Voxr is designed to be straightforward: activate, speak, done. But there are a few things you can do to get noticeably better results. Here are practical tips from regular use.
Getting better transcriptions
Speak at a natural pace. You don’t need to slow down or over-enunciate. Apple’s speech recognizer is trained on natural speech patterns, and speaking naturally actually produces better results than speaking artificially slowly.
Pause between sentences. Brief pauses help the recognizer identify sentence boundaries, which leads to cleaner output from the LLM processing step. You don’t need to say “period.” Just pause naturally where you’d put a period in writing.
Reduce background noise. This is the single biggest factor in transcription quality, since everything is processed locally on your Mac. A quiet room makes a dramatic difference. If you’re in a noisy environment, consider using a directional microphone or headset mic rather than your MacBook’s built-in microphone.
Use a dedicated microphone. Your Mac’s built-in mic works fine in a quiet room, but an external microphone, even a basic USB one, will improve accuracy noticeably, especially if you’re more than a couple feet from your laptop.
Think before you speak. Unlike typing, where you can easily delete and retype, voice input works best when you have a rough idea of what you want to say before you start. Take a second to organize your thoughts, then speak in complete sentences.
Environment tips
Grant permissions upfront. The first time you use Voxr, macOS will prompt you for Microphone and Speech Recognition permissions. You may also need to grant Accessibility access in System Settings > Privacy & Security for the paste functionality to work in all applications.
Workflow tips
Position your cursor first. Before activating Voxr, make sure your cursor is in the text field where you want the text to appear. Voxr pastes into whatever app has focus, so have your destination ready.
Use it for drafts, not final copy. Voice input excels at getting ideas down quickly. Dictate a rough draft of an email or document, then do a quick manual edit pass. This is almost always faster than typing from scratch.
Dictate in chunks. For longer content, dictate a paragraph at a time rather than trying to speak an entire document in one go. This gives you natural breakpoints to review and lets the LLM process shorter, more manageable pieces of text.
Works in any app. Voxr pastes text through the clipboard, which means it works everywhere: Slack, VS Code, your browser, Notes, Mail, Pages, and any other app that accepts text input. There’s nothing special to configure per-app.
Getting the most from the LLM
The local language model handles cleanup automatically, but the quality of its output improves when your input is cleaner to begin with. The tips above about natural pacing, clear speech, and quiet environments aren’t just about the speech recognizer. They also give the LLM better raw material to work with.
If you find the output isn’t matching the style you want, try adjusting your speaking pace and clarity. The AI produces better results when given cleaner input to work with.