# VOD Deep Dive Part 3: Audio Fundamentals — Making Sound Small

> How digital audio works: sampling rates, bit depth, channels, AAC vs Opus vs Dolby Atmos, multi-language tracks, loudness normalization, and practical ffmpeg recipes.

- Author: zhuermu
- Published: 2026-05-10
- Web version: https://zhuermu.com/en/blog/vod-deep-dive-part-3-audio-fundamentals/

---
> *This is Part 3 of the [VOD Streaming Deep Dive](/blog/vod-deep-dive-part-1-video-fundamentals/) series.*

---

## How Sound Becomes Digital

Sound is air vibration — a continuous waveform. Computers can only store numbers, not continuous waves. Two steps are needed:

1. **Sampling**: Measure the wave's height at regular intervals
2. **Quantization**: Convert each measurement into a number

```
Amplitude
 ▲
 │  ●   ●                     ●    Sample points
 │     ●  ●  ●
 │           ●  ●
 │              ●  ●   ●
 └──────────────────────► Time
     ↑ ↑ ↑ Sample N times per second — N is the "sample rate"
```

---

## Sample Rate

**Unit**: Hz (hertz — samples per second)

| Sample rate | Use case |
|------------|----------|
| 8 kHz | Telephone voice |
| 16 kHz | Speech recognition, VoIP (Zoom/Teams) |
| 22.05 kHz | Retro games, AM radio |
| **44.1 kHz** | **CD audio, music preferred** |
| **48 kHz** | **Video industry default** (film, streaming, broadcast) |
| 96 kHz | Hi-fi recording |
| 192 kHz | Professional studio |

**Nyquist theorem**: To reproduce a frequency F, you need a sample rate of at least 2F. Human hearing tops out around 20 kHz, so 44.1/48 kHz is just enough (with a small margin).

For VOD, standardize on **48 kHz**. If your source is 44.1 kHz, resample during transcoding with `-ar 48000`.

---

## Bit Depth

How many bits per sample:

| Bit depth | Loudness levels | Use case |
|-----------|----------------|----------|
| 8-bit | 256 | Retro games, telephony |
| **16-bit** | **65,536** | **CD, consumer streaming** |
| 24-bit | ~16.7M | Professional recording |
| 32-bit float | Astronomical | Audio production internal format |

Most video audio is 16-bit, 48 kHz.

---

## Channels

A **channel** is an independent audio track:

| Channels | Name | Configuration | Used in |
|----------|------|---------------|---------|
| 1.0 | Mono | Single channel | Telephony, old TV |
| 2.0 | Stereo | Left + Right | **Music, most video** |
| 5.1 | Surround | Front L + Center + Front R + Rear L + Rear R + **LFE** (.1 = subwoofer) | Cinema, home theater |
| 7.1 | Surround | 5.1 + two side channels | Premium home theater |
| 7.1.4 | Atmos etc. | 7.1 + **4 overhead channels** | Dolby Atmos |

```
5.1 surround layout (top-down view):

         FL ──── C ──── FR
              │  🧑  │
              │      │
         SL ──┻━━━━──SR
                LFE
```

---

## Audio Bitrate: How Many kbps Is Enough?

Audio bitrate is also bits per second, but much smaller than video — typically **5–10% of the video bitrate**.

| Bitrate | Perception | Typical use |
|---------|-----------|-------------|
| 32 kbps | Voice OK, music broken | Extreme low bandwidth |
| 64 kbps | Voice clear, music passable | Low-bitrate scenarios |
| 96 kbps | Music acceptable | Broadcast, YouTube default |
| **128 kbps** | **Music sounds good** | **Streaming default** |
| 192 kbps | High fidelity | Premium music streaming |
| 256 kbps | Audiophile-grade | Apple Music |
| 320 kbps | MP3 maximum | Music enthusiasts |
| Lossless (FLAC) | Transparent | Hi-fi niche |

For VOD: **stereo AAC at 128 kbps** is the correct answer for the vast majority of scenarios.

---

## Major Audio Codecs

### AAC (Advanced Audio Coding) — The Streaming Default

- **By**: MPEG (same organization behind H.264)
- **Year**: 1997
- **Compatibility**: **every** video platform, browser, and phone
- **Variants**:
  - **AAC-LC** (Low Complexity): Most common. HLS/DASH default.
  - **HE-AAC** (High Efficiency): Better at low bitrates (<64 kbps)
  - **HE-AAC v2**: HE-AAC + parametric stereo, decent at 48 kbps

### MP3 — Retired

Classic but less efficient than AAC. Original patents expired in 2017. No reason to use MP3 in new projects.

### Opus — The Web Newcomer

- **Open-source, royalty-free**
- Excellent from 6 kbps (voice) to 510 kbps (music)
- **WebRTC default**, used by Discord
- But HLS/DASH compatibility lags behind AAC; limited iOS/Safari support

### Dolby Family — Cinema-Grade

| Codec | Use case |
|-------|----------|
| AC-3 (Dolby Digital) | 5.1 surround, Blu-ray, legacy HDTV |
| E-AC-3 / DD+ (Dolby Digital Plus) | 5.1/7.1, streaming movies |
| **Dolby Atmos** (E-AC-3 + JOC or AC-4) | **Spatial audio**, premium platforms |

Dolby Atmos on Netflix, Disney+, and Apple TV+ is a hallmark of premium subscriptions.

