Live Captioner by Osaeris

This is an attempt to create a live captioner which can be fed audio by any chosen source and provide 2-line subtitles on a chroma-green background. Created on M series Mac. Uses metal for acceleration.

It's early days, but it's working! 🙂

---

Requirements

The Docker instance expects a Whisper server to be running.
I tried to run this inside the Docker container, but performance was too slow on Apple silicon, so Whisper runs natively on the host machine (Apple M2).

---

Install Whisper.cpp

brew install cmake
git clone https://github.com/ggerganov/whisper.cpp
cd whisper.cpp
make -j

Add a model (example: large-v3)

cd models
./download-ggml-model.sh large-v3

Run the Whisper server

./build/bin/whisper-server -m models/ggml-large-v3.bin -l en --port 9000

---

Run the Docker Captioner Service

Navigate to your project directory:

cd ~/Documents/scripts/docker-projects/speech-captioner/

How to stop and rebuild the container

docker stop captioner
docker rm captioner
docker build -t captioner .
docker run -p 8000:8000 --name captioner captioner

---

Use the Captioner

Visit:

http://127.0.0.1:8000/

Click "Start captions" (allow microphone permissions from your chosen source).