Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Vorbis settings for SE listening test (Read 4761 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Vorbis settings for SE listening test

I want to add two recent vorbis encoders to SoundExpert - libVorbis v1.2.3 and  aoTuVb5.7. Most vorbis suggestions received from site visitors relate to low bit rates, so some corresponding setting have to be chosen for both of them. Need your advice, guys. Comments about versions and bit rate chosen are welcome as well. Thanks.
keeping audio clear together - soundexpert.org

Vorbis settings for SE listening test

Reply #1 –
-q 4 for ~128 kbps.
-q 3 for ~112 kbps.
etcetera...

Vorbis settings for SE listening test

Reply #2 –
I incline to (-q3) for both of them. For nine SE sound samples it gives:

  • libVorbis v1.2.3 (-q3) – 107.10 kbit/s
  • aoTuVb5.7 (-q3) – 109.55 kbit/s


Another variant is to adjust target bit rate with fractional q values. aoTuVb5.7 (-q2.99) gives 102.04 kbit/s, so the only option to equal them is to use fractional qualities for both. Something like this:

  • libVorbis v1.2.3 (-q2.99) – 98.35 kbit/s
  • aoTuVb5.7 (-q2.47) – 98.34 kbit/s


But I think the second variant is less preferable because those quality settings will hardly be used in practice and so less interesting for general public. Also huge gap in target bit rate between (-q3) and (-q2.99) probably means that integer values are specially tuned.

If there are no better ideas, I will add the first ones to SE 105 kbit/s group.
keeping audio clear together - soundexpert.org

Vorbis settings for SE listening test

Reply #3 –
IIRC libVorbis v1.2.3 is still based on aoTuV b2 (as libvorbis 1.1.0)

Vorbis settings for SE listening test

Reply #4 –
Quote from: lvqcl on
IIRC libVorbis v1.2.3 is still based on aoTuV b2 (as libvorbis 1.1.0)

May be. But it is widely used now (Theora, VLC, Helix).
Is it worth to compare it with Aoyumi's latest version?
Also I'm not sure about the bit rate. May be lower (-q) settings would be more interesting as they were tuned the most by Aoyumi.
keeping audio clear together - soundexpert.org
 

Vorbis settings for SE listening test

Reply #5 –
Finally short investigation put everything into place. Below are the output bit rates of both Vorbis encoders depending on -q parameter with 0.1 step. Nine SE samples are used.

Code: [Select]
-q   libVorbis  aoTuVb
	1.2.3	 5.7
------------------------
-2		31,14
-1,9		31,97
-1,8		32,93
-1,7		33,84
-1,6		34,85
-1,5		35,63
-1,4		36,64
-1,3		37,7
-1,2		38,86
-1,1		39,91
-1	48,02	42,73
-0,9	50,06	43,68
-0,8	51,26	44,64
-0,7	53,34	45,6
-0,6	55,48	46,52
-0,5	57,7	47,61
-0,4	58,81	48,61
-0,3	61,19	49,82
-0,2	63,63	50,93
-0,1	64,91	52,33
0	55,17	61,69
0,1	56,73	63,16
0,2	58,74	65,08
0,3	60,37	66,73
0,4	62,01	68,37
0,5	63,66	70
0,6	65,31	71,57
0,7	67,22	73,49
0,8	68,93	75,27
0,9	70,63	77,13
1	71,96	80,43
1,1	72,92	81,24
1,2	74,49	82,59
1,3	75,44	83,41
1,4	76,47	84,25
1,5	77,48	85,09
1,6	78,49	85,93
1,7	79,87	87,08
1,8	80,96	87,95
1,9	82,03	88,84
2	88,14	94,06
2,1	89,03	94,98
2,2	90,49	96,41
2,3	91,81	97,85
2,4	92,7	98,78
2,5	93,61	99,72
2,6	94,74	100,82
2,7	95,66	101,77
2,8	96,58	102,7
2,9	97,51	103,65
3	107,1	110,64
3,1	108,6	112,23
3,2	110,05	113,75
3,3	111,54	115,14
3,4	112,71	116,64
3,5	114,49	118,22
3,6	115,76	119,53
3,7	117,27	121,34
3,8	118,45	122,77
3,9	120,33	124,38
4	119,44	123,74
4,1	121,88	126,15
4,2	124,25	128,93
4,3	126,78	131,47
4,4	129,24	134,05
4,5	132	136,69
4,6	134,55	139,59
4,7	136,95	142,47
4,8	139,89	145,41
4,9	142,21	147,93
5	151,47	162,03
5,1	154,44	164,21
5,2	157,1	166,08
5,3	159,83	167,93
5,4	162,58	169,75
5,5	165,43	171,79
5,6	168,43	173,73
5,7	171,29	175,59
5,8	174,12	177,42
5,9	177,07	179,54
6	185,66	183,22
6,1	188,55	185,57
6,2	191,33	187,89
6,3	194,17	190,2
6,4	197,11	192,51
6,5	200	195,05
6,6	202,79	197,73
6,7	205,5	200,3
6,8	208,56	202,81
6,9	211,52	205,42
7	215,47	211,38
7,1	218,19	214,16
7,2	220,94	216,96
7,3	223,92	219,82
7,4	226,72	222,55
7,5	229,71	225,35
7,6	232,53	228,16
7,7	235,9	230,96
7,8	238,79	233,73
7,9	241,97	236,5
8	245,19	238,89
8,1	250,42	244,96
8,2	256,35	250,71
8,3	261,96	256,47
8,4	267,31	262,62
8,5	273	268,96
8,6	278,61	275,22
8,7	284,22	282,08
8,8	290,16	289,3
8,9	296,15	299,08
9	313,2	312,46
9,1	329,23	329,87
9,2	344,06	346,31
9,3	359,02	363,21
9,4	372,9	379,36
9,5	384,53	393,96
9,6	393,86	406,43
9,7	401,25	416,7
9,8	407,88	426,55
9,9	414,22	436,33
10	419,97	445,53
Graphical charts are available in SE news article.

There are several distinct monotonous segments. I think fair comparison of the encoders has to be done within the boundaries of corresponding segments. So the first pair has been added to SE rating system with the following settings:
  • aoTuVb5.7 (-q1) – 80.43 kbit/s
  • libVorbis v1.2.3 (-q1.75) – 80.41 kbit/s
Volunteers are welcome!
keeping audio clear together - soundexpert.org