This commit is contained in:
2025-06-17 18:30:25 +01:00
parent 1c89b29844
commit a308ef6258
20 changed files with 422 additions and 0 deletions

70
docs/slides/60-audio.md Normal file
View File

@@ -0,0 +1,70 @@
Audio
### Clean Audio = Believable Video
> Viewers forgive shaky images far sooner than noisy sound.
Note:
Start by playing two identical clips—one with background hiss, one cleaned—and ask which looks “more professional.”
--
Audio
#### 1. Capture at the Source
* Use the best mic available; position 1530cm from mouth.
* Monitor with headphones; peaks should stay below 6dB.
* Record at least 10s **roomtone** in every location.
Note:
Emphasise that good capture saves hours in post; show how redlined peaks distort irreversibly.
--
Audio
#### 2. RoomTone & Noise Control (Audacity)
1. Import clip.
2. Select clean roomtone → **Effect ▸ Noise Reduction ▸ Get Profile**.
3. Select entire track → repeat Noise Reduction (612dB reduction, Sensitivity 6, Frequency Smoothing 3).
4. Apply **HighPass Filter 80Hz** to remove rumble.
Note:
Warn that overreduction causes “swirly” artefacts—dial Sensitivity down if it sounds robotic.
--
Audio
#### 3. Equalize & Compress
* Roll off <80Hz; dip 4kHz to tame harshness.
* Compressor: Threshold 6dB, Ratio 3:1, Attack 10ms, Release 150ms.
* Limit peaks to 1dB; aim for dialogue ≈‑12LUFS.
Note:
Show LUFS meter in Kdenlive (View Audio Meter) and explain web loudness target (14LUFS) versus broadcast (23LUFS).
--
Audio
#### 4. Sync & Layering
* Align tracks via the **handclap spike** in waveforms.
* Keep voiceoff on dedicated top audio track; ambience/FX below.
* Add 3frame fades at every cut to kill pops.
Note:
Demo Kdenlives *Align Audio to Reference* for quick waveform syncthe same tool used in multicam earlier.
--
Audio
#### 5. Export Best Quality
* Export **WAV 48kHz 24bit** from Audacity.
* In Kdenlive Render: AAC 48kHz, 192kb/s (YouTube preset is fine).
* Watch meters during render; ensure no clipping.
Note:
Clarify that archive masters stay lossless; delivery formats can be lossy.