Hi!
First, thank you for the great feedback, testing and samples...
OptimFROG Dual Stream is aimed at filling the big gap between
perceptual coding and lossless coding. The goal is to offer real
transparent audio coding (not only perceptually transparent) at
half or less the bitrate generally used by lossless coding, and
also to permit progressive consistent increase of the quality
level, until lossless coding is reached.
These are the changes for the latest OptimFROG DualStream,
version 4.507b, compared to version 4.507a2:
- stable format, not anymore alpha stage
- new correction file format, works now with the --ans option
- corrected the --ans (advanced noise shaping) option, which
previously used suboptimal formulas, and works better now
- no longer creates pure lossless files, use OFR instead
- default (no command line switches) encodes at --quality 3
- corrected some small decoder problems with high amplitude signals
- the --check option works with --correction to check MD5 hash
- new input plug-ins (compiled also for dBpowerAMP and Winamp3)
- decoding compatible with the alpha versions (4.507a and 4.507a2),
but not with their correction files
- removed raw file support, which was not implemented
- small fixes and improvements
Download it (Win32 console+foobar2000+dBpowerAMP+Winamp3+
Winamp2 input plug-ins) at
http://LosslessAudioCompression.com/Downlo...m_4507b_zip.php
You may check all the details directly at
http://LosslessAudioCompression.com/DualStream.php
or by visiting
http://LosslessAudioCompression.com
Speek made a front-end for OptimFROG DualStream which you
can download at
http://home.wanadoo.nl/~w.speek/kermit.htm
You can find the previous HA discussion about DualStream at:
http://www.hydrogenaudio.org/forums/index....showtopic=10642
@den: A few more tests for you... :-)
Florin Ghido
Full Version: OptimFROG DualStream 4.507b released!
QUOTE
@den: A few more tests for you... :-)
Excellent, I'm there. Will post some comments in a couple of days!
B)
Den.
OK, some initial comments from early testing.
I used the Ring Of Fire sample, same as I used last time.
ABX scores:
ofsb --bitrate 275 15/15: Slight increase in background hiss. Not bad, but there. Seems less annoying than the last test I did with the "a" release. Is that possible Florin? Did you improve the straight lossy mode at all?
ofsb --bitrate 275 --ans 15/15: Slight background hiss, but better than without ans.
ofsb --bitrate 320 15/15: Hiss is barely noticeable, I almost missed it. Again seems better than the "a" release without ans IIRC.
ofsb --bitrate 320 --ans 15/15: Very good. If I really concentrate, I can just pick slight movement in the noise, but much, much better than the ans noise movement from the ofs "a" release.
ofsb --bitrate 400 8/15, 13/15: Getting very hard now. Can just detect a change in the hiss when listened at high volumes.
ofsb --bitrate 400 --ans, 15/15: Again, very hard to pick, could only do it immediately after the above test. I think I can still hear some noise moving with the notes, but again, much better than how I remember the ans on the "a" release.
ofsb quality 6, which yielded a 431 kbit file 8/15, 7/15, 5,15: Got me. Transparent.
B)
Nice. Certainly has a significant improvement over the previous release. I was concerned about the auto noise shaping before, particular with pulsing type music. It is much better than before, in terms of the noise moving around with the music and I suspect that some may not be able to pick it. I'm still not sure whether ans is a huge benefit, but I need to try it further across a range of samples. When you use ans with the quality mode, it appears to give you the same quality file, just with a few less bits (~5-15 kbits less at quality 2 for the samples I've tested so far), and as you go up in quality setting/bitrate it becomes neglible.
I really like the quality mode though, and this is the way to go I suspect, unless you must have control over the bitrate. The quality measurement and control implemented seems to work quite consistently, alleviating the need to use a range of bitrate attempts and custom commandlines to get a certain quality encoding.
Rather than bomb this thread with lots of tests, I'll just continue having a play with this and if something interesting comes up, I'll post something.
@Florin
I have another question. In your post and readme file you state that ofs is no longer completely lossless. Why have a correction file then? Perhaps using the correction file gives you a near lossless file, but not a perfect copy?
