Audio QC API

AI-powered audio processing: transcribe speech, remove noise, and trim audio automatically.

What You Need

Must have:

• Linux computer (Ubuntu/Debian recommended)
• Docker installed

Nice to have:

• NVIDIA GPU (makes it way faster)

Setup

Step 1: Install Docker

sudo apt update
sudo apt install docker.io docker-compose
sudo usermod -aG docker $USER

Log out and back in after this step

Step 2: GPU Support (Optional but Recommended)

sudo apt install nvidia-container-toolkit
sudo systemctl restart docker

Step 3: Run the API

git clone <your-repo-url>
cd audio-api
mkdir models workspace temp
docker-compose up --build

That's it! The API will be running at http://localhost:8000

What It Does

• Transcribe - Converts speech to text with timestamps
• Denoise - Removes background noise from audio
• Clean - Does all of the above + trims audio to just the speech

How to Use

Go to http://localhost:8000/docs in your browser for an easy web interface.

Or use these commands:

# Just transcribe
curl -X POST "http://localhost:8000/transcribe/" \
  -F "file=@your-audio.wav" \
  -F "expected_text=what you think it says"

# Clean everything (recommended)
curl -X POST "http://localhost:8000/clean/" \
  -F "file=@your-audio.wav" \
  -F "expected_text=what you think it says" \
  -o cleaned-audio.wav

Important Notes

• First time will be slow (downloads AI models ~2GB)
• Works with WAV, MP3, FLAC files
• GPU makes it 5-10x faster
• Without GPU it still works, just slower