RealtimeSTT Releases

Introduced a unified client class that automatically manages the server lifecycle and updated the CLI commands to simplify server and client management.

Oct 21, 2024

A major update focused on performance and stability, introducing early transcription, soundcard compatibility, optimized communication, and improved logging and chunk handling.

Oct 3, 2024

Added a method to clean the audio buffer and prepared functionality for automatic downsampling on soundcards that do not support 16kHz recording.

Sep 26, 2024

Aug 18, 2024

Introduced a new parameter to allow sharing the main model for both realtime and final transcriptions, offering a trade-off between memory usage and optimization.

Aug 17, 2024

Implemented language detection features for both full sentence and real-time transcription, including confidence probability metrics.

Aug 16, 2024

Added a new parameter to utilize the Silero model for end-of-speech detection, improving accuracy in noisy environments.

Aug 7, 2024

Fixed CUDA initialization errors on Linux by consolidating thread creation methods and updated dependencies.

Jul 19, 2024

Major update introducing OpenWakeWord support for wake word detection, including training guides and conversion instructions.

Jun 28, 2024

Updated multiprocessing settings and upgraded the faster_whisper library to the newest version.

Jun 2, 2024

Added decoding parameters and a method to dynamically switch between microphone input and audio injection.

Apr 14, 2024

Added beam size parameters to optimize for faster transcriptions and exposed raw bytes from the last transcription.

Apr 8, 2024

Fixed a queue size issue for macOS and upgraded the PyTorch requirements.

Mar 30, 2024

This release introduces a callback mechanism to allow client-side processing of audio chunks captured by the microphone. It enhances the recording pipeline by enabling users to hook into the audio stream directly.

Mar 16, 2024

This release migrates to PyTorch's multiprocessing backend and introduces configuration parameters for device selection and computing precision. It also improves usability by making the transcription process interruptible and fixes a specific issue.

Jan 29, 2024

Fixes startup issues and login placeholders, and updates translations.

Jan 21, 2024

Introduces support for multiple Dockge instances that can proxy requests between each other.

Jan 3, 2024

Improves the stack list UI by hiding folders without yaml files and updates the Docker Compose plugin.

Dec 16, 2023

The update brings support for the large-v3 Whisper model and a new browser-based client example. It also introduces methods for manual audio feeding and microphone usage.

Dec 15, 2023

This release focuses on stability and cross-platform compatibility, specifically addressing installation issues on macOS and improving interrupt handling. It resolves specific bugs related to multiprocessing queues and UI spinners.

Nov 9, 2023