How WhisperARC uses Whisper on Windows
See how WhisperARC runs Whisper through whisper.cpp, how Cloud works and which model fits your Windows PC.
Last updated
What Whisper means in WhisperARC
Whisper is OpenAI's speech-recognition model. WhisperARC uses it on Windows for dictation, system audio and files without Python or a terminal.
WhisperARC is an independent app and is not endorsed by OpenAI. Whisper is the name of the speech technology. It is not an official OpenAI app.
Local Whisper through whisper.cpp
Local Mode runs downloadable Whisper weights through whisper.cpp. GGML models stay on the PC and can use the CPU or a compatible Vulkan GPU. WhisperARC falls back to the CPU when needed.
Microphone audio, system audio and imported media stay on the PC. Local Mode is available on the Free plan.
Online Whisper in Cloud
Cloud sends audio online to Whisper Large v3 Turbo. The result returns to WhisperARC without saving it to cloud history.
Cloud translation is a separate step. Whisper handles recognition only.
Choose a Whisper model
WhisperARC offers Base q8, Small q8, Large v3 Turbo q5/q8, Large v3 q5 and full Large v3. Smaller or more quantized models use fewer resources; larger models need stronger hardware.
Turbo q5 is the default balance of quality, speed and size. Results vary by language, audio and hardware, so test a representative recording.
What WhisperARC adds around Whisper
- One global dictation shortcut with hold-to-talk or start/stop behavior from other Windows applications.
- Text delivery to the focused editable field with a copy panel when direct insertion is unavailable.
- System-audio capture, persistent file queues and local History.
- Project glossaries for names, acronyms, products and specialist vocabulary.
- A separate translation step plus TXT, SRT, VTT, JSON, DOCX and audio exports from History.
Primary sources
These sources cover Whisper and whisper.cpp. WhisperARC's hardware guidance leaves room for Windows use and is not a benchmark guarantee.