Skip to main content
Superwhisper recording window showing live realtime transcription while dictating

​
Overview

Realtime transcription shows your words on screen as you speak, so you get immediate feedback on accuracy. It works only with voice models that support streaming:
  • Nova (cloud): Nova 3, Nova 2, and Nova Medical. These need an internet connection
  • Scribe (cloud, ElevenLabs): needs an internet connection
  • Parakeet (local): the Parakeet models run on your device. On Windows, only the two Parakeet Real-time models support it. See Realtime on Windows

​
Realtime on Windows

These voice models support Realtime on Windows:
  • Nova 3, Nova 2, and Nova Medical (cloud)
  • Scribe (cloud)
  • Parakeet Real-time (English) and Parakeet Real-time (Multilingual) (local)
The two Parakeet Real-time models are experimental. They appear in the model list only when Settings → AI Models → Show experimental models is on. Each is a 2.4 GB download. They run through ONNX Runtime, so they don’t need Vulkan. The standard Parakeet model on Windows doesn’t support Realtime. Cohere Transcribe, the Whisper models, and S1-Voice don’t support it either.

​
Why the toggle is missing or grayed out

  • Voice model: your mode uses a model that doesn’t support streaming, such as Cohere Transcribe or the Whisper models. Switch to one of the models above. On Windows, see Realtime on Windows.
  • License: Realtime needs a Pro license.
  • Identify Speakers: for Scribe and Nova models, Realtime and Identify Speakers can’t be on together. Turn Identify Speakers off to use Realtime, or keep it on and leave Realtime off.

​
Enable realtime transcription

1

Open the mode

Open Superwhisper and go to the mode you want to edit
2

Select a realtime-capable voice model

In the mode editor, set Voice Model to a Nova, Scribe, or Parakeet model
3

Enable Realtime

With a realtime-capable model selected, the Realtime toggle becomes available. Turn it onSuperwhisper mode editor with a Parakeet model selected and Realtime enabled