|
Post by Kron on May 10, 2014 16:44:43 GMT
I tried l3enc with Wine, didn't work, but it did suggest using DOSBox. I'm trying it right now and seems to work. (Rather slow, as in single digit frames per second) but works none the less.
PS: I am but a mere audio/codec amateur, your typical armchair expert on the subject. Should I not be adding anything of any use, I will gladly keep from spamming.
|
|
|
Post by kamedo2 on May 11, 2014 5:06:33 GMT
Kron, You are very welcome. Please stay here, post even more useful info. I did a primitive non-blind listening test, using l3enc and lossless music samples from here, and encoded Bladeenc samples. listening-tests.hydrogenaudio.org/sebastian/mp3-128-1/results.htmI prefer Bladeenc 96kbps as a low anchor, because no music tracks are of >2.5 score. It means people will have no trouble distinguishing from the original. If we use l3enc 128kbps, samples like V Vangelis_Chariots_of_fire and I linchpin should be hard for beginners to spot the difference. Vangelis_Chariots_of_fire scored 2.59 and linchpin scored 3.01 in the 2008 MP3 128kbps listening test.
|
|
|
Post by Kron on May 11, 2014 11:54:22 GMT
I tried l3enc v2.71 on a light test (Male voice, mono, with light stereo orchestral background) l3enc did surprisingly well (don't have a good headset for spotting artefacts). BladeEnc faltered badly at 96k. I don't know how BladeEnc fares against FAAC in regards low vs. midlow anchor. I currently don't have the time for more extensive tests.
I was wondering, when the test officially starts, will a casting call be put on hydrogen audio? You're going to need as much testers as you can get, especially since most codecs are fairly good at 96kbps already.
|
|
|
Post by kamedo2 on May 11, 2014 13:21:18 GMT
I tried l3enc v2.71 on a light test (Male voice, mono, with light stereo orchestral background) l3enc did surprisingly well (don't have a good headset for spotting artefacts). BladeEnc faltered badly at 96k. I don't know how BladeEnc fares against FAAC in regards low vs. midlow anchor. I currently don't have the time for more extensive tests. I was wondering, when the test officially starts, will a casting call be put on hydrogen audio? You're going to need as much testers as you can get, especially since most codecs are fairly good at 96kbps already. Thank you for testing. I know the 2008 public listening test of MP3 128kbps used l3enc, but using BladeEnc as a low anchor seems to be more wise. I'm thinking of a casting call on both HA and slashdot when the test officially starts. I don't have much experience in publicity and advertisement jobs, so any offer is very helpful.
|
|
IgorC
Administrator
Posts: 41
|
Post by IgorC on May 11, 2014 18:45:23 GMT
While the lowest anchor will be significantly worst than mid-low anchor it can be any (old) lossy codec (MP3, AAC ...). It will be appropriate (actually better) if it will have lower bandwith as well: mid-low anchor - 10 kHz, lowest anchor - somewhere at ~5 kHz. The same technique of two low anchors (with different bandwith) is/was used during standarization of codecs (AAC, etc...). It helps a lot to process submitted results from listeners. Related info Should HA promote a more rigorous listening test protocol?
|
|
|
Post by kamedo2 on May 17, 2014 19:24:10 GMT
I tried these low-anchors. cutoff commandline 10kHz ffmpeg_r62950 -y -i original.wav -acodec libfaac -b:a 96k out.libfaac.b96.mp4
8kHz ffmpeg_r62950 -y -i original.wav -acodec libfaac -q:a 30 out.q30.mp4 6kHz ffmpeg_r62950 -y -i original.wav -acodec mp2 -b:a 96k -ac 2 out.mp2_96k.mp2 7kHz ffmpeg_r62950 -y -i original.wav -acodec libvo_aacenc -b:a 32k -ac 2 -ar 22050 out.vo32k_22k.mp4 16kHz bladeenc -quit -nocfg original.wav out.be96k.mp3 -96
|
|
|
Post by kamedo2 on May 23, 2014 13:23:39 GMT
I'm considering a preliminary blind listening test of low anchors: - FAAC 96kbps
- FAAC VBR q30 (48kbps)
- FFmpeg's native mp2 96kbps
- vo-aacenc 96kbps
- BladeEnc 96kbps
- FFmpeg's native aac 96kbps
- FFmpeg's native aac, v8g patch 96kbps
I don't want low anchors to be better than the FAAC 96kbps. This test will reveal which low anchor is most suitable.
