Settings for hybrid lossless usage. Goal: economy and faster operations.
-b3x3c ~270k --portable medium
-b4x3c ~370k --portable HQ
Settings for lossy archiving (no wvc) . Goal: quality first, less economy and encoding speed. NO impact on decoding speed.
-b5x4 ~460k --normal HQ
-b6x4 ~540k --extreme
-b7x4 ~630k --insane
The -h switch may be added for slightly more compression / audio quality but at the expense of encode and decode time.
Thanks for posting these shadowking (and of course for trying them all out with exhaustive listening!)
I generally use a higher -x mode just because I'm usually limited by ripping speed anyway, and of course there's no cost once the encoding is done. Obviously transcoding an entire library is a different thing.
We can all agree that WavPack is a present from bryant to the world. There's *nothing* that compares. It is exquisite and unmatched in regards to quality/engineering/portability/etc.
Now, since we're sharing, let me share my own pursuit to the *ultimate* WavPack setting to end them all.
I've came to the conclusion that everything above -b4.35 is an overkill. Don't quote me on this, but I think even bryant uses -b4.35 personally (correct me if I'm wrong). You can feed -b4.35 to WavPack and it will do the right thing no matter the source, be it red book audio or higher, the ratio will stay the same.
Unlike shadowking, I won't separate settings into various use cases, this preset is meant to be singular, *the* one.
Without further ado, this is my personal gold master setting, where I delete any other source material and leave only the WavPack copy:
I don't use portables, if I'm out and about I enjoy the nature and surroundings. A recap of the settings:
Around 384 kbps, usually quite a bit higher. The best thing, this scales along with source material and number of channels.
Best quality/size possible. Enough said. Hurts encoding/decoding time, but with today's hardware this is *not* a problem.
Again, the best possible quality improvement while not affecting decoding time.
Store md5 sum of the encode, generate a new RIFF wav header (discard any extra junk from existing header) and verify after encode.
If you asked me a few years back, there is no way I would use those settings. Today's CPUs are monsters, and using higher WavPack modes is a no brainer.
Encoding time is a non-issue, good things come to those who wait, as with everything in life.
From my extensive experimentation, there is no reason *not* to use highest `-x` and `-hh` settings.
While shadowking's settings are *excellent*, I can only shudder when I see `-b5`, `-b6` or `-b7`. Storage *is* cheap, but why waste it?
To expand a bit on this, I have a nice DAC and Genelec monitors (no sense in revealing exact models). *Never* did I regret using `-b4.35`, not to mention high `hh` and `x6` modes. According to me, I'm *done*. This is my gold master and bryant is a hero! Thanks so much, and thanks again!
Also shadowking, I see your signature is changing all the time.
Right now it's `wavpack 4.80 -b6.24x4`. Previously it was `-b6x4` and before that it was `-b5x4`... Seriously, do you think you hear a difference between those (high-end) settings? I though I have OCD but I digress...
Thanks for the kind words about WavPack; it's greatly appreciated and I'm glad it's working out for you!
Generally I agree with your settings. However I have been working on a project lately where I'm encoding real time on a Raspberry Pi 2, and so -hhx6 is definitely not viable (at least not on a single core) so I am using essentially -b4hx and have been quite happy with the results (I'm not quite as critical as I used to be).
On PCs though I tend to rip directly to lossless (using abcde) and then use the transcoding function to convert to my target settings, and I usually use -x6 then because I'm not pressed for time.
Hello oro and David. I have taken into account the suggestions and arrived at 3 settings:
300k - b3.5hx4lc - hq portable
400k - b4.5hx4lc - hq normal
550k - b6.25hx4l - hq transcoding
The -hx4 setting is used instead of hhx6 as compression is nearly as good and onpar or better than -hh.
The first two settings are similar to 320..384 range and expected to be used with correction files. The last setting was based around 500k , inflated by additional 50k for extra headroom. This one is for those who wish to encode direct as lossless alternative without wvc files. Also for hi -res audio this scales into around 1100k and this is what David recommends for those sampling rates. -b4.5 or 400k can be used for an 'economy' approach if one can allow for (rare) audible deviations.
Hello, tell me please:
1) Which of these settings is preferable for high quality lossy part, not compression?
-m -i -q -b3.5hhx1lc - %d
-m -i -q -b3.5hhx3lc - %d
-m -i -q -b3.5hhx6lc - %d
2) Does key -x* affect quality lossy audio part or only compress?
3) Does key -h or -hh affect quality lossy audio part or only compress?
Sorry for my English, but HYBRID very interesting option. :)
With shadowking or bryant being obviously more qualified to give you the full monty - including the -h and -hh options, I can at least give you a (rough) grasp of how the -x option works, more or less:
Without it, the encoder, as a standard, works in a way where encoding and decoding speeds are roughly the same.
Its use therefore makes the encoder spend more time encoding (in order to yield better compression) without compromising decoding speed.
The higher the number, the larger the bias towards the time the encoders spends, erm, encoding :D (therefore reaching higher compression).
The Wavpack documentation (http://www.wavpack.com/wavpack_doc.html) claims it works more efficiently with synthesized sounds and/or non-standard sampling rates.
Edit: sorry for contributing to a resuscitation, (https://www.thesaurus.com/browse/resuscitation)guys/mods:
In wishing to provide him/her with a reply, only after hitting "Post" did I realize this was quite old a thread.
what is the "trick" of using "bits/sample" instead of more clear (I think for many) "kbits/second". b3.5 - how much is that in kbps? but b3.7 ? where bits can take on decimal ?