Use your own OpenAI API key
Set up an OpenAI API key in WhisperARC for Windows. Select speech-to-text and Assistant models, understand separate API billing and troubleshoot errors.
Last updated
Prepare your OpenAI account
OpenAI KEY is available on every WhisperARC plan, including Free. It sends audio and text requests directly to OpenAI using your own API account. It requires an internet connection and does not consume WhisperARC Cloud credits.
WhisperARC handles the API calls. You do not need to install an SDK, set an environment variable or write a script to use this source.
- Sign in to the OpenAI platform and select the project you want to use.
- Create a project API key with access to the models you intend to use.
- Check that the project's billing and usage settings allow API requests.
Save the key and activate OpenAI KEY
- Open Model in WhisperARC and select OpenAI KEY.
- Paste the key into API key and choose Save API key.
- Wait for verification to finish.
- Select the OpenAI KEY card again to activate this source.
- Choose Transcription and Assistant models separately.
- Try a short dictation or import a short recording, then review the result.
Choose Transcription and Assistant models
Transcription turns speech into text. The initial selection is GPT-4o mini transcribe. The supported models below are the current WhisperARC catalogue; access also depends on your OpenAI account.
Assistant works on text for dictation modes and microphone translation. Its model is selected separately. A dictation mode takes priority over translation; imported files remain in their detected source language.
| Transcription model | What to check |
|---|---|
| GPT-4o mini transcribe | Default transcription selection in WhisperARC. |
| GPT Transcribe | Current OpenAI transcription option offered by the app. |
| GPT-4o transcribe | Alternative transcription model; compare a representative sample. |
| Whisper-1 (Legacy) | Supports the detailed timestamps requested by WhisperARC for file transcription. |
| GPT-4o transcribe diarize | Provides speaker labels when they are returned by the model. |
Understand API billing
OpenAI bills API usage to your OpenAI account. These charges are separate from WhisperARC's paid plans and Cloud credits. Rewriting or translation can add text-model usage to audio transcription.
Consult current OpenAI pricing and your Usage dashboard before long recordings or batches. Keep a short test representative of the transcription and Assistant settings you will actually use.
Replace or remove a saved key
A verified key is locked for editing. To replace it, use the delete button, paste the new key and choose Save API key again. If OpenAI KEY was active, removing its key returns the selected source to Local.
Resolve key, access and credit errors
Use the message shown by WhisperARC to choose the next step. Processing errors do not automatically retry with a different source.
| Message | Next step |
|---|---|
| Invalid OpenAI API key | Check whether the key was revoked; remove and replace it. |
| OpenAI denied access | Check the project's permissions and selected model access. |
| OpenAI request limit reached | Wait before retrying. |
| Insufficient OpenAI credits | Review OpenAI billing, credit balance and spending limits. |
Audio, text and local storage
The saved key is kept in encrypted local settings. Audio is sent directly to OpenAI for transcription; text is sent for Assistant transformations or translation. History remains local.
OpenAI says API data is not used for training by default unless you opt in. Provider retention depends on the endpoint and account controls. Use Local Mode when you want recognition to run on this device.