PER v1.6.15 – PER is getting better and better

PER 1.6.15 is here

PER is a blind listening test program for Mac and Windows.

Its job is simple: help you make audio decisions without knowing which version you are listening to.

PER started in a slightly silly way. I had a discussion with a client, who is also a friend, about how he was comparing different masters and mixes. I got the feeling that he was not really giving each version a fair chance, and that the decision was becoming more emotional than useful.

So I made the first version of PER partly as a joke.

Then I tried it myself.

And that was the annoying part, because it actually worked.

I have compared masters and mixes inside REAPER for years, with loudness matching and careful switching. But even then, I still know what I am listening to. I can see the tracks. I can see the filenames. I know which version went through a certain plugin, a certain piece of hardware, or a certain processing chain.

And once you know that, you are already biased.

You may want the expensive compressor version to win. You may want the Ozone version to lose. You may want the clever version to sound better than the boring one. But ears are not always as independent as we want them to be.

PER tries to remove some of that.

You load a few audio files, for example two or three different masters of the same song, or a few different mixes you want to compare. PER then analyzes the files, normalizes them to a LUFS value chosen by you, hides the file identities, and lets you listen without knowing which file is playing.

You only get the Listen buttons.

No filenames.
No “this one is louder so it must be better.”
No staring at the session and trying to convince yourself you already know the answer.

After each listen, you rate the version from 1 to 6. When you have completed the number of ratings per file that you selected, PER shows you a ranking with the average score for each file.

The original LUFS-I value is also shown in the result. Not because it affects the blind test, it does not, but because it can be useful afterwards if you want to see how loud the files were before PER normalized them.

The default number of ratings per file is 3. That is intentionally low because people are often in a hurry. But if you want a more useful result, I would recommend something closer to 16–20 ratings per file. More repetitions usually give you a clearer picture of what you actually prefer, instead of what you think you prefer.

PER can be used for:

• Comparing different masters of the same song

• Comparing different mixes

• Letting a band vote on which version they prefer

• Checking whether a louder master is actually better, or just louder

• Testing processing chains without knowing which one is which

• Taking some of the expectation bias out of audio decisions

When the test is finished, you can listen to the ranked files again, export the result, redo the test, continue the test with more ratings, or start a new listening test.

You can also remove the normalization after the test is finished if you want to hear the original level differences.

PER also has a project export function. This lets you create a shared listening project so someone else can do the same test with the same files. For example, a band could compare two or three mixes independently and then compare the results afterward.

Important note about exported projects:
When you export a PER project, the audio files are included so the other person can listen to the same material. That also means you are sending the audio files to them. A technically curious person could find and use those files, so only export and share projects if you have the right to share the audio, and only send them to people you trust.

You can also choose whether all files should be converted to the same sample rate, or whether the original files should be kept as they are. So if one file is 88.2 kHz and another is 96 kHz, that is fine too.

What is new in PER 1.6.15?

This update adds a few workflow improvements.

There is now a button that shows which files are included in the current listening round, without revealing which hidden version is which. So if you load a lot of files, or if you export a project for someone else, you can still check what material is actually included in the test without breaking the blind part of the listening process.

The version number is now shown directly in the interface, so it is easier to check that you are running the latest version.

The old 8-file limit has also been removed. You can now load as many files as your computer can handle.

PER can now normalize ratings for a fairer result. This means the result is less affected by whether you used a narrow rating range in one round, like 1 and 2, and a wider range in another round, like 1 and 6. The important thing is which version you preferred in that round, not how dramatic your number choice happened to be.

Rating normalization can be turned off in Settings if you prefer to use the raw scores instead.

There is also a new New button that clears the current test and lets you start over from scratch.

PER is free, but it requires that you follow this Patreon campaign. You do not need to be a paying member to use it.

When you start PER, you enter your email address in the application. PER then checks whether that email is connected to a Patreon account following Noisebud. If you stop following the campaign, PER will eventually stop working.

Formats:
Mac
Windows

Download:
Files are attached directly to this post.

---

Original Patreon post:
https://www.patreon.com/posts/160240913

This link is included so the post can be shared on social media and via RSS while always pointing back to the original Patreon post. If you’re already reading this on Patreon, the link will simply take you back here.