HN 표시: TypeWhisper – 여러 엔진, 프로필을 갖춘 음성-텍스트 변환
hackernews
|
|
🔬 연구
#ai
#openai
#review
#typewhisper
#음성-텍스트 변환
#자동 번역
#프로덕트 리뷰
원문 출처: hackernews · Genesis Park에서 요약 및 분석
요약
마르코가 개발한 오픈소스 음성 인식 앱 'TypeWhisper'가 출시되었습니다. 이 프로그램은 macOS와 Windows를 지원하며, 사용자의 목소리가 기기 외부로 유출되지 않는 완전한 오프라인 환경을 특징으로 합니다. WhisperKit이나 Parakeet 등 다양한 엔진을 GPU 없이 CPU만으로 구동할 수 있으며, 앱이나 웹사이트별로 언어와 엔진을 다르게 설정하는 프로필 기능을 제공합니다. 또한 내장된 AI 프롬프트를 통해 번역이나 요약이 가능하고, 플러그인 시스템을 통해 사용자가 직접 기능을 확장할 수 있습니다.
본문
[TypeWhisper](https://www.typewhisper.com/en/)[Features](https://www.typewhisper.com/en/#features)[Use Cases](https://www.typewhisper.com/en/use-cases)[Add-ons](https://www.typewhisper.com/en/addons)[Docs](https://www.typewhisper.com/en/docs)[Benchmark](https://www.typewhisper.com/en/benchmark)[Changelog](https://www.typewhisper.com/en/changelog) [DE](https://www.typewhisper.com/de/)[](https://ko-fi.com/seofood)[](https://discord.gg/pUFR4a65SD)[](https://github.com/TypeWhisper) # TypeWhisper 1.0, now on macOS. Private speech-to-text for your Mac with system-wide dictation, file transcription, prompts, profiles, history, dictionary, and snippets. [Download for macOS](https://github.com/TypeWhisper/typewhisper-mac/releases)[Read the macOS docs](https://www.typewhisper.com/en/docs/mac) Windows Beta and iOS Alpha are available for early testing.  ## See it in action. Your browser does not support the video tag. Press a hotkey. Speak. Text appears. ## The macOS 1.0 core. The stable release focuses on dictation, transcription, prompts, profiles, history, dictionary, and snippets. ### Private by default. Run speech-to-text locally on your Mac with no telemetry, no subscriptions, and no mandatory cloud dependency.  ### System-wide dictation. Use a global hotkey to dictate into any app, with fast insertion and configurable behavior.  ### Prompts and automation. Process text with built-in prompt actions, then go deeper with the local API, CLI, and plugins as advanced surfaces.  ### Profiles, history, dictionary. Keep app-aware settings, searchable history, correction rules, and snippets in one place.  ### File transcription. Drop in audio or video files, batch transcribe them, and export subtitles with timestamps.  ## Why not just use built-in dictation? ### Per-app profiles Automatically switch language, engine, and behavior per app or website. ### Six speech engines Pick the right engine for speed, accuracy, or language support. ### Audio and video files Transcribe full files with drag and drop. Export subtitles as SRT or WebVTT. ### History and automation Searchable transcription history and a local HTTP API for workflows. ## Choose your engine Six speech engines - three built-in, more via add-ons. | Feature | WhisperKitVersatile | Parakeet TDT v3Fast | Apple SpeechAnalyzerZero Setup | | --- | --- | --- | --- | | Languages | 99+ | 25 European | ~40 | | Streaming | | | | | Translation | 20 languages | 20 languages | 20 languages | | Speed | Fast | Up to 5x faster | Fast | | Model Sizes | Tiny to Large v3 | 1.1B params | System-managed | | Model Download | Manual in-app | Manual in-app | Automatic by macOS | | Best For | Multilingual & translation | European languages | Quick setup | | Accuracy | Excellent | Excellent | Good | Apple-optimized Whisper models. Best for multilingual use and streaming preview. NVIDIA's latest TDT architecture. Extremely fast transcription for European languages with excellent accuracy. Apple's native speech recognition. No manual model downloads - models are managed by macOS. Requires macOS 26+. ### WhisperKit Versatile Apple-optimized Whisper models. Best for multilingual use and streaming preview. Languages99+ Streaming Translation20 languages SpeedFast Model SizesTiny to Large v3 Model DownloadManual in-app Best ForMultilingual & translation AccuracyExcellent ### Parakeet TDT v3 Fast NVIDIA's latest TDT architecture. Extremely fast transcription for European languages with excellent accuracy. Languages25 European Streaming Translation20 languages SpeedUp to 5x faster Model Sizes1.1B params Model DownloadManual in-app Best ForEuropean languages AccuracyExcellent ### Apple SpeechAnalyzer Zero Setup Apple's native speech recognition. No manual model downloads - models are managed by macOS. Requires macOS 26+. Languages~40 Streaming Translation20 languages SpeedFast Model SizesSystem-managed Model DownloadAutomatic by macOS Best ForQuick setup AccuracyGood Additional engines (Qwen3 ASR, Groq Whisper, OpenAI Whisper) are available as [add-ons](https://www.typewhisper.com/en/addons). ## Available now, with clear release stages. macOS is the supported 1.0 path. Windows and iOS remain preview releases. 1.0 Stable ### macOS Stable 1.0 release for daily use on your Mac. [Download for macOS](https://github.com/TypeWhisper/typewhisper-mac/releases) Beta ### Windows Public beta for Windows 10 and 11. Expect ongoing polish and rapid iteration. [Download Windows Beta](https://github.co
Genesis Park 편집팀이 AI를 활용하여 작성한 분석입니다. 원문은 출처 링크를 통해 확인할 수 있습니다.
공유