ロボット声にさせない！ポッドキャスト・ナレーションの自然なブレス音減衰法

The Physiology of Speech: Why Total Breath Removal Sounds Unnatural

Breathing is the physical engine of human speech. When we speak, our diaphragm contracts, air fills our lungs, and vocal fold phonation begins during the expiratory phase.

When novice editors purchase automated cleanup tools, their first impulse is often to delete every audible breath from the timeline. The resulting audio is acoustically unnatural and emotionally alienating.

The Subconscious Suffocation Response in Listeners

Human mirror neurons subconsciously synchronize with the breathing rhythms of the people we listen to. When an audio track strips every inhalation, the listener's brain detects an impossible biological anomaly: an individual delivering continuous paragraphs without ever taking a breath.

This creates subconscious cognitive tension, physical fatigue, and an uncanny "suffocation" sensation that pulls the listener out of the story.

Inspiratory Beats: Rhythmic Timing and Emotional Sincerity

In audiobooks and narrative podcasts, an intake of breath conveys vital emotional context:

• A sharp, rapid intake signals sudden surprise or excitement.
• A slow, deep breath signals solemn reflection or emotional gravity.
• An audible exhale communicates relief, resignation, or humor.

Stripping these sounds flattens the performer's acting and destroys the natural rhythmic meter of the prose.

Attenuation vs Deletion: The Golden Rule of Breath Editing

The standard in professional voiceover post-production is attenuation, never elimination.

The -6 dB to -12 dB Attenuation Sweet Spot

Instead of cutting the breath down to digital silence (-inf dBFS), reduce the amplitude of the breath passage by 6 dB to 12 dB:

• At -6 dB: The breath is halved in acoustic energy, sitting comfortably beneath conversational dialogue.
• At -12 dB: The breath is soft and subtle—inaudible on laptop speakers, but maintaining organic warmth and flow in high-end studio headphones.

Preserving Pre-Phonetic Inhale Cushions

When a breath immediately precedes an explosive consonant (such as "P", "T", or "K"), the air pressure build-up in the mouth is part of the consonant's attack envelope.

Cutting too close to the first syllable truncates this pressure onset, turning a crisp "People" into a clipped "'eople". Always preserve a 0.08s to 0.15s cushion before the start of the first spoken phoneme.

Decision Matrix: Breath Types, Severity & Recommended Action

Use this clinical decision matrix when reviewing dialogue in the Breath Remover:

Step-by-Step Breath Control Workflow in AutoSilenceCut

1. Import Audio Track: Drop your dialogue recording into the browser-local Breath Remover.
2. Run Automatic Transient & Breath Scan: The Web Audio engine detects quiet high-frequency inhalation passages flanking spoken phrases.
3. Select Attenuation Mode: Set gain reduction to -8 dB (Balanced Natural). This softens heavy gasps without dropping to dead silence.
4. Verify Speech Cushions: Confirm that the 0.32s Natural Cushion is engaged, ensuring breaths and speech decay blend smoothly into room ambience.
5. Audition Transitions: Click [Hear Before] on candidate breaths to verify that phoneme attacks remain pristine.
6. Export Lossless Master: Export an uncompressed 16-bit PCM WAV master with perfectly leveled natural dynamics.

Realistic Production Scenarios: Audiobook vs Punchy Commercial

Scenario A: Audiobook Narration (Audible / ACX Compliance)

A voice actor narrates a 6-hour historical novel. The condenser microphone is positioned 5 inches away, capturing rich low-end vocal warmth but also picking up audible chest inhalations.

• Why Deletion Fails: ACX quality inspectors reject audiobooks with "choppy gating" or unnatural dead-air pauses between sentences.
• The Solution: Attenuate all inhalations exceeding -30 dBFS by -9 dB, and preserve 0.25s of natural acoustic room tone.
• Result: Seamless, human, audiobook-grade audio that passes ACX automated QC on the first upload.

Scenario B: 15-Second High-Energy Social Video Ad

A fast-paced promotional ad requires a voiceover artist to deliver 42 words in exactly 14.2 seconds over a heavy electronic music bed.

• Why Attenuation Fails: Every millisecond of time is needed for spoken words; the music bed masks all silence gaps.
• The Solution: Remove 100% of breaths, tighten sentence cushions to 0.05s, and apply 8ms micro-crossfades.
• Result: Ultra-punchy, maximum-urgency advertising delivery.

音声編集・処理で避けるべき代表的な失敗と注意点

• Mistake 1: Hard Gating to Digital Black (-inf dBFS): Muting breath passages to absolute zero creates "room tone dropouts" where the room hiss abruptly cuts in and out like a broken wire.
• Mistake 2: Cutting Off Consonant Attack Phases: Truncating the pre-breath cushion chops off the unvoiced initial transient of words starting with "h", "p", or "t".
• Mistake 3: Over-Processing with Harsh De-Essers: Attempting to tame breath noise with a generic high-frequency de-esser dulls natural speech sibilance, making the speaker sound like they have a lisp.
• Mistake 4: Not Listening to Full Paragraphs: Always audition breath edits in full multi-sentence context to ensure natural human cadence is maintained.