Platform edition
Cohere Transcribe (Local) for Windows
Cohere Transcribe 03-2026 offline on your PC through a managed CrispASR runtime, on the CPU or a GPU with CUDA or Vulkan.
- Version
- 1.2.4
- Plugin ID
- com.
typewhisper. cohere-transcribe - Requirements
- TypeWhisper 1.1.0 or newer · Windows 10 or newer · Disk space for a 1.5 to 2.4 GB model plus the CrispASR runtime, which is about 0.7 GB for NVIDIA CUDA
Inside this edition

Features in this edition
- Local speech-to-text processing
- Processing without uploading audio or text
- Four model sizes from about 1.5 to 2.4 GB, with Q5_0 recommended
- 14 languages with automatic language identification
- Live preview while you speak, with the final text after you stop
- Processing on the CPU, an NVIDIA GPU with CUDA, or a GPU with Vulkan
- Model and runtime downloads checked against pinned SHA-256 values
Settings in this edition
- Optionally enter a Hugging Face token and select Save token
- Select Download model for Cohere Transcribe 2B (Q5_0) or another size, then Use model
- Choose the Processing device for the next model load and select Save settings
- Use Unload and remove selected model to delete the selected model's files
Guide for this edition
Set up the add-on and verify its platform-specific capabilities with a first practical workflow.
Roles in TypeWhisper
Setup
- In TypeWhisper, open Settings > Discover plugins and install Cohere Transcribe (Local). The add-on then appears in the Settings sidebar, and its page opens. If it is already installed, open its page in the sidebar.
- On the add-on's page, review the options available in this edition.
- Optionally enter a Hugging Face token and select Save token
- Select Download model for Cohere Transcribe 2B (Q5_0) or another size, then Use model
- Choose the Processing device for the next model load and select Save settings
- Use Unload and remove selected model to delete the selected model's files
- Under Settings > Dictation, select Cohere Transcribe (Local) as the Provider and choose a Model. A Dictation Only workflow can also use its own transcription model.
In practice
- Transcription
- Dictate a few sentences with Cohere Transcribe (Local) and check the language, model, and result before you use it for longer recordings.