Speakey

Speakey is a real-time dictation application that processes speech directly on the user's computer, without relying on cloud services. By holding down the Caps Lock key, the user can speak and the text is instantly inserted at the cursor, working in any application and text field. The priority is privacy: all audio remains on the device, no accounts are required, and after an initial activation, the program operates without an internet connection.

★★★★★

4.0(1 ratings)

Download Speakey (Official links)

File size: 356 MB

The latest version of Speakey is: 1.3.0

Operating system: Windows

Languages: Spanish, English

Developer: Slipwave Group LLC

Price: $45.00 USD (15-day free trial)

Private, local processing. All speech recognition runs on the user's machine. No audio, transcribed text, or usage data is sent to external servers. There is no data collection, telemetry, or need for user accounts. After license validation and model downloads, it works offline.
Key-activated dictation. The main function is activated by holding down the Caps Lock key (reconfigurable). Releasing it inserts the text at the cursor's location. This eliminates the need to switch windows or use copy-paste commands, functioning in any program.
Two optimized transcription engines. The software automatically selects the most suitable engine for the hardware. On NVIDIA GPUs, it uses faster-whisper with CUDA acceleration to achieve latency under one second. On other GPUs, it employs Parakeet via DirectML. It is also functional with just a CPU.
Personal vocabulary learning. Speakey incorporates an adaptive system that learns user-specific terms, such as proper nouns or technical jargon. Words that are transcribed incorrectly once are automatically corrected in the future.
Correction management and replacement rules. The user can click on any transcribed word to see phonetic suggestions from their personal dictionary. It allows adding manual replacement rules, organizing them into categories, and importing word lists. Changes are applied without needing to restart the application.
Support for 25 languages. The application includes models for multiple languages out of the box, such as Spanish, English, French, German, Italian, Portuguese, Dutch, Polish, Russian, Ukrainian, Czech, Romanian, Swedish, Norwegian, Finnish, and Hungarian. Language switching is done from the settings without additional downloads.
Complete suppression of the Caps Lock key. When using Speakey, the original function of the Caps Lock key is disabled, preventing accidental typing in capitals while dictating. This solves a common problem in chat applications.
Configurable session history. The program allows managing the transcription history. The user can choose to keep it only during the session, preserve it between restarts, or disable it completely. Each history entry can be edited or deleted individually.
Continuous dictation mode. For longer dictation sessions, there is a toggle mode that activates listening continuously without needing to hold down the key, facilitating the drafting of long texts.
License validation without telemetry. The software's only network communication is limited to three functions: initial license validation, downloading transcription models, and optional update checks. No usage or performance data is sent.

The development of Speakey began with the goal of creating a dictation tool that prioritized privacy and speed by leveraging local hardware. It is built with Rust using the Tauri 2.0 framework for the interface and the application core. Its audio processing includes a digital signal processing (DSP) chain with features such as adaptive mono mixing, DC offset removal, an 80 Hz high-pass filter, a soft limiter, and a 16-tap sinc resampler. The application is offered as a one-time purchase with no subscription, including a 15-day trial period that does not require credit card registration.

Alternatives to Speakey:

Glimp — Free Download. Real-time AI assistant for job interviews

Speakey

Alternatives to Speakey:

Glimp

Willow Voice

VoiceOS

Typeless

BB Recorder

Vowen

VoiceInk

OpenWispr

Pipit

Whispering Tiger

StarWhisper

Murmure