### FLAC / ALAC — Lossless

Lossless compression reduces size by 50–70% while perfectly preserving the original PCM data. Used in Apple Music lossless tier and audiophile contexts. Not practical for video streaming — bitrate is too high.

---

## Multi-Language Audio Tracks

A single video file can carry **multiple audio tracks**:

```
MP4 file
├── video track   (H.264)
├── audio track 1 (AAC, English)
├── audio track 2 (AAC, Chinese)
├── audio track 3 (AAC, Japanese)
└── subtitle track (WebVTT)
```

Streaming protocols (HLS/DASH) support **independent audio track delivery** — the player only downloads the language the user selected:

```m3u8
#EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="audio",LANGUAGE="en",NAME="English",DEFAULT=YES,URI="audio/en/index.m3u8"
#EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="audio",LANGUAGE="zh",NAME="中文",URI="audio/zh/index.m3u8"
```

More on this in [Part 5: Streaming Protocols](/blog/vod-deep-dive-part-5-streaming-protocols-hls-dash/).

---

## Loudness Normalization

Ever noticed the volume spike when a commercial cuts in? That's because different content has wildly different loudness levels.

**Loudness normalization** adjusts all content to a uniform *perceived* loudness level (not peak volume).

### Common Standards

| Standard | Target loudness | Used by |
|----------|----------------|---------|
| **EBU R128** | -23 **LUFS** | European broadcast |
| **ATSC A/85** | -24 LUFS | North American broadcast |
| **Apple Music / Spotify** | -14 LUFS | Music streaming |
| **YouTube** | -14 LUFS | Default |
| **Short-form / mobile** | -16 to -14 LUFS | Phone speaker range |

**LUFS** (Loudness Units Full Scale) is the international standard for perceived loudness.

### ffmpeg Loudness Normalization

```bash
# Normalize audio to -14 LUFS
ffmpeg -i input.mp4 -af loudnorm=I=-14:TP=-1.5:LRA=11 -c:v copy output.mp4
```

---

## Hands-On: Inspect and Transcode Audio

### Check audio tracks in a video

```bash
ffprobe -v error -show_streams -select_streams a input.mp4
```

Typical output:

```
codec_name=aac
sample_rate=48000
channels=2
channel_layout=stereo
bit_rate=128000
```

### Standardize to AAC 48 kHz 128 kbps stereo

```bash
ffmpeg -i input.mov \
  -c:a aac -b:a 128k -ar 48000 -ac 2 \
  -c:v copy \
  output.mp4
```

- `-c:a aac`: Audio codec AAC
- `-b:a 128k`: 128 kbps bitrate
- `-ar 48000`: 48 kHz sample rate
- `-ac 2`: 2 channels (stereo)
- `-c:v copy`: Copy video stream as-is (saves time)

---

## Key Takeaways

1. Digital audio requires **sampling rate** (temporal density) and **bit depth** (amplitude precision).
2. VOD default: **48 kHz** sample rate, **16-bit** depth.
3. Consumer streaming defaults to **stereo (2.0)**; cinema uses **5.1 / Atmos**.
4. **AAC-LC at 128 kbps** is the default audio setting for VOD.
5. A single video file can carry multiple audio tracks (multi-language).
6. **Loudness normalization** (EBU R128 / -14 LUFS) prevents the "ads are too loud" problem.

---

*Previous:* [Part 2: Video Codecs](/blog/vod-deep-dive-part-2-video-codecs-h264-h265-av1/)

*Next:* **[Part 4: Container Formats — MP4, fMP4, and CMAF](/blog/vod-deep-dive-part-4-container-formats-mp4-fmp4-cmaf/)**

---

## Frequently asked

### What audio bitrate should I use for streaming video?

For the vast majority of VOD scenarios, stereo AAC at 128 kbps is the right answer — music sounds good at that rate and it's the streaming default. Audio typically needs only 5–10% of the video bitrate. Go lower (64–96 kbps) only for bandwidth-constrained cases, and higher (192–256 kbps) mainly for premium music streaming rather than video.

### What is the difference between 44.1 kHz and 48 kHz sample rates?

44.1 kHz is the CD audio standard and the preferred rate for music, while 48 kHz is the default across the video industry — film, streaming, and broadcast. Both satisfy the Nyquist theorem, which requires sampling at least twice the highest frequency you want to reproduce (human hearing tops out around 20 kHz). For VOD, standardize on 48 kHz and resample 44.1 kHz sources during transcoding.

### What is loudness normalization and why does it matter?

Loudness normalization adjusts all content to a uniform perceived loudness level, measured in LUFS, rather than peak volume. It fixes the classic problem of commercials sounding much louder than the show they interrupt. Common targets: -23 LUFS for European broadcast (EBU R128), -14 LUFS for YouTube and music streaming. In ffmpeg you can apply it with the loudnorm audio filter.


---

## References

- [RFC 6716: Definition of the Opus Audio Codec](https://datatracker.ietf.org/doc/html/rfc6716) — IETF
- [FFmpeg documentation](https://ffmpeg.org/documentation.html) — FFmpeg