I've always been curious about audio compression: at what bitrate does Opus become truly transparent? Can most people actually distinguish between 96kbps or 160kbps and lossless? Rather than relying on ego / reddit / forum debates and other subjective opinions, I decided to build a proper double-blind test.
Thanks to recent advances in AI coding, this was so simple to set up.
Double-blind audio comparison tool
The tool
I've created a web-based comparison tool that lets you listen to four versions of the same audio tracks:• FLAC (24 MB) – Lossless reference
• Opus 160kbps (5 MB) – Should be transparent
• Opus 96kbps (3 MB) – Good quality
• Opus 24kbps (0.8 MB) – Obviously compressed
The catch? You don't know which is which. The variants are randomly labelled A, B, C, and D. Your job is simple: listen, compare, and rank them by perceived quality.
Why Opus?
Opus is a modern, open-source codec that's incredibly efficient. It powers everything from Discord voice chat to YouTube audio streams. But there's ongoing debate about where it becomes perceptually lossless. This test aims to gather real data.
The tracks
The audio are a selection of my own compositions. I chose something I'd created myself to (a) avoid any copyright complications, (b) because I wanted material with varied instrumentation, dynamics, and frequency content to properly stress-test the codecs, and (c) I'd like more people to hear what I make!
How it works
• Click through the four variants (A, B, C, D) while the audio plays.• Drag them into order from best to worst quality
• Optionally share your age range, listening device, and confidence level
• See how you did!
Optionally share your age range, listening device, and confidence level
The test is completely anonymous, but you can leave your email if you'd like to receive the aggregated results once we reach statistical significance.
Early findings
I'm hoping to gather at least 100-200 responses before drawing any conclusions. The demographic data (especially age) will be fascinating – high-frequency hearing loss means younger listeners might perceive differences that older listeners can't detect anymore.
Try it yourself
Take the test! It doesn't take long, and hopefully it's enjoyable too... Use decent headphones or speakers if you have them, and listen in a quiet environment. I'd genuinely love to know: can you tell the difference?
Once the results are in, I'll publish a follow-up post with some of the findings and data visualised