Recording Voiceovers

4.1 Mic & Room Setup

What you're creating: A clean audio environment with the right mic, position, and room treatment for clear, reflection-free recordings.

Mic choice: USB dynamic mics like the Shure MV7 or Samson Q2U are excellent for untreated rooms — they naturally reject background noise and don't pick up every echo like cheap condenser mics

Mic distance: stay 10–12 cm away, use a pop filter, and speak past the mic — not directly at it — to reduce plosives on harsh "p" and "b" sounds

Room treatment: hard walls bounce sound — face a closet, hang a blanket, or record in a carpeted room to cut reflections — even DIY treatment makes a huge difference

4.2 Performance & Delivery

What you're creating: A natural, well-paced voiceover with deliberate pauses after on-screen actions.

Check for problem sounds

Say a few "p" and "b" words (plosives), and "s" and "sh" words (sibilance). If they distort, adjust your mic angle or distance.

4.3 Using AI Voiceovers

What you're creating: A tastefully directed AI voiceover — the right voice, varied tone by section, and precise delivery controls for natural pacing.

AI-generated voiceovers have come a long way — natural pacing, expressive tone, even subtle warmth. But tasteful AI voiceovers still takes craftsmanship. The secret lies in how you direct it.

Voice selection

Pick a voice that mirrors your brand personality. For most tutorials, that means calm, clear, and neutrally warm.

Avoid voices that sound like:

• Overly polished news anchors
• Cartoon mascots
• Hyper-casual YouTubers

A great AI voice should feel steady, human, and… forgettable in the best way. You want the message to stand out, not the voice itself.

Tone variation

Don't use one flat tone for everything. Vary the delivery style based on what's happening in the screen recording. Here's a simple breakdown:

This variation keeps the voiceover feeling dynamic and intentional.

Directing the delivery

Most AI voiceover tools support SSML-style (Speech Synthesis Markup Language) controls. It is a standard way to control how text is spoken — adjusting pitch, speed, pauses, emphasis, and pronunciation. While not every tool shows you raw tags, many still offer these controls under simpler names like "Add pause" or "Change pronunciation". Use them:

• Emphasis: Add <emphasis> or bold important verbs and nouns.
• Pauses: Insert <break time="500ms"> after big clicks, UI transitions, or to mark new steps.
• Adjust pitch/speed: maps to SSML's pitch, rate, or volume attributes
• Pronunciation: Add phonetic cues for tricky brand names or keyboard shortcuts.

Example:

• "Clueso" → kloo-zoh
• "Cmd+Shift+5" → Command Shift Five

If your tool doesn't support SSML, break up the text into short, clear sentences in your script so the AI handles pacing naturally.

Mixing AI + human voiceovers

Mixing human and AI can work, but don't alternate line-by-line; it's jarring. The trick is consistency within each section:

✅ What works:

• Human VO for intro + result
• AI VO for steps

❌ Avoid:

• Alternating every other sentence or step
• Mixing within the same paragraph

Switching too frequently also confuses viewers, especially in educational content where clarity is top priority.

4.4 Final Quality Checks

What you're creating: A QC pass that catches sibilance, audible breaths, and room tone mismatches before publishing.

Always QC your AI voiceover before publishing:

• Listen through laptop speakers and earbuds, not just system monitors.

Watch out for:

• Harsh sibilance (add a light de-esser to tame harsh "s/sh" sounds.)
• Audible breaths between lines
• Room tone mismatches (if mixing with human VO)
• Use light compression if your AI tool doesn't already apply it