Keyless is an open-source, locally run AI dictation and speech-to-text tool, available as a Desktop App (Beta) and a TUI terminal version. Its goal is to deliver a fast, private voice input workflow with as few dependencies as possible, built on the Rust stack. Users can download desktop installers or terminal binaries from Releases, or build it via Cargo or from source.
Based on the main description, Keyless’s AI capabilities are built on Whisper models, with a model download required on first launch. It supports real-time transcription, live preview while speaking, push-to-talk controls, final transcript generation, automatic language detection, voice activity detection, temperature fallback decoding, and quality metric tracking. For multilingual support, it claims support for 99 languages, but does not separately specify Chinese recognition quality or a Chinese-language interface.
Keyless’s biggest highlight is its privacy-first design: 100% local processing, no cloud, no API key, no account required, and all audio and transcription data stay on the device. It is also open source and auditable. On the performance side, it supports GPU acceleration, with macOS Metal enabled automatically and optional CUDA support on Linux/Windows. It also supports quantized models, automatic microphone sample-rate handling, and high-quality resampling. For integration, the main text does not mention an API, but it offers both desktop and terminal usage modes.
The main text does not mention commercial pricing, subscriptions, paid plans, or free-tier limits, so it can be understood as an open-source download/self-hosted tool. In terms of usability, the desktop version is more suitable for general users, though it is still in Beta. First-time setup requires granting microphone and accessibility permissions, downloading a Whisper model, and configuring hotkeys and output mode. The terminal version is better suited to developers and command-line users, but may present a barrier for non-technical users.
Its strengths include strong local privacy, no need for accounts or cloud services, cross-platform support, GPU acceleration, multilingual support, and open-source auditability. Limitations include transcription speed and quality depending on local hardware and the selected model, the desktop version’s maturity still needing to be proven, and no stated commercial support, team management, SLA, or Chinese-specific optimization. It is best suited for privacy-conscious individuals, developers, writers, terminal users, and office scenarios where users do not want to upload voice data to the cloud.
The main text does not provide information about official service availability, network connectivity, or payment. Since the tool itself runs locally, once the installer and model are obtained, usage does not depend on the cloud. However, accessing GitHub Releases or downloading models from China may be affected by the network environment. Alternatives include whisper.cpp, MacWhisper, Buzz, OpenAI Whisper API, or built-in system dictation.
⚠ This review is compiled from public sources and does not constitute a purchase recommendation. Verify all facts on the vendor's official site. Verify on keyless.sh official site.
keyless.sh is an Unknown AI Apps provider. TG4G tracks its product information, an overall rating of 7.0/10, and a China-accessibility score of China direct-connect friendly. Click "Visit Official Site" to reach keyless.sh directly.