How much faster is Parakeet than Whisper?

In internal measurements, Parakeet reaches a real-time factor of about 0.10–0.13× on CPU, while Whisper measures 0.43–1.16× depending on model size and backend, meaning Parakeet is several times faster at transcribing the same audio, even without a GPU.

Which languages does each engine support?

Whisper supports 99 languages, including mixed-language input and code-switching mid-sentence. Parakeet covers around 25 European languages. If your language isn't on Parakeet's list, Whisper is the one to use.

Can I switch engines without losing my settings?

Yes. Both engines live under the same "On Device" provider in Settings → Speech-to-Text, switching is a single toggle, and your hotkeys, snippets, and history are untouched.

Whisper or Parakeet? Two local engines, one clear trade-off.

WhisPaste ships two fully offline speech engines side by side, so you are never locked into one trade-off. Whisper covers 99 languages with GPU acceleration; Parakeet trades that breadth for raw CPU speed. Both run entirely on your device: pick per session, or set a default and forget it.

Whisper: the accurate all-rounder

Whisper runs as whisper.cpp, compiled directly into WhisPaste via dart:ffi: no subprocess, no server. It covers 99 languages in three quality tiers (compact, balanced, premium), so you can trade download size and speed for accuracy. On supported hardware it accelerates via Vulkan on Windows and Linux or Metal on Apple Silicon, and falls back to CPU automatically if a GPU crashes or is unavailable.

Parakeet: the speed champion

Parakeet runs NVIDIA's Parakeet-TDT model through sherpa-onnx, also fully in-process. It covers around 25 European languages (a smaller set than Whisper) but runs purely on CPU, no GPU required, and measures several times faster than Whisper on identical hardware (a real-time factor of roughly 0.10–0.13× versus Whisper's 0.43–1.16×). The one-time model download is about 640 MB.

Which engine should you use?

Reach for Whisper when you need maximum language coverage or the highest accuracy, and you have a capable GPU or don't mind a CPU render that takes closer to real-time. Reach for Parakeet when your hardware is modest, you work mainly in a supported European language, and speed matters more than the last percent of accuracy. You can switch between them any time in Settings → Speech-to-Text, no reinstall, no data loss.

How much faster is Parakeet than Whisper?: In internal measurements, Parakeet reaches a real-time factor of about 0.10–0.13× on CPU, while Whisper measures 0.43–1.16× depending on model size and backend, meaning Parakeet is several times faster at transcribing the same audio, even without a GPU.
Which languages does each engine support?: Whisper supports 99 languages, including mixed-language input and code-switching mid-sentence. Parakeet covers around 25 European languages. If your language isn't on Parakeet's list, Whisper is the one to use.
Can I switch engines without losing my settings?: Yes. Both engines live under the same "On Device" provider in Settings → Speech-to-Text, switching is a single toggle, and your hotkeys, snippets, and history are untouched.