Platform edition
Cloudflare ASR for macOS
OpenAI-compatible transcription behind a Cloudflare Tunnel with an Access service token.
- Version
- 1.0.6
- Plugin ID
- com.
typewhisper. cloudflare-asr - Requirements
- TypeWhisper 1.5.0 or newer · macOS 14.0 or newer
Inside this edition
Features in this edition
- Cloud speech-to-text processing
- Support for custom endpoints or proxies
- Cloudflare Access authentication with a service token
- Selection or refresh of available models
Settings in this edition
- Set the server URL for the OpenAI-compatible endpoint
- Store the CF-Access-Client-ID and CF-Access-Client-Secret
- Optionally store a bearer token for the service behind the tunnel
- Choose the transcription model in add-on or provider settings
Platform differences
On macOS, Cloudflare ASR connects to your OpenAI-compatible transcription backend through a Cloudflare Tunnel. On Windows, it calls Cloudflare Workers AI directly, after you sign in with Cloudflare in your browser or enter your account ID and an API token.
Setup on macOS: Cloudflare Tunnel
- Deploy an OpenAI-compatible transcription backend behind a Cloudflare Tunnel.
- If the service is protected by Cloudflare Access, create a service token for it.
- In TypeWhisper, open Settings > Discover plugins and install Cloudflare ASR. The add-on then appears in the Settings sidebar, and its page opens.
- Enter the backend’s base URL. TypeWhisper appends
/v1/modelsfor model discovery and/v1/audio/transcriptionsfor transcription, so do not enter the full transcription endpoint. - For an Access-protected service, enter CF-Access-Client-Id and CF-Access-Client-Secret.
- Refresh the backend’s models and select one, or enter the model name manually when discovery is unavailable.
- Select Cloudflare ASR as the transcription engine in Settings or a profile.
Audio is sent to the backend you operate through Cloudflare. Its model and processing behavior depend on that backend.
Setup on Windows: Workers AI
- In TypeWhisper, open Settings > Discover plugins and install the add-on, which is listed there as Cloudflare Workers AI. It then appears in the Settings sidebar, and its page opens.
- Select Connect with Cloudflare and sign in in your browser. If you granted access to several accounts, choose the Cloudflare account and select Save settings.
- Or enter your 32-character Cloudflare account ID, paste an API token with Workers AI access into API key, select Save settings, then Check connection.
- Under Transcription model, choose Whisper Large V3 Turbo or Whisper and, if it appears, select Use selected model.
The Windows plugin calls Workers AI under https://api.cloudflare.com/client/v4/accounts/{account-id}/ai/run/ itself. It offers Whisper, with automatic language detection, and Whisper Large V3 Turbo (@cf/openai/whisper-large-v3-turbo), which also accepts a spoken language and dictionary terms; no custom endpoint, Tunnel, or Access service token is needed. Audio is sent to Cloudflare Workers AI.