": "https://example.com/sample.mp3"
  }
}

OR

{
  "input": {
    "audio_b64": "<base64_string>"
  }
}

Output Format:

{
  "text": "transcribed text content",
  "model_path": "/workspace/models/parakeet-tdt-1.1b.nemo",
  "model": "parakeet-tdt-1.1b"
}

Contributors

joiemoie

7 commits

joiemoie/parakeet-runpod

0

stars

7

commits

Python

primary language

Jan 19, 2026

updated

README

Parakeet-TDT-1.1B RunPod Serverless

This is a RunPod serverless worker for the NVIDIA Parakeet TDT 1.1B ASR model.

Prerequisites

  • RunPod Network Volume: This template expects the model file to be present in a Network Volume.
    • Mount path: /workspace
    • Model file path: /workspace/models/parakeet-tdt-1.1b.nemo

You must download the .nemo file and place it in your network volume before starting the endpoint.

Usage

1. Build and Deploy

Build the Docker image:

docker build -t <your-repo>/parakeet-runpod:latest .
docker push <your-repo>/parakeet-runpod:latest

Deploy on RunPod Serverless using this image. Ensure you attach your Network Volume to /workspace.

2. Inference

Send a POST request to your endpoint.

Input Format:

You can provide either audio_url (publicly accessible URL) or audio_b64 (base64 encoded audio file).

{
  "input": {
    "audio_url": "https://example.com/sample.mp3"
  }
}

OR

{
  "input": {
    "audio_b64": "<base64_string>"
  }
}

Output Format:

{
  "text": "transcribed text content",
  "model_path": "/workspace/models/parakeet-tdt-1.1b.nemo",
  "model": "parakeet-tdt-1.1b"
}

Contributors

joiemoie

7 commits

Languages

Python

97.1%

Dockerfile

2.9%