Hello Diary uses Sherpa ONNX for speech recognition—a powerful open-source framework that runs entirely on your device. This means your voice never travels to cloud servers and never becomes training data.
What Is Sherpa ONNX?
Sherpa ONNX is an open-source speech recognition framework designed specifically for on-device processing. Unlike cloud-based systems, it runs models directly on your phone or computer without requiring internet connectivity.
The Privacy Architecture
When you speak into Hello Diary, your voice is captured and processed locally. The audio waveform is converted to text on your device using pre-trained models. These models are static—they don't update based on your voice or send data back to us.
The Processing Flow
- Audio Capture: Your device records your voice.
- Local Processing: Audio is fed to Sherpa ONNX models locally.
- Real-Time Transcription: Neural networks convert speech to text instantly.
- Zero Network Usage: No data is sent to the cloud.
Why Open Source Matters for Privacy
Sherpa ONNX is open source, meaning its code is public. This allows security researchers to audit the implementation and verify that there are no backdoors or hidden data collection. Transparency builds trust.
Comparing to Cloud Alternatives
Cloud Services (Google, Amazon): require uploading audio to their servers. They often retain rights to use your data for service improvement.
Sherpa ONNX: processes everything locally. You own your data completely.