Dictation with visual references and a chosen finish

Before you begin

• Use the Denker desktop app on Mac or Windows. Global dictation and capture require the desktop app.
• Follow the welcome guide for microphone access and the permissions needed to capture your screen and paste into other apps. On Mac, these include Screen Recording and Accessibility; the speech engine may also request Speech Recognition access. Windows permission and capture behavior differ.
• Open Preferences → Voice. Choose your microphone, speech engine and the language you speak. A translation prompt chooses the written output language separately. Available recognition languages depend on the engine.
• Under Dictation, enable Dictate in any app. Open Shortcut and prompts, check your start/stop shortcut and save the prompts you want to use. Paste automatically controls whether Denker attempts to insert the result. AI rewriting needs a connected AI provider; plain dictation can use the speech engine without an AI rewrite.
• Try a short, harmless sample in the exact destination app and input field. Image and file support varies by app, field, platform and Denker build. A text field may accept words while refusing attachments.

For installation and AI connection, follow Getting Started. Use the controls shown in your installed app if they differ from this guide.

Start with a text-only message

Speech recognition turns your voice into text. An optional AI prompt rewrites or translates that text. Try one sentence first so you can see how your chosen destination receives it.

1. Click into an unsent draft in your destination app.
2. Use the start/stop shortcut shown in your dictation preferences and speak one sentence.
3. Keep the intended field active and use that shortcut again to stop with Default. To keep the recognized wording, turn off the Default rewrite switch before starting.
4. Inspect the destination. If Denker reports that the text was copied, paste it yourself into the intended field and check it there.

Dictation prepares text; you decide when to send or submit it. Correct recognition errors before relying on names, dates or amounts.

Explain two details with marked screenshots

1. Start dictation from the destination draft. Show the first detail, hold Shift and drag to mark it while explaining what matters.
2. Move to a distinct view and mark the second detail, then explain how it relates to the first. Multiple marks on an unchanged view can share one screenshot. Say “the first screenshot” and “the second screenshot” so your request remains understandable.
3. Return focus to the intended destination field. While listening, use the shortcut displayed beside a saved prompt to stop and apply it, or stop with Default.
4. Inspect the finished draft. Check that both marked images arrived and that each sentence still refers to the correct image.

A still image is useful for a visible state, such as a wrong total or an open menu. Capture can include the whole display with your marks, so move unrelated or sensitive information out of view first. Screenshots paste before the text and need a destination that accepts images.

If your installed build provides automatic numbered references, labels such as “[1]” can identify the associated screenshot. Check their placement and preserve the labels when rewriting. Otherwise, name the images explicitly in your spoken explanation, such as “the train tickets in the first screenshot.” Check references again when starting a separate dictation or draft.

Recording attachments and their limits

If your installed app offers recording during dictation, use a short clip to show a sequence of actions and describe those actions aloud. Try the available controls with a short capture first, then replay the saved clip before including it in your message.

A local path identifies a file on your computer. To share the recording, attach or upload the actual file through the destination app. Open the accepted attachment there before sending so the recipient receives a clip they can use. Mac and Windows attachment behavior can differ.

The rewrite prompt organizes or translates the words you dictate. It leaves the attached screenshots and recording unchanged. Video editing, subtitle creation and translating text inside an image need a separate workflow.

Rewrite and translation prompts you can inspect

Save an instruction in Shortcut and prompts, then use the shortcut shown beside it while listening. The examples below pair an illustrative input with a prompt and the kind of draft to aim for. Review the resulting wording, facts and meaning before using it.

Review and recover a missing result

• Read the destination before retrying. Keep any text or attachments that already arrived so a second paste does not create duplicates.
• In Preferences → Voice → Recent dictations, inspect the saved entry promptly; history keeps the last three entries locally. Copy retrieves the original text and Copy improved retrieves an available rewrite. Retry appears after a failed or interrupted AI step and copies a new result; it does not paste the attachments again. Review it before replacing the existing draft.
• If a screenshot is missing, click its saved screenshot preview to copy that image, then paste it into an image-capable destination. Check the image and its reference after the paste.
• If a clip is missing, open the saved recording where available and attach the actual file through your destination app. A preview or path alone is not the recording.
• If AI rewriting fails, preserve the recognized original and use it as a draft, or reconnect your provider and retry from history. A translation request can fail back to the original language; check the language before sending.
• If capture never started, check the microphone, speech engine and relevant system permissions, then try a short text-only sample. Avoid repeating the whole message until you know what reached the destination.

For a useful handoff, check meaning, facts, screenshot order, attachment availability and recipient access separately. Send only after the completed draft says and shows what you intended.

Availability and examples

Use dictation prompts and marked screenshots in compatible apps. Text, image and file support depends on the destination field and your installed build. Start with a short sample in the app you plan to use, then check the finished text and accepted attachments.

Read the October announcement for the feature overview. Its films illustrate the workflows with fictional sample data.