In the interest of performing a transcode test, I need to see if the samples I'm going to choose artifact well. Unfortunately, there is some irony in me conducting a listening test, because I can't hear artifacts very well. So this thread is for Den, who said he'd help me here, and for others who have good hearing and expressed an interest in a transcode test.
The question here is, Does this sample artifact well when transcoded? Alternatively though, if you DO have a sample that produces transcoding artifacts, PLEASE upload it so that I can use it for the test.
The sample I've included in this post is from the Twister Soundtrack, but I can try and rip from the artist's original album if preferred. It uses female vocal solo, and a harpsicord solo, so it has SOME of the things that can rip up a codec.
The focus on this test will be various codecs -> Lame --preset medium (found on rarewares as a MODIFIED version of Lame).
Quintissential Metal: I don't think I could hear artifacts in this with Blade 112, but maybe you can.
This one is JUST over 30 seconds. I think it can still be considered fair use though, since I wanted to cut it off at a good point.
Just grabbed them. Will report back in a while.
OK, I took the Mastrery of Puppets sample, and encoded it to preset medium using the modified version of LAME 3.90.3 from Rarewares.
Then took the same, encoded to preset standard, and then preset medium.
ABX'd the first against the second, and there is additional stereo flanging in the distorted guitar within the first few seconds in the transcoded example. Quite obvious. ABX'd 10/10.
Interestingly, the preset standard -> preset medium also requires a few extra bits against the direct preset medium.
With repeated listens, there are some other artifacts too, wobbles in the decay of the cymbals in the first few seconds, mushy attacks, etc.
Give this a go yourself with the ABX foobar tool. I'm sure you will notice it, particularly the change in the stereo flanging in the first few seconds.
Haven't got to Tatula yet.
OK Talula. Did the same as above. Where the harpsichord section comes in, the direct encode to medium has slightly cleaner attacks than the transcoded example, but the difference is much harder to pick than the Mastery of Puppets sample. At first I was guessing, but once I noticed the slight muddiness in the harpsichord part, I could ABX it 10/10 times.
This is only with LAME standard to LAME medium. I'm sure there would be problems with other codecs also, although perhaps to a lesser degree.
Hope this helps.
Right, my last post on these two samples, I promise...
Just repeated the above but went via Musepack, ie sample->LAME preset medium vs sample->-mppenc (1.15r) normal->LAME preset medium.
Much harder to pick this time. With Puppets, same distorted section ABX 7/10, so probably not significant. Talula, couldn't pick it in the harpsichord section, 5/10. It may be possible to improve these scores with some very attentive listening, but already I can tell that these transcodes are much better than the previous ones via LAME --preset standard.
So just on these two samples, mpc->LAME medium appears to be less risky than LAME standard to medium.
This suggests to me that conducting this test using samples from your collection across a range of genres would work, at least for me anyway.
Perhaps someone else out there can check these too, but I'm sure I don't have "golden ears".
Thanks, Den. Anyone else wanting to comment on how well these will transcode, speak up.
I own absolutely no jazz CDs, but for those interested in the transcode test, I was wondering if this particular sample is 'jazz enough' for the purposes of this test.
I found an evil country sample (Sorry to all country fans).
This sample represents the worst of country, IMHO.
Here are a couple of electronic samples, 2 different cuts of KraftWek Autobahn. They feature some unusual vocal treatments, and one has a HF sweep.
See if you thin they might be suitable. If not, I can always post one of the more "musical" sections".
And here's the other one.