Getting Started with Voice Coding: A Practical Guide to Ummless

9 min read · March 7, 2026

Getting Started with Voice Coding: A Practical Guide to Ummless

Ummless is a macOS tray application that lets you dictate text anywhere on your computer, refine it with AI, and paste the result — all triggered by a single keyboard shortcut. This guide walks you through installation, configuration, and your first productive dictation session.

Prerequisites

Before installing Ummless, make sure you have:

• macOS 14 (Sonoma) or later — Ummless uses Apple's on-device SFSpeechRecognizer, which requires macOS 14+.
• Microphone access — A built-in or external microphone. Any mic that works with macOS will work with Ummless.
• An Ummless account — Sign up at ummless.com to enable AI refinement features and sync presets across devices.

No additional software, runtimes, or dependencies are required.

Installation

Step 1: Download Ummless

Download the latest version from ummless.com/download. The download is a standard .dmg disk image for macOS.

Step 2: Install the Application

1. Open the downloaded .dmg file
2. Drag Ummless to your Applications folder
3. Eject the disk image

Step 3: First Launch

Open Ummless from your Applications folder. On first launch, macOS will prompt you for two permissions:

Microphone Access Ummless needs microphone access to capture your speech. Click "Allow" when prompted. Without this permission, dictation will not work.

Speech Recognition Apple requires explicit permission for apps to use the SFSpeechRecognizer framework. Click "Allow" to enable on-device speech recognition. This permission is specifically for local processing — your audio does not leave your device.

Accessibility (Optional) For the global hotkey to work in all applications, Ummless may request accessibility access. Navigate to System Settings > Privacy & Security > Accessibility and enable Ummless.

After granting permissions, Ummless will appear as an icon in your menu bar. It runs as a tray application — there is no dock icon or main window. Everything happens through the menu bar icon and the dictation palette.

Step 4: Sign In

Click the Ummless icon in your menu bar and select "Sign In." This opens a browser window where you can authenticate with your Ummless account. Signing in enables:

• AI-powered text refinement
• Preset syncing across devices
• Custom preset creation and management

You can use Ummless without signing in for basic transcription (no AI refinement), but the experience is significantly better with an account.

Your First Dictation

Opening the Palette

Press Cmd+Shift+Space (the default global hotkey). A compact dictation palette appears in the center of your screen. The palette is transparent and always-on-top, so it does not obscure your current work.

Recording

The palette opens in a ready state. Click the microphone button or simply start speaking — Ummless begins capturing audio immediately. As you speak, you will see a real-time transcript appear in the palette. This is the raw, unrefined output from Apple's speech recognizer.

Try dictating something simple to start:

"This is my first dictation with Ummless. I want to see how well it handles natural speech including pauses and filler words."

Stopping the Recording

Press Cmd+Shift+Space again or click the stop button in the palette. Ummless will stop recording and, if you are signed in and have a preset selected, automatically send the transcript for AI refinement.

Viewing the Result

After a brief processing moment, the refined text appears in the palette. You will see the difference immediately — filler words removed, punctuation added, sentences properly structured.

Copying to Clipboard

The refined text is automatically copied to your clipboard. Switch to any application and paste with Cmd+V. That is the complete workflow: hotkey, speak, hotkey, paste.

Choosing and Managing Presets

Presets control how your raw transcript is refined. Choosing the right preset for each situation is the key to getting useful output consistently.

Built-in Presets

Ummless includes several built-in presets designed for common use cases:

Clean Transcript Minimal processing. Removes filler words ("um," "uh," "like") and adds basic punctuation. The output closely mirrors your speaking style. Use this when you want light cleanup without significant restructuring.

Professional Email Transforms your dictation into a properly formatted email. Adds greeting and sign-off structure, formalizes language, and organizes thoughts into clear paragraphs. Ideal for composing emails to colleagues, clients, or stakeholders.

Casual Message Light refinement that preserves conversational tone. Perfect for Slack messages, Discord, or texts where formal language would feel out of place. Keeps contractions, allows sentence fragments, and maintains your natural voice.

Technical Documentation Structures output with clear headings, precise technical language, and consistent terminology. Expands abbreviated phrasing into complete explanations. Use this for README files, API documentation, or technical guides.

Concise Summary Distills your spoken input into a brief, focused summary. Strips tangential remarks and redundancy, keeping only the core message. Good for executive summaries, meeting notes, or status updates.

Creative Writing Preserves literary flourishes, metaphors, and stylistic choices while cleaning up verbal artifacts. Does not over-formalize or flatten your prose. Use this for blog posts, stories, or any creative content.

