Comment Monter un Podcast : Le Guide Complet de la Prise Brute au Mastering

The 7-Stage Spoken-Word Post-Production Chain

Editing a podcast is fundamentally different from mixing music. In spoken-word audio, listener intimacy and cognitive ease are paramount. A listener will wear in-ear headphones for 45 minutes straight; any harsh sibilance, electrical hum, sudden volume leaps, or robotic silence cuts will create severe listening fatigue.

Why Processing Order Dictates Final Audio Clarity

The most common reason podcast audio sounds muddy, harsh, or distorted is incorrect signal order. If you compress or EQ a track before removing low-end rumble and mouth clicks, the compressor will clamp down on inaudible sub-bass vibrations and the EQ will amplify high-frequency saliva pops.

The Law of Audio Post-Production: "Perform corrective surgery first (rumble, hum, clicks, dead air), dynamic balancing second (expansion, gentle compression), tonal enhancement third (vocal EQ), and loudness mastering last (LUFS normalization & peak limiting)."

Phase-by-Phase Deep Dive: Raw Recording to Master

Étape 1 : Ingest, Raw Backup & Gain Staging

Always duplicate your raw recording files before touching an edit. Store your raw multi-track recordings in a read-only archive folder. Verify that your peak recording level sits between -18 dBFS and -12 dBFS, leaving at least 12 dB of headroom to avoid digital clipping during subsequent processing.

Étape 2 : Silence Triage & Pacing (0.32s Cushion)

Load your audio into the Podcast Mixing Studio. Automated silence detection scans for gaps where speakers paused to check notes, take a sip of water, or endure remote VoIP latency.

• Threshold: Set to -38 dB (Balanced Sensitivity).
• Minimum Gap: Set to 0.80 seconds.
• Retained Cushion: Apply the 0.32-second Natural Speech Cushion so vocal decay and breathing beats are preserved.

Étape 3 : Sub-Rumble High-Pass & 50/60Hz Hum Notch

Human speech rarely produces meaningful acoustic fundamental frequencies below 85Hz for men and 120Hz for women.

• 80Hz 2nd-Order Butterworth High-Pass: Strips HVAC drone, passing traffic, mic handling bumps, and desk vibrations without affecting vocal resonance.
• 50Hz / 60Hz Notch Filter: Eliminates electrical ground-loop hum caused by AC power lines, audio interface power supplies, or USB ground loops.

Étape 4 : Mouth Sound & Saliva Pop De-Clicking

Saliva bubbles and dry palate clicks produce ultra-fast, high-frequency transients between 2kHz and 10kHz lasting 1ms to 4ms. Run our cubic spline de-clicker to detect transient energy bursts exceeding 4.5x running standard deviation and smoothly interpolate across the damaged samples without dulling consonant clarity.

Étape 5 : Breath Attenuation & Downward Expansion

Never delete 100% of speaker inhalations—doing so produces an uncanny, suffocating sensation. Instead, apply -6 dB to -12 dB attenuation or a gentle downward room-tone expander that softens loud gasps into natural, imperceptible breath intakes.

Étape 6 : Broadcast Vocal EQ & Warmth

Shape the dialogue tone for broadcast presence:

• Low-Mid Body (200Hz): +1.0 dB to +1.5 dB subtle boost for vocal warmth.
• Boxiness Cut (400Hz–600Hz): -1.5 dB narrow dip to eliminate room reflections.
• Speech Presence (3.5kHz): +2.0 dB wide bell boost for vocal articulation.
• Air Sheen (10kHz High-Shelf): +1.0 dB subtle shelf for modern studio sheen.

Étape 7 : ITU-R BS.1770 Loudness & True-Peak Limiting

The final stage balances perceived loudness to international broadcast standards:

• Stereo Target: -16.0 LUFS Integrated (±1.0 LUFS).
• Mono Target: -19.0 LUFS Integrated (±1.0 LUFS).
• True-Peak Ceiling: -1.0 dBTP to prevent inter-sample distortion when listeners' podcast apps transcode your WAV master to MP3 or AAC.

Podcast Audio Standards Matrix: Major Platforms

Mastering to the correct standard ensures podcast players don't forcibly turn down your volume or introduce limiting artifacts:

Step-by-Step Editing Workflow in AutoSilenceCut

1. Import Recording: Open the Podcast Mixing Studio and drop in your WAV or MP3 recording.
2. Run Silence & Cadence Scanner: Review detected dead-air gaps. AutoSilenceCut automatically protects in-sentence dialogue and applies a 0.32s speech cushion.
3. Enable Studio DSP Rack: Toggle on 80Hz High-Pass, 50/60Hz Hum Notch, Vocal Presence EQ, and True-Peak Limiter.
4. Audition and Room-Tone Bridge: Listen to transition splices. Enable Room Tone Matching so trimmed pauses blend seamlessly with ambient room acoustics.
5. Normalize to -16 LUFS & Export: The engine scales gain to exact ITU-R BS.1770 standards and exports an uncompressed 16-bit PCM WAV master.

Cas concret : Dual-Host Interview

A 45-minute remote interview recorded with a host in London (50Hz power grid) and a guest in New York (60Hz power grid) had noticeable electrical hum and remote latency.

• Hum Removal: Dual-frequency notch filtering stripped 50Hz and 60Hz ground hum without coloring vocal formants.
• Latency Shortening: 42 remote pauses averaging 1.8s were shortened to 0.35s turn-taking cushions, eliminating 4 minutes and 12 seconds of sluggish drag.
• Loudness Matching: The guest's quiet USB mic (-24 LUFS) and host's studio mic (-15 LUFS) were balanced and normalized to a uniform -16.0 LUFS Integrated master.

Erreurs fréquentes à éviter lors du traitement vocal

• Mistake 1: Chasing "As Loud As Possible": Over-limiting dialogue to -10 LUFS destroys dynamic expression, amplifies room reflections, and causes streaming services to harshly turn down your audio.
• Mistake 2: Hard Noise Gating: Gating to absolute digital black sounds unnatural and creates noticeable room-noise pumping.
• Mistake 3: Zero-Length Pauses: Splicing words together with no cushion makes conversations sound breathless and anxiety-inducing.
• Mistake 4: Not Testing on Phone Speakers and In-Car Audio: Always do a final QA pass on cheap earbuds, phone speakers, and a noisy car stereo to ensure dialogue intelligibility.