|
|
|
Post by kamedo2 on May 28, 2014 17:50:09 GMT
I'm about to start a preliminary listening test of: - FAAC abr 96k.mp4
- FAAC vbr q30(~48kbps).mp4
- FFmpeg native mp2 encoder 96k.mp2
- vo-aacenc 0.1.3 abr 96k.mp4
- Bladeenc 96k.mp3
- FFmpeg native AAC encoder 96k.mp4
- FFmpeg native AAC encoder+v8g patch 96k.mp4
using first 15 samples of a 2011 public multiformat listening test (30 samples). I'll make sure that the lowest anchor is worse than the mid-low anchor.
|
|
|
Post by kamedo2 on Jun 4, 2014 13:19:45 GMT
Because of my simple mistake, I'm testing these 6 low anchors with a duplicate: - FAAC abr 96k.mp4
- FAAC vbr q30(~48kbps).mp4
- FAAC vbr q30(~48kbps).mp4
- vo-aacenc 0.1.3 abr 96k.mp4
- Bladeenc 96k.mp3
- FFmpeg native AAC encoder 96k.mp4
- FFmpeg native AAC encoder+v8g patch 96k.mp4
87% done. Expect to see the results soon.
|
|
|
Post by kamedo2 on Jun 5, 2014 19:26:56 GMT
|
|
|
Post by kamedo2 on Jun 8, 2014 19:49:21 GMT
Now it's clear which anchor we should use. The bitrate distribution of audio files encoded by 6 encoders using 40 samples. I've checked encoder parameters and conditions. Now we should prepare packages. AAC Opus Ogg MP3 V5 FAAC 96k FAAC q30 103.143 82.088 96.568 143.804 97.622 58.074 SinceAlways 103.5 118.89 113.215 154.478 98.781 48.067 9-Have-big-expensive-car.441 109.635 106.495 112.413 142.579 97.705 51.759 Hotel California 100.345 120.148 93.984 131.324 97.702 53.808 Severance__1.31-1.51_ 116.17 97.405 143.165 160.293 98.472 54.33 liberate 99.463 116.257 110.799 143.561 97.399 53.828 Can't Wait Until Tonight (Dry Wurlitzer Mix) 98.704 98.46 95.301 125.235 98.082 53.42 25-This-is-the-end.441 103.384 103.47 98.657 144.106 97.982 56.467 With Love (Outro) 95.015 95.686 99.479 121.091 98.309 45.635 4-Sound-English-male.441 110.859 120.63 108.39 125.575 97.808 53.072 Girl_In_The_Fire__Sample_ 101.955 92.133 96.128 130.513 97.885 55.463 Hey Tonight 100.594 110.81 85.33 140.947 97.979 46.883 ExitMusic 100.865 122.456 134.219 154.968 97.912 51.001 girl 100.757 132.751 121.369 175.761 99.208 45.142 24-Greensleeves-Korean-male-speech.441 99.38 96.783 116.613 134.87 97.616 58.991 Through The Fire And Flames 99.146 123.935 95.72 126.82 97.684 54.88 TrosYGareg 105.863 112.681 118.512 137.172 97.661 50.599 Changes 99.813 103.17 89.394 120.056 97.665 54.069 Greatest_Love_of_All_2min57 106.845 104.16 102.39 134.828 98.283 48.479 Coral 100.98 132.906 100.614 156.076 97.633 60.029 trumpet 104.905 111.689 88.972 132.966 98.467 51.608 21-classic.441 110.112 95.969 100.417 129.988 97.833 48.02 Only Time 109.066 104.388 122.045 145.766 97.671 49.171 The Chastising of Renegade 108.954 114.173 132.913 137.706 97.779 60.169 take_your_finger_frin_my_head 109.151 102.793 116.081 147.589 97.679 47.953 