Show HN: Whispering – Open-source, local-first dictation you can trust (github.com)

Article summary

Whispering is an open-source, local-first dictation application that can run as a browser-hosted app or a desktop surface within Epicenter. It records speech, transcribes it using a chosen provider, and delivers the text, with options for on-device transcription and direct provider connections. The application is designed to prioritize user privacy and security. Whispering is built using Svelte and Tauri, and its code is available on GitHub.

Main themes

• Open-source dictation
• Local-first approach
• Privacy and security
• Speech-to-text technology
• Cross-platform compatibility

What commenters say

• The addition of Deepgram listen API compatibility would enable live transcription via Deepgram or OWhisper, enhancing the application's functionality.
• Some users have expressed interest in using the Parakeet model locally, citing its superior performance and accuracy compared to Whisper.
• There are concerns about the security of the application, with some antivirus software flagging the installer as potentially infected or containing a trojan.
• The use of on-device transcription and local models raises questions about the trade-offs between privacy, security, and functionality, with some users prioritizing one over the others.
• The application's support for multiple transcription providers and models is seen as a key advantage, allowing users to choose the best option for their needs.
• Some users have reported issues with the application's compatibility with certain languages or platforms, highlighting the need for further development and testing.