Posting this here in case anyone wants to test it.
Click to view attachment
Click to view attachment
Full Version: modified version of ABC/HR-java for Linux
Oh dear... It seems I broke everything except the fastswitching in ABX mode. 
Here is the list of things I broke, now fixed:
Looping
Fastswitching toggle (stuck on in ABX, stuck off in main ABC/HR mode)
Automatic gain setting
Also I fixed that click at the start of playback, and made the range bar adjustable without stopping playback.
Schnofler, if you're interested in integrating any of these changes, it still needs some work on my part to clean it up, ie deleting all my commented out System.out.println's etc. - I'll do that if you ask
One other thing I don't understand is the randomization of samples - if they are randomized, you can't tell which sample you're comparing to the original until the results are saved, which for me was very confusing. On my first test I was confused because I thought I couldn't ABX mpc at -q 1 (~40kbs) but then I found out it was a random sample at -q 6 (200kbs). (where's the emoticon for PHEW!) At least there should be some visual indication that randomization is on. Just in this version I switched it off for ABX and left it on for ABC/HR mode.
The way I'm testing is to encode a song at about 20 bitrates, then approximately do a binary search to find the lowest bitrate at which I fail the ABX tests - would that be called a transparency threshold? This gets the minimum size while still being transparent (to me).
Laurence
Here is the list of things I broke, now fixed:
Looping
Fastswitching toggle (stuck on in ABX, stuck off in main ABC/HR mode)
Automatic gain setting
Also I fixed that click at the start of playback, and made the range bar adjustable without stopping playback.
Schnofler, if you're interested in integrating any of these changes, it still needs some work on my part to clean it up, ie deleting all my commented out System.out.println's etc. - I'll do that if you ask
One other thing I don't understand is the randomization of samples - if they are randomized, you can't tell which sample you're comparing to the original until the results are saved, which for me was very confusing. On my first test I was confused because I thought I couldn't ABX mpc at -q 1 (~40kbs) but then I found out it was a random sample at -q 6 (200kbs). (where's the emoticon for PHEW!) At least there should be some visual indication that randomization is on. Just in this version I switched it off for ABX and left it on for ABC/HR mode.
The way I'm testing is to encode a song at about 20 bitrates, then approximately do a binary search to find the lowest bitrate at which I fail the ABX tests - would that be called a transparency threshold? This gets the minimum size while still being transparent (to me).
Laurence
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.