Asleep__4.11-4.31_ 106.968 103.74 122.072 145.315 98.326 52.618 clapton_44k 114.349 112.294 123.584 168.79 97.845 54.738 velvet 99.239 93.86 100.284 129.374 98.233 52.08 15-Good-evening.441 109.76 103.996 113.745 146.506 97.958 47.525 Last_Of_The_Mohicanz__Sample_ 100.114 90.019 106.919 131.076 97.834 56.076 SlavesOfFear 99.05 96.917 105.963 134.395 97.649 49.015 sandman 108.411 100.371 104.54 127.652 97.974 43.582 Robots_old 106.178 101.153 93.912 119.552 97.559 53.366 bonhemian_rhapsody 105.935 110.637 106.679 137.144 97.935 53.844 Waiting 95.082 99.544 103.59 115.804 98.051 46.634 12-German-male-speech.441 102.801 104.706 89.438 122.5 97.61 56.364 Jupiter, the Bringer of Jolity 108.694 104.42 87.617 107.999 98.026 47.998 Shinsho_pool_3min45_4min4 102.829 92.584 101.581 130.981 97.747 55.44 NewYorkCity 85.32 130.625 79.441 77.263 83.341 41.275 35_SQAM_glockenspiel_cut 104.637 103.638 112.205 127.758 98.995 48.038 27-last-song-drums-and-trampets.441 %samples SinceAlways %samples 9-Have-big-expensive-car.441 %samples Hotel California %samples Severance__1.31-1.51_ %samples liberate %samples Can't Wait Until Tonight (Dry Wurlitzer Mix) %samples 25-This-is-the-end.441 %samples With Love (Outro) %samples 4-Sound-English-male.441 %samples Girl_In_The_Fire__Sample_ %samples Hey Tonight %samples ExitMusic %samples girl %samples 24-Greensleeves-Korean-male-speech.441 %samples Through The Fire And Flames %samples TrosYGareg %samples Changes %samples Greatest_Love_of_All_2min57 %samples Coral %samples trumpet %samples 21-classic.441 %samples Only Time %samples The Chastising of Renegade %samples take_your_finger_frin_my_head %samples Asleep__4.11-4.31_ %samples clapton_44k %samples velvet %samples 15-Good-evening.441 %samples Last_Of_The_Mohicanz__Sample_ %samples SlavesOfFear %samples sandman %samples Robots_old %samples bonhemian_rhapsody %samples Waiting %samples 12-German-male-speech.441 %samples Jupiter, the Bringer of Jolity %samples Shinsho_pool_3min45_4min4 %samples NewYorkCity %samples 35_SQAM_glockenspiel_cut %samples 27-last-song-drums-and-trampets.441
|
|
|
Post by win32vb on Jun 9, 2014 3:27:38 GMT
It looks to me like MP3 is using an unfairly high overall bitrate (136 Kb/s). I believe you could use non-integer quality numbers on the LAME MP3 encoder to bring the overall bitrate closer to 96 Kb/s. For example, "V 5.5".
|
|
|
Post by leonza on Jun 9, 2014 12:34:06 GMT
It looks to me like MP3 is using an unfairly high overall bitrate (136 Kb/s). I believe you could use non-integer quality numbers on the LAME MP3 encoder to bring the overall bitrate closer to 96 Kb/s. For example, "V 5.5". I agree
|
|
|
Post by kamedo2 on Jun 9, 2014 15:39:05 GMT
|
|
IgorC
Administrator
Posts: 41
|
Post by IgorC on Jun 9, 2014 15:57:16 GMT
|
|