on-device-transcription

• Purpose: A minimal application that converts speech to text entirely on-device using WebGPU technology, ensuring privacy by processing audio locally without cloud services.
• Key Features:
  • Local processing via [Ratchet](https://github.com/huggingface/ratchet fa-solid fa-up-right-from-square) (Hugging Face's on-device ASR framework).
  • Supports deployment as a web app or desktop application (Electron).
  • Requires WebGPU API support in browsers for web usage.
• Target Audience: Developers or users needing privacy-focused, offline-capable speech-to-text functionality.
• Framework Support: Built with SvelteKit for the frontend and Electron for desktop packaging.
• Demo: Web version available on [Hugging Face Spaces](https://huggingface.co/spaces/HugoDzz/ratchet-demo fa-solid fa-up-right-from-square).
• Underlying Tech: Leverages [OpenAI Whisper](https://openai.com/index/whisper/ fa-solid fa-up-right-from-square) model architecture via Ratchet.