Installing OpenAI Whisper on AWS EC2 using Docker:

Deploy OpenAI Whisper on AWS EC2 Using Docker: A Step-by-Step Guide

OpenAI Whisper is a powerful speech-to-text model capable of handling multiple languages and audio formats. Deploying Whisper on an AWS EC2 instance allows you to run it efficiently and make it accessible via an API. In this guide, I’ll show you how to launch an EC2 instance, install Docker, and run Whisper using a Docker container.

Step 1: Launch an EC2 GPU Instance

For optimal performance, choose a GPU-enabled instance. Here’s a recommended configuration:

Once the instance is launched, SSH into it:

ssh -i /path/to/key.pem ubuntu@<ec2-public-ip>

Step 2: Install Docker

Install Docker on your EC2 instance to run Whisper in a container:

sudo apt update
sudo apt install -y docker.io
sudo systemctl enable docker
sudo systemctl start docker
sudo usermod -aG docker $USER

Log out and back in for the group changes to take effect.

Step 3: Clone the Whisper Repository

Clone the Dockerized Whisper repository:

git clone https://github.com/taimoor511/whisper.git
cd whisper

Step 4: Build the Docker Image

Build the Whisper Docker image locally:

docker build -t whisper-dev .

This process may take some time, depending on the instance type and internet speed.

Step 5: Run the Docker Container

Run Whisper in detached mode and expose port 8000:

docker run -d --name whisper-dev -p 8000:8000 whisper-dev

The Whisper API will now be accessible at:

http://<ec2-public-ip>:8000

Step 6: Test the API

Single file:

curl -X POST "http://<ec2-public-ip>:8000/whisper/" -F "files=@<file-name.mp3>"

Multiple files:

curl -X POST "http://<ec2-public-ip>:8000/whisper/" \
  -F "files=@<file1.mp3>" \
  -F "files=@<file2.mp3>"

Step 7: Access API Documentation

You can explore all endpoints using Swagger UI:

http://<ec2-public-ip>:8000/docs

Step 8: Stop and Remove Container

When you’re done, stop and remove the container:

docker stop whisper-dev
docker rm whisper-dev

Conclusion

Deploying Whisper on AWS EC2 via Docker makes it easy to run high-performance speech-to-text models with minimal setup. You can now build apps, batch audio transcription pipelines, or integrate it with other services.