ayushdh96/whisper-cpp repository overview

📦 whisper.cpp Dockerized ASR

This project provides a fully Dockerized setup of whisper.cpp, enabling users to transcribe audio files locally using OpenAI’s Whisper model with GGML optimization — no manual setup required. It includes support for both standard (ggml-base.en.bin) and quantized (ggml-base.en-q5_0.bin) models. 🧱 Repository Structure whisper.cpp/ ├── Dockerfile # Dockerfile for base.en model (downloads at build time) ├── Dockerfile.q5 # Dockerfile for q5 model (uses pre-copied local model) ├── models/ # Contains locally available models (e.g., ggml-base.en-q5_0.bin) ├── Catching.mp3 # (Optional) Sample audio file for testing └── README.md # Instructions and usage 🐳 How to Pull Prebuilt Images from Docker Hub For standard model: docker pull ayushdh96/whisper-cpp:base.en For quantized model: docker pull ayushdh96/whisper-cpp:q5 🎧 How to Transcribe Your Own Audio

Make sure your .mp3/.wav/.flac file is in the same directory as where you run the command.

🔹 Using Standard Model (base.en): docker run --rm -v "$PWD:/audio" ayushdh96/whisper-cpp:base.en
-m /app/models/ggml-base.en.bin
-f /audio/Catching.mp3
--output-txt --output-json --output-srt 🔹 Using Quantized Model (q5): docker run --rm -v "$PWD:/audio" ayushdh96/whisper-cpp:q5
-f /audio/Catching.mp3
--output-txt --output-json --output-srt ✅ No need to pass -m for q5 — it’s built into the container. 📄 Output Files

After running either image, the following files will appear in your directory: • Catching.txt — plain transcript • Catching.srt — subtitle file • Catching.json — structured transcript data

📌 Notes • base.en image downloads ggml-base.en.bin during build • q5 image comes preloaded with ggml-base.en-q5_0.bin • Supported formats: .mp3, .wav, .flac, .ogg • You can substitute the model path in base.en image with other variants (ggml-medium.en.bin, etc.)