Dictate coding prompts and technical text on Mac

Give your prompt a clear shape

1. Place the cursor in the input field where you want the text and start attunz dictation.
2. Say the goal, the relevant component or API names, and any constraints. For documentation, say who it is for and what it should explain.
3. Review the text, correct important details, then submit it to the tool you are using.

Try a coding prompt or documentation request

Hypothetical coding prompt: “In UserSettings, add an empty state when the API returns no projects. Keep the existing loading and error states. Explain the proposed change and suggest tests before editing code.”

Hypothetical documentation request: “Document how retryAfterMs is used after an HTTP 429 response. Include the unit and one example.”

These examples illustrate text you could dictate. Check that UserSettings, retryAfterMs and HTTP 429 match your actual task before submitting the request.

Understand the context attunz uses

During dictation, attunz can automatically use the app name, AX window text, selected text, existing input-field text and available caret position as needed. Window text comes from macOS Accessibility (AX), not screenshots. Available text and caret information depend on the app and permissions; they help AI enhancement understand the input scenario and refine ASR text and formatting.

Clipboard capture for model context is off by default. Automatic window context does not scan project source files. Importing documents into memory is a separate action you choose; permitted local Claude/Codex history is another distinct input. You can turn off automatic context collection.

Check the details that change meaning

Review identifiers, capitalization, file paths, numbers, units and punctuation. Check constraints such as “do not change the public API” especially carefully. If you dictate command text in a terminal, inspect it before executing it.

For recurring names and specialist terms, read the memory guide. For local model setup and the difference between local and cloud processing, read the offline dictation guide. Context collection and processing location are separate questions.

Set up AI enhancement and preview your prompt

Enable AI Enhancement in Settings and choose a model. After dictation, it follows your prompt to correct recognition errors, remove filler and repetition, and adjust tone and formatting to the task. Short text can bypass editing, depending on your settings.

Open AI Enhancement → Prompt to set tone and formatting, then preview the result in the test area. You can insert context variables such as the current app. You can also edit personal writing preferences in your memory style profile.

The Prompt test area uses your selected enhancement model. If it is a cloud model, running the test sends the test text to that provider. Use a nonsensitive sample and review the output; a test preview is not a guarantee of the result in every app.

All Attunz software features are free forever with local models or BYOK (your own API key). Attunz charges no fees for these options. You pay any third-party model fees directly to the provider. The Attunz hosted cloud service uses manual usage-credit top-ups for optional cloud ASR and AI enhancement, with no automatic renewal at present. It is available. Pricing

Choose a voice input workflow

Live voice conversations let you talk back and forth with an agent. Dictation creates editable prompt text that you can review before submitting. Native tools can support one or both; check the particular app or CLI rather than treating all voice features as the same. A system-wide Mac dictation workflow can help when you move between prompt fields, editors and notes.

Check Claude Code voice requirements

Claude Code’s native dictation is tuned for coding terms and uses project and branch names as hints. It requires Claude.ai authentication and sends audio to Anthropic for transcription. Direct API-key and third-party provider authentication are not supported for this feature; its documentation also excludes SSH/cloud and VS Code Remote sessions. Check the current documentation for your surface.

Claude Code voice requirements

Codex already supports native dictation. The Codex app’s May 5, 2026 update added cleanup and a dictionary for names, paths and code symbols. Separately, CLI release 0.156.0 on September 22, 2026 enabled voice conversations by default. App dictation and CLI conversations are different surfaces and workflows.

Codex app changelog · Codex CLI 0.156.0 release

Choose what you need across apps

• One app: check its native voice controls first. Claude Code and Codex offer native options, and Cursor’s Agents Window also has voice input. Free, local or BYOK options are not exclusive to Attunz.
• Several Mac apps: use a dictation workflow suited to moving between your editor, prompt field and technical notes. Text insertion still depends on the target app and its permissions.
• Local processing: follow the offline guide for local speech recognition, enhancement and memory settings. Review which services receive your text when you submit it.

Cursor’s Agents Window · Offline guide

For work accounts, use only apps and processing services allowed by your organization. Check the text before submitting a prompt or executing a command.

Common questions

Related guides

• Names and technical terms in dictation
• Set up offline dictation on Mac