Selecting a Preset

In the dictation palette, click the preset selector dropdown before or after recording. The selected preset is remembered between sessions, so you only need to change it when switching contexts.

Creating Custom Presets

Custom presets let you tailor refinement to your exact needs. To create one:

1. Open the Ummless dashboard at ummless.com/dashboard/presets or click "Manage Presets" from the tray menu
2. Click "Create Preset"
3. Fill in the preset details:
   • Name — A short, descriptive label (e.g., "PR Description," "Standup Update")
   • Description — What the preset is for (helps you remember when selecting)
   • Instructions — The core prompt that guides refinement behavior

Here is an example custom preset for git commit messages:

Name: Git Commit Message
Description: Formats dictation as a conventional commit message

Instructions: Transform the input into a git commit message following
conventional commits format. First line should be a concise summary
under 72 characters in imperative mood (e.g., "Add," "Fix," "Refactor").
If the speaker provided detail about why or how, include it as a body
paragraph separated by a blank line. Preserve all technical terms,
file names, and function names exactly. Remove all filler words and
conversational language.

Custom presets sync to your account and are available on all your devices.

Tips for Better Accuracy

Microphone Setup

The quality of your microphone directly affects transcription accuracy:

• Position — Keep the microphone 6-12 inches from your mouth. Too close causes plosive distortion; too far picks up room noise.
• Environment — A quiet room with soft surfaces (carpet, curtains) produces better results than a hard-walled room with echo.
• External mics — If you dictate frequently, a USB condenser microphone or a headset mic will outperform your laptop's built-in array.

Speaking Technique

Natural speech works well, but a few adjustments improve accuracy:

1. Speak at a natural pace — Do not slow down artificially. The speech recognizer is tuned for natural cadence.
2. Enunciate technical terms — Words like "Kubernetes," "PostgreSQL," or "OAuth" benefit from clear pronunciation the first few times. The recognizer adapts.
3. Pause between distinct thoughts — A brief pause (half a second) helps the recognizer identify sentence boundaries.
4. Avoid covering your mouth — Sounds obvious, but habits like resting your chin on your hand muffle audio significantly.
5. Stay consistent — The on-device recognizer learns from your speech patterns. Consistent daily use improves accuracy over time.

Handling Recognition Errors

When the recognizer gets a word wrong:

• Do not restart — Finish your thought, then fix the one word in the refined output. It is faster than re-dictating.
• Check the refined output — AI refinement often corrects recognizer errors using context. "Post gress" in the raw transcript may become "Postgres" after refinement.
• Add domain terms to your preset — If your preset instructions mention specific technical terms, the refinement model is more likely to correct misrecognitions.

Integrating Into Your Daily Workflow

The key to adopting dictation is identifying the moments where you currently type prose and replacing them with speech. Start with the lowest-friction use cases:

Week 1: Messages

Use Ummless for all Slack messages, emails, and chat replies. This is the easiest starting point because the stakes are low and the time savings are immediate.

Week 2: Git Workflow

Add dictation to your commit messages and PR descriptions. Create custom presets for each. You will notice your commit history becoming significantly more descriptive.

Week 3: Documentation

Start dictating documentation, README updates, and inline code comments. Use the Technical Documentation preset or create domain-specific ones.

Week 4: Everything Else

By now, dictation should feel natural. Use it for meeting notes, design documents, issue descriptions, and any other prose you write during your day.

Keyboard Shortcuts Reference

Action	Shortcut
Toggle dictation palette	Cmd+Shift+Space
Start/stop recording	Cmd+Shift+Space (when palette is open)
Copy result	Automatic (also Cmd+C)
Close palette	Escape

The global hotkey can be customized in Ummless settings if Cmd+Shift+Space conflicts with another application.

Troubleshooting

"Microphone access denied" Go to System Settings > Privacy & Security > Microphone and ensure Ummless is enabled.

"Speech recognition unavailable" Check System Settings > Privacy & Security > Speech Recognition. Also verify you are running macOS 14 or later.

Hotkey does not work in certain apps Some applications (particularly those running with elevated privileges) may block global hotkeys. Grant Ummless accessibility access in System Settings > Privacy & Security > Accessibility.

Refinement is slow or unavailable AI refinement requires an internet connection and an active Ummless account. Basic transcription (without refinement) works offline.

Poor accuracy with technical terms Create a custom preset that explicitly lists the technical vocabulary you use. This guides the refinement model to correct common misrecognitions.

You are now set up and ready to dictate. The initial awkwardness of speaking instead of typing fades quickly — most users report that dictation feels natural within a few days. Start with messages, expand to documentation, and let the habit build from there.