OpenAI Speech

Redock can use OpenAI speech-to-text models to turn recordings into terminal input. This is useful when you dictate longer prompts, Git messages, test plans, or mixed-language technical text on mobile.

What Redock Calls

Redock sends audio to:

POST https://api.openai.com/v1/audio/transcriptions

The request uses:

• Authorization: Bearer <OpenAI API Key>
• multipart/form-data
• Audio field: file
• Response format: json
• Model: the OpenAI speech model selected in Redock

Redock records 16 kHz mono WAV audio before sending it.

Prepare an API Key

1. Open OpenAI Platform.
2. Confirm the selected organization and project can call the API.
3. Confirm billing, quota, or project budget is available.
4. Create or copy a secret key from the API Keys page.
5. Keep the key private.

https://platform.openai.com/api-keys

Redock stores the API key in Keychain. It is not written into app data snapshots.

Fill It in Redock

1. Open My > Speech.
2. Set Speech Provider to OpenAI.
3. Choose a model.
4. Paste the OpenAI API Key.
5. Return to a terminal workspace and test the speech button.

Model Choices

• Fast Transcribe: gpt-4o-mini-transcribe, the default for daily use.
• Accurate Transcribe: gpt-4o-transcribe, better when quality matters more.
• Diarize: gpt-4o-transcribe-diarize, mainly for speaker separation. Redock currently reads the final text field and does not show speaker labels.

Common Issues

• Missing API key: check My > Speech.
• Authentication failed: confirm the key was copied correctly and belongs to the current project.
• No permission: check project permissions, organization role, and API key scope.
• Quota or billing error: check Platform billing and project budget.
• Poor mixed-language results: try Accurate Transcribe and speak commands or paths more clearly.

Official References

• OpenAI Platform
• API Keys
• Speech to text guide
• Audio transcription API reference