On 15/02/16 09:36, Chris Howard - CBR wrote:
Does anyone have an experience integrating r128gain with rivendell?
I have a wide variety of loudness levels in my imported carts.
I'd like to find an automated way to levelize everything,
or at least a large percentage.
Chris,
there is no reliable automated library leveliser. There may be one but
it may not produce a consistent result across varied source material.
There are a number of ways to set the average electrical voltage level
of an audio track to be the same as others in a collection.
The process is called normalisation and the result is based on the
maximum peak voltage in a track so the loudest point sets the level for
the whole track.
It's electrically accurate.
Unfortunately most recorded material involves very complex sounds
containing lots of different frequencies and varying rates of change
between soft and loud sounds.
Some complex sounds sound louder then others even though they may peak
at the same 'level' measured by a voltmeter.
“Compression” processing audio so the difference between high and low
levels is reduced can also make a recording appear to be louder.
So many many dollars and lots of time have been applied to trying to
invent systems that can process various recordings so they sound more
or less the same loudness.
With varying success.
The problem being that the human ear and brain processes sound over a
wide dynamic range and a significant frequency range in an elastic
logarithmic manner. Low level sounds get processed differently to high
level sounds. Loudness is affected by how much of the frequency range
you can hear. A mix that sounds great in a studio can sound very
different in a car or in a kitchen. Every attempt to emulate the ear
with some dynamic sound processor involves some compromise and I have
only ever found one system that indicates some study of the human ear in
it's design.
Audio processing is usually applied to the signal leaving the studio in
order to capture everything and to eliminate peaks or overloads on the
transmission system.
With the music library stored on a hard drive as wav files it's tempting
to use some sort of electronic iron to make them all sound the same
loudness but I don’t think it's a good idea.
Ideally you have used the best possible sound card to get the tracks
from source [CD, Vinyl, Cassette, etc ] into your library. In the
Rivendell system tracks are stored at -13db which allows very necessary
headroom. If you have normalised them the loudest sound will be at -13db
which may not be perfect for all of the track but the audio won't be
distorted.
When you play tracks in your production studio they will all sound
different. Depending on what processing as applied at source some will
be louder then others.
You can set the play-out level higher or lower which has no effect on
the actual audio, just on the sound card when this track is played.
Some songs peak at 0 and will sound fine. I have some that get played as
low as -6 and some classical numbers that get quite a boost. The
outgoing limiter will handle the louder bits of the classical tracks.
The soft bits need a bit of lift. Use some speakers that a typical
listener would use. No matter how you grade tracks there will be a
listener who has a system that does not respond the same as others.
There will be other listeners who hear differently because we are not
constructed on production lines. You can only do as best you can.
Rivendell has a numeric and a visual indication of the level setting for
each track in RD Library which is a good place to start, but you still
need to check each track. Looking at the track waveform you soon get to
know where to do spot checks so you can grade a complete album in a few
minutes.
In the same way that the creation of an mp3 by figuring what can be left
out and still 'sound the same' produces tracks that on a A:B comparison
are immediately obviously 'lossy'; anything that is going to modify your
tracks to make them the same loudness is a risk I would not take.
The human ear is logarithmic, frequency conscious at different levels,
and has a dynamic range about 3 or 4 times greater than any broadcast
system. We all hear differently, and differently at different ages.
Take care to import tracks with a high quality sound card. Grade levels
as appropriate. Have a system in the outgoing chain which is going to
manage peaks [Limiter] and perhaps tighten up the dynamic range
[Compressor].
There are a couple of multi-band processors that don't do a bad job.
Jamin can be set up to deliver a consistent sound in a Jack matrix.
But don't compromise those audio tracks because they are very labour
intensive to replace!
Regards
Robert Jeffares
_______________________________________________
Rivendell-dev mailing list
[email protected]
http://caspian.paravelsystems.com/mailman/listinfo/rivendell-dev
_______________________________________________
Rivendell-dev mailing list
[email protected]
http://caspian.paravelsystems.com/mailman/listinfo/rivendell-dev
