Killgore
6th June 2004, 14:10
Hi!
I have these two versions of Lame: 3.95.1 downloaded from doom9.org and 3.96 from http://mitiok.cjb.net/ . Additionally it appears that lame_enc.dll in doom9 package is 3.96 alpha version. I?ve noticed really big differences in final bitrate during AC3 to MP3 encoding (VBR V6 - I haven?t tried other VBR levels). These differences don?t occur when I encode CD-Audio music (actually they are very small, like 1 kbps).
I made a quick comparison below using Wonder Boys audio track. For dlls I?ve used Besweet 1.5b26. Before encoding with lame.exe I?ve decoded ac3 to wav using Besweet. Options used -azid( -c normal -L -3db ) -ota( -hybridgain ) and for Lame -v --vbr-new -V 6 -b 32.
lame.exe
- 3.95.1 ? 97 kbps
- 3.96 ? 110 kbps
lame_enc.dll
- 3.96 alpha ? 92 kbps
- 3.96 ? 108 kbps
And one more thing - at the beginning of this audio track there is a long period of pure silence. 3.95.1 and 3.96 alpha encode it at 32 kbps, and 3.96 (exe and dll) encode it at 320 kbps.
My question is: isn?t it a bug or some faulty optimization? On the other hand I remember that using pre-3.95 lame versions I had bitrates rather similar to 3.96 than 3.95. What do you think?
I have these two versions of Lame: 3.95.1 downloaded from doom9.org and 3.96 from http://mitiok.cjb.net/ . Additionally it appears that lame_enc.dll in doom9 package is 3.96 alpha version. I?ve noticed really big differences in final bitrate during AC3 to MP3 encoding (VBR V6 - I haven?t tried other VBR levels). These differences don?t occur when I encode CD-Audio music (actually they are very small, like 1 kbps).
I made a quick comparison below using Wonder Boys audio track. For dlls I?ve used Besweet 1.5b26. Before encoding with lame.exe I?ve decoded ac3 to wav using Besweet. Options used -azid( -c normal -L -3db ) -ota( -hybridgain ) and for Lame -v --vbr-new -V 6 -b 32.
lame.exe
- 3.95.1 ? 97 kbps
- 3.96 ? 110 kbps
lame_enc.dll
- 3.96 alpha ? 92 kbps
- 3.96 ? 108 kbps
And one more thing - at the beginning of this audio track there is a long period of pure silence. 3.95.1 and 3.96 alpha encode it at 32 kbps, and 3.96 (exe and dll) encode it at 320 kbps.
My question is: isn?t it a bug or some faulty optimization? On the other hand I remember that using pre-3.95 lame versions I had bitrates rather similar to 3.96 than 3.95. What do you think?