Audio Examples

Table of Contents

  1. Why This Layout Works
  2. How to Listen
  3. Track Categories
  4. Listening Families By Track
  5. Notes

Qualitative Results

Listen to representative codec comparisons the way qualitative examples are usually presented in papers

This page is organized around the challenge structure first: Track 1 and Track 2 are separated into distinct sections and asset paths, then subdivided into listening families, bitrate groups, and numbered examples that make it easy to reference the same case across reviews and discussions.

Challenge tracks 2
Listening families 6
Representative cases 35
Playable clips 290

Why This Layout Works

This structure is close to the usual qualitative-results pattern from papers and challenge writeups:

  • examples are grouped by evaluation condition rather than by system
  • each case keeps the speech content fixed while the systems vary
  • consistent numbering makes it easy to point multiple listeners to the same example
  • role-based color coding keeps the comparison set easy to scan

Jump to top

How to Listen

Start With The Reference

Listen to the clean or noisy reference first when it is available so the rest of the row has a stable target.

Compare Within One Case

Stay inside a single case before moving on. That keeps the content fixed while artifacts, intelligibility, and naturalness change.

Scan Across Cases

After you learn a system's character, move across the same family to see whether the pattern repeats.

  • Use headphones when possible.
  • Audio players use native browser controls and preload="none" to avoid downloading every file at page load.

Jump to top

Track Categories

Track 1 Transparency Codecs

Coding-only evaluations focused on clean-speech quality, mild real-world distortions, and preservation of overlapping speakers.

Track 2 Speech Enhancement Codecs

Enhancement-capable evaluations focused on clean-speech transparency and intelligibility under enhancement-oriented conditions.

Jump to top

Listening Families By Track

Track 1

Transparency Codecs

Coding-only evaluations focused on clean-speech quality, mild real-world distortions, and preservation of overlapping speakers.

Clean Speech

Single-speaker clean-speech listening examples, grouped by bitrate.

6 kbps Aligned comparison set for this bitrate condition

Example 1

Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Anchor

Example 2

Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Anchor
1 kbps Aligned comparison set for this bitrate condition

Example 1

Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Anchor

Example 2

Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Anchor

Example 3

Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Anchor

Real-World Conditions

Speech examples recorded in realistic acoustic conditions, grouped by bitrate.

1 kbps Aligned comparison set for this bitrate condition

Example 1

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 2

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 3

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 4

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 5

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 6

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 7

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
6 kbps Aligned comparison set for this bitrate condition

Example 1

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 2

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 3

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 4

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 5

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 6

Noisy Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Word Intelligibility

Diagnostic word-pair examples focused on intelligibility at 1 kbps.

1 kbps word pairs Aligned comparison set for this bitrate condition

Example 1

Target: net Alternative: met
Clean Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 2

Target: neck Alternative: deck
Clean Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 3

Target: got Alternative: dot
Clean Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 4

Target: shoes Alternative: choose
Clean Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 5

Target: dote Alternative: note
Clean Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Overlapping Speakers

Examples containing overlapping speakers, grouped by bitrate.

6 kbps Aligned comparison set for this bitrate condition

Example 1

Reference
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Additional Comparison
1 kbps Aligned comparison set for this bitrate condition

Example 1

Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Jump to top
Track 2

Speech Enhancement Codecs

Enhancement-capable evaluations focused on clean-speech transparency and intelligibility under enhancement-oriented conditions.

Clean Speech

Single-speaker clean-speech listening examples, grouped by bitrate.

6 kbps Aligned comparison set for this bitrate condition

Example 1

Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Anchor

Example 2

Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Anchor
1 kbps Aligned comparison set for this bitrate condition

Example 1

Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Anchor

Example 2

Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Anchor

Example 3

Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Anchor

Word Intelligibility

Diagnostic word-pair examples focused on intelligibility at 1 kbps.

1 kbps word pairs Aligned comparison set for this bitrate condition

Example 1

Target: net Alternative: met
Clean Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 2

Target: neck Alternative: deck
Clean Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 3

Target: got Alternative: dot
Clean Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 4

Target: shoes Alternative: choose
Clean Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ

Example 5

Target: dote Alternative: note
Clean Reference
Baseline
aitd-go
Boya Audio
NanoCodec
NJU-AA Lab
Pdura7
TeamWZQAQ
Jump to top

Notes

These examples are intended to complement the written challenge materials by making qualitative behavior easy to hear quickly, not by replacing the formal metrics or the written system reports.

Jump to top