Audio Mixing Patterns
Comprehensive guide to audio mixing for video production using ffmpeg. Covers narration/music balancing, automatic ducking, timing control, and loudness normalization.
Core Principle
Quality Audio = Clear Narration + Supportive Music + Appropriate Levels
The human voice occupies 85-255 Hz (fundamental) with harmonics up to 8kHz. Music must support, not compete.
Volume Balancing Formula
Standard Video Mix Ratios:
--------------------------
Narration: 100% (reference level)
Music: 15-20% of narration level
SFX: 70-100% of narration level (contextual)
dB Relationships:
-----------------
Narration: -14 dB LUFS (dialogue standard)
Music bed: -30 to -35 dB LUFS (under narration)
Music only: -16 dB LUFS (no narration sections)
SFX: -18 to -20 dB LUFS
Volume Multiplier Quick Reference
| Ratio | Multiplier | Use Case |
|---|---|---|
| 100% | 1.0 | Full volume (narration) |
| 70% | 0.7 | Prominent SFX |
| 50% | 0.5 | Equal blend |
| 30% | 0.3 | Noticeable background |
| 20% | 0.2 | Subtle bed (recommended music) |
| 15% | 0.15 | Minimal presence |
| 10% | 0.1 | Barely audible |
Basic ffmpeg Mixing Commands
Two-Track Mix (Narration + Music)
# Basic mix: narration at full, music at 15%
ffmpeg -i narration.mp3 -i music.mp3 \
-filter_complex "[0:a]volume=1.0[narr];[1:a]volume=0.15[music];[narr][music]amix=inputs=2:duration=first" \
-c:a aac -b:a 192k output.m4a
Three-Track Mix (Narration + Music + SFX)
ffmpeg -i narration.mp3 -i music.mp3 -i sfx.mp3 \
-filter_complex "\
[0:a]volume=1.0[narr];\
[1:a]volume=0.15[music];\
[2:a]volume=0.7[sfx];\
[narr][music][sfx]amix=inputs=3:duration=first:weights='3 1 2'" \
-c:a aac -b:a 192k output.m4a
Timing with adelay Filter
The adelay filter positions audio at precise timestamps.
Syntax
adelay=delays[|delays...][,all=1]
# delays: milliseconds or samples (with 'S' suffix)
# all=1: apply same delay to all channels
Position Music at Specific Time
# Start music at 5 seconds
ffmpeg -i narration.mp3 -i music.mp3 \
-filter_complex "\
[0:a]volume=1.0[narr];\
[1:a]adelay=5000|5000,volume=0.15[music];\
[narr][music]amix=inputs=2:duration=first" \
output.m4a
Multiple Timed Audio Cues
# Narration starts at 0, music at 2s, SFX at 5.5s
ffmpeg -i narration.mp3 -i music.mp3 -i sfx.wav \
-filter_complex "\
[0:a]volume=1.0[narr];\
[1:a]adelay=2000|2000,volume=0.15[music];\
[2:a]adelay=5500|5500,volume=0.7[sfx];\
[narr][music][sfx]amix=inputs=3:duration=longest" \
output.m4a
Audio Ducking
Automatically lower music when speech is present.
Simple Sidechain Compression (Ducking)
ffmpeg -i narration.mp3 -i music.mp3 \
-filter_complex "\
[0:a]asplit=2[narr][sc];\
[1:a][sc]sidechaincompress=threshold=0.02:ratio=10:attack=50:release=500[ducked];\
[narr][ducked]amix=inputs=2:duration=first" \
output.m4a
Parameters Explained
| Parameter | Value | Effect |
|---|---|---|
| threshold | 0.02 (default 0.125) | Lower = more sensitive to speech |
| ratio | 10:1 | How much to reduce (10:1 = significant duck) |
| attack | 50ms | How fast to duck when speech starts |
| release | 500ms | How fast to return after speech ends |
| knee | 2.82843 | Softness of compression curve |
Advanced Ducking with Precise Control
ffmpeg -i narration.mp3 -i music.mp3 \
-filter_complex "\
[0:a]asplit=2[narr][sc];\
[1:a]volume=0.5[music_pre];\
[music_pre][sc]sidechaincompress=\
threshold=0.015:\
ratio=15:\
attack=30:\
release=800:\
makeup=1:\
knee=6[ducked];\
[narr][ducked]amix=inputs=2:duration=first:weights='1 0.4'" \
output.m4a
Mix Ratios by Content Type
Content Type | Narration | Music | SFX | Notes
---------------------|-----------|-------|------|------------------
Tutorial/How-to | 100% | 10% | 50% | Voice clarity critical
Corporate/Business | 100% | 15% | 60% | Professional presence
Social Media | 100% | 20% | 80% | Higher energy
Documentary | 100% | 25% | 100% | Cinematic feel
Promo/Advertising | 100% | 30% | 100% | Impactful
Music Video | 50% | 100% | 80% | Music dominant
Podcast | 100% | 5% | 30% | Minimal distraction
E-learning | 100% | 8% | 40% | Focus on retention
Loudness Normalization (LUFS)
LUFS (Loudness Units Full Scale) is the broadcast standard for perceived loudness.
