whisper_echo
is a server service repository, tailored for voice-to-text transcription using the whisper-cpp. The service takes in audio data, processes it through the Whisper system, and returns the corresponding text transcription. Implemented in C++20, specifically designed for Nix environments.
These instructions will guide you on how to clone, build, and test this project in your local Nix environment.
- Nix package manager installed and configured on your system.
This project utilizes Nix, a potent package manager that simplifies dependency management. Run the following command to install all necessary dependencies:
nix develop
git clone https://github.com/faker2048/whisper_echo.git
cd whisper_echo
mkdir build && cd build
cmake ..
make
Upon successful build, the resulting binary can be found in the build
directory.
./bin/whisper_echo ./models/ggml-base.en.bin
Instead ./models/ggml-base.en.bin
use your own model_path
Here is a usable Python version of the client example. Run nix develop
to enter the Python environment, and then execute python ./whisper_echo/examples/py_client/client.py --file path/to/your_audio.wav
.
This project is licensed under the MIT License.