OpenVerb – offline voice dictation for macOS using one model instead of two (github.com)

OpenVerb has introduced a groundbreaking offline voice dictation tool for macOS that operates using a single model instead of the typical two, enhancing both privacy and performance. By leveraging Google's Gemma 4 model for transcription and text polishing, OpenVerb eliminates the need for cloud connectivity and reduces memory usage, running efficiently with approximately 3 GB of RAM. The installation process is straightforward, requiring only a few minutes on a clean machine, and the app operates entirely offline, making it ideal for on-the-go dictation without compromising user data. This development is significant for the AI/ML community as it challenges traditional dictation tools that rely on multiple models, which often involve data transmission to the cloud, raising privacy concerns and introducing latency. By utilizing a single model, OpenVerb not only streamlines the process but also ensures that all data remains on the device, offering a robust alternative to subscription-based services. The app supports Apple Silicon processors and is designed for seamless integration into macOS workflows, positioning it as a versatile option for users seeking efficient and secure voice dictation solutions.