In my own case, because of my relatively slow PC, and the fact that I decode slabs of files everyday for my Minidisc, I will probably be sticking to Wavpack, but I'm going to continue playing with Dualstream on a semi daily basis, just to see what this puppy can do!
Thanks again for your efforts Florin. The future is looking very interesting with DualStream and Wavpack hybrid/lossy for my particular needs. I really appreciate what you and David has given the audio community with these programs.
Den.
I used the Ring Of Fire sample, same as I used last time.
ABX scores:
ofsb --bitrate 275 15/15: Slight increase in background hiss. Not bad, but there. Seems less annoying than the last test I did with the "a" release. Is that possible Florin? Did you improve the straight lossy mode at all?
ofsb --bitrate 275 --ans 15/15: Slight background hiss, but better than without ans.
ofsb --bitrate 320 15/15: Hiss is barely noticeable, I almost missed it. Again seems better than the "a" release without ans IIRC.
ofsb --bitrate 320 --ans 15/15: Very good. If I really concentrate, I can just pick slight movement in the noise, but much, much better than the ans noise movement from the ofs "a" release.
ofsb --bitrate 400 8/15, 13/15: Getting very hard now. Can just detect a change in the hiss when listened at high volumes.
ofsb --bitrate 400 --ans, 15/15: Again, very hard to pick, could only do it immediately after the above test. I think I can still hear some noise moving with the notes, but again, much better than how I remember the ans on the "a" release.
ofsb quality 6, which yielded a 431 kbit file 8/15, 7/15, 5,15: Got me. Transparent.
B)
Nice. Certainly has a significant improvement over the previous release. I was concerned about the auto noise shaping before, particular with pulsing type music. It is much better than before, in terms of the noise moving around with the music and I suspect that some may not be able to pick it. I'm still not sure whether ans is a huge benefit, but I need to try it further across a range of samples. When you use ans with the quality mode, it appears to give you the same quality file, just with a few less bits (~5-15 kbits less at quality 2 for the samples I've tested so far), and as you go up in quality setting/bitrate it becomes neglible.
I really like the quality mode though, and this is the way to go I suspect, unless you must have control over the bitrate. The quality measurement and control implemented seems to work quite consistently, alleviating the need to use a range of bitrate attempts and custom commandlines to get a certain quality encoding.
Rather than bomb this thread with lots of tests, I'll just continue having a play with this and if something interesting comes up, I'll post something.
@Florin
I have another question. In your post and readme file you state that ofs is no longer completely lossless. Why have a correction file then? Perhaps using the correction file gives you a near lossless file, but not a perfect copy?
In my own case, because of my relatively slow PC, and the fact that I decode slabs of files everyday for my Minidisc, I will probably be sticking to Wavpack, but I'm going to continue playing with Dualstream on a semi daily basis, just to see what this puppy can do!
Thanks again for your efforts Florin. The future is looking very interesting with DualStream and Wavpack hybrid/lossy for my particular needs. I really appreciate what you and David has given the audio community with these programs.
Den.
QUOTE(den @ Jul 21 2003, 09:32 PM)
@Florin
I have another question. In your post and readme file you state that ofs is no longer completely lossless. Why have a correction file then? Perhaps using the correction file gives you a near lossless file, but not a perfect copy?
I have another question. In your post and readme file you state that ofs is no longer completely lossless. Why have a correction file then? Perhaps using the correction file gives you a near lossless file, but not a perfect copy?
I think he means that OFS will no longer create pure lossless files when you don't specify a quality or bitrate, which is what the previous versions did. Now, if you don't specify a quality or bitrate you get --quality 3, so there's no "official" way to go lossless (although --bitrate 2000 does it
).The correction files still work losslessly.
BTW, thanks for all your testing den and good job Florin!
"slaps head" 
Yep, you're right David. It is lossless with the correction file, I just tried it...
Den.
Yep, you're right David. It is lossless with the correction file, I just tried it...
Den.
QUOTE(ghido)
OptimFROG Dual Stream is aimed at filling the big gap between perceptual coding and lossless coding. The goal is to offer real transparent audio coding (not only perceptually transparent) at half or less the bitrate generally used by lossless coding, and also to permit progressive consistent increase of the quality level, until lossless coding is reached.
Sweeeet
Does OptimFrog have a predictor as well?
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.