Target Levels by Platform
| Platform | Target LUFS | True Peak | Notes |
|---|---|---|---|
| YouTube | -14 LUFS | -1 dB TP | Auto-normalized |
| Spotify | -14 LUFS | -1 dB TP | Loudness penalty applied |
| Apple Music | -16 LUFS | -1 dB TP | Sound Check |
| Broadcast TV | -24 LUFS | -2 dB TP | EBU R128 standard |
| Podcast | -16 to -19 LUFS | -1 dB TP | Apple spec |
| TikTok/Reels | -14 LUFS | -1 dB TP | Mobile optimization |
Loudness Normalization Command
# Normalize to -14 LUFS (YouTube/Spotify standard)
ffmpeg -i input.mp3 \
-af loudnorm=I=-14:TP=-1:LRA=11 \
-c:a aac -b:a 192k output.m4a
Two-Pass Normalization (More Accurate)
# Pass 1: Analyze
ffmpeg -i input.mp3 \
-af loudnorm=I=-14:TP=-1:LRA=11:print_format=json \
-f null - 2>&1 | grep -A 12 "output_i"
# Pass 2: Apply measured values
ffmpeg -i input.mp3 \
-af loudnorm=I=-14:TP=-1:LRA=11:\
measured_I=-18.5:measured_TP=-2.3:measured_LRA=8.2:\
measured_thresh=-28.5:\
linear=true \
-c:a aac -b:a 192k output.m4a
Multi-Track Production Pipeline
Complete Video Audio Mix
ffmpeg -i video.mp4 -i narration.wav -i music.mp3 -i sfx.wav \
-filter_complex "\
[1:a]volume=1.0,aformat=sample_fmts=fltp:sample_rates=48000:channel_layouts=stereo[narr];\
[2:a]volume=0.15,aformat=sample_fmts=fltp:sample_rates=48000:channel_layouts=stereo[music];\
[3:a]adelay=3000|3000,volume=0.7,aformat=sample_fmts=fltp:sample_rates=48000:channel_layouts=stereo[sfx];\
[narr][music][sfx]amix=inputs=3:duration=first:normalize=0[mixed];\
[mixed]loudnorm=I=-14:TP=-1:LRA=11[final]" \
-map 0:v -map "[final]" \
-c:v copy -c:a aac -b:a 192k \
output.mp4
Audio-Only Master Mix
ffmpeg -i narration.wav -i music.mp3 -i intro_sfx.wav -i outro_sfx.wav \
-filter_complex "\
[0:a]volume=1.0[narr];\
[1:a]volume=0.15[music];\
[2:a]adelay=0|0,volume=0.8[intro];\
[3:a]adelay=55000|55000,volume=0.8[outro];\
[narr][music][intro][outro]amix=inputs=4:duration=longest:weights='3 1 2 2'[mix];\
[mix]loudnorm=I=-14:TP=-1[final]" \
-map "[final]" -c:a aac -b:a 256k master_audio.m4a
Quick Reference: Common Patterns
Pattern 1: Narration + Background Music
ffmpeg -i narration.mp3 -i music.mp3 \
-filter_complex "[0:a]volume=1.0[n];[1:a]volume=0.15[m];[n][m]amix=inputs=2:duration=first" \
output.m4a
Pattern 2: Music with Auto-Duck
ffmpeg -i narration.mp3 -i music.mp3 \
-filter_complex "[0:a]asplit=2[n][sc];[1:a][sc]sidechaincompress=threshold=0.02:ratio=10[d];[n][d]amix=inputs=2" \
output.m4a
Pattern 3: Timed Intro Music Fade
ffmpeg -i narration.mp3 -i intro_music.mp3 \
-filter_complex "\
[1:a]afade=t=out:st=8:d=2,volume=0.3[intro];\
[0:a]adelay=10000|10000[narr];\
[intro][narr]amix=inputs=2:duration=longest" \
output.m4a
Pattern 4: Crossfade Between Segments
ffmpeg -i segment1.mp3 -i segment2.mp3 \
-filter_complex "\
[0:a]afade=t=out:st=28:d=2[s1];\
[1:a]adelay=28000|28000,afade=t=in:d=2[s2];\
[s1][s2]amix=inputs=2:duration=longest" \
output.m4a
Troubleshooting
| Issue | Cause | Solution |
|---|---|---|
| Clipping/distortion | Combined levels too high | Reduce individual volumes or add limiter |
| Narration buried | Music too loud | Reduce music to 10-15%, add ducking |
| Hollow/thin sound | Phase cancellation | Check mono compatibility |
| Pumping artifacts | Aggressive ducking | Increase attack/release times |
| Inconsistent levels | No normalization | Apply loudnorm filter |
Add Limiter to Prevent Clipping
ffmpeg -i input.mp3 \
-af "alimiter=level_in=1:level_out=0.9:limit=0.95:attack=5:release=50" \
output.m4a
Related Skills
video-pacing: Video rhythm and timing patternsremotion-composer: Programmatic video generationdemo-producer: Product demo video productionthumbnail-first-frame: Video thumbnail optimization
References
- ffmpeg Filters - Complete audio filter reference
- Volume Balancing - Detailed formulas and calculations
- Ducking Patterns - Automatic ducking implementation