RocketWhisper turns speech into text on Windows, Mac and Linux. It combines local speech recognition with file transcription, correction rules and optional AI text processing. The application offers a free trial and perpetual Personal or Business licenses. Personal covers one computer, while Business covers computers throughout an organization.
Microphone dictation
You choose a microphone, recognition language and speech model, then start and stop recording. RocketWhisper processes speech locally and displays the result for copying or output to another application. Windows and Linux use Whisper; Mac offers WhisperKit and Apple SpeechAnalyzer, depending on the system.
Global shortcuts control recording without returning to the main window. Windows supports holding Right Alt while speaking and releasing it to transcribe. A double tap starts continuous recording. Mac offers similar hold and toggle controls through Right Option or Fn.
On Mac, shortcut recording can paste into the application that previously held focus. Recording through the app’s button requires manual copying instead. Linux’s Wayland environment can limit global shortcuts, clipboard access and application detection.
Files, subtitles and history
Batch transcription accepts multiple files through an Add command or drag and drop. Windows and Linux handle audio and video files, extracting video sound through FFmpeg. The Mac batch workflow processes audio files sequentially. You choose TXT for transcripts or SRT and VTT for subtitle output.
RocketWhisper keeps recognition history locally. Opening an entry retrieves earlier text, while keyword search locates matching results. You can copy that text for reuse. Windows also provides deletion and export controls for history entries.
Terms and correction rules
Windows custom terms give Whisper hints before recognition. You enter names or specialist vocabulary, and RocketWhisper includes those terms in the recognition prompt. Prompt capacity limits how many terms the model can consider.
Correction rules work afterward. Each rule pairs a search pattern with replacement text; regular expressions can match patterns beyond a literal word. The Mac word dictionary also replaces terms in recognition results.
Punctuation controls operate at different stages. A recognition prompt encourages Whisper to include punctuation, while postprocessing rules insert punctuation and line breaks afterward. Windows can also use voice activity detection to skip silence and noise before decoding speech.
Application-specific text modes
You can map applications to processing modes. RocketWhisper detects the target application and applies the mode you assign, falling back to global settings when no mapping exists. Simple mode outputs recognition text directly. Smart mode adds punctuation and line breaks.
AI modes can formalize wording, make it casual, summarize, translate or correct grammar. You choose a provider and model, supply cloud credentials or a local server address, and test the connection. LM Studio and Ollama provide local processing options. Mac can also use Apple’s on-device Foundation Models when available.
Local speech recognition does not require sending microphone audio to an AI service. Optional cloud processing sends recognition text directly to the provider. A local AI server keeps that processing on the computer. If Windows AI processing fails, RocketWhisper retains the original text.
AI Command and custom instructions
AI Command works on text you highlight. A dedicated shortcut starts voice capture for an instruction, such as shortening or translating that text. RocketWhisper sends the text and instruction to the AI provider. Windows displays the result in its text area; Mac can replace the text you highlighted.
Custom instructions begin with a prompt you define and a separate shortcut. You dictate new speech, then RocketWhisper applies that prompt to the recognition result. This workflow does not require highlighting existing text. Provider configuration determines whether the transformation uses local or cloud AI.
Voice editing, launchers and search
Voice commands can insert a new line or paragraph and perform deletion commands. An application launcher instead matches a keyword you register with an executable path. When a match occurs, RocketWhisper launches the application rather than inserting the keyword as dictation. Mac launcher entries can also open URLs.
Voice Search recognizes a search phrase, extracts its query and opens Google in the default browser. This function uses the browser’s network connection even though speech recognition runs locally.
Mac live translation
Live Translation captures system playback rather than microphone input. You choose source and target languages and start capture from the menu bar. Apple SpeechAnalyzer produces speech text, and Apple Translation turns it into captions in a floating panel.
The first use of a language pair downloads its translation model; later translation runs locally. Available pairs depend on Apple’s translation models. The panel can show both source text and translation. Microphone dictation can continue alongside this separate playback-caption workflow.
