Tuesday, 18 December 2012
Friday, 7 December 2012
New Version Of Sonic Field: 0.2.1
Here is the latest drop:
http://www.sonicfield.co.uk/downloads/SonicField-0-2-1.zip
I have expedited this this minor number release to make the new dynamic range compression patches I have been talking about recently available to everyone. The front page of sonicfield.co.uk has been amended so the download link points to the new version.
Version 0.2.1
=============
This version (0.2.1) of Sonic Field has many enhancements including but not limited to:
1) Wave limiting and wave maximizing
2) Automatic reading in a swapped out signals
3) Enhanced granular system with bounded random grain size
4) Performance enhancements in the Do engine
5) Performance enhancements in the parallelisation drier
6) More accurate line number reporting in errors
7) Many minor tweaks
8) Addition of youtube-voice.sfp single task sfp file
9) Performance counters are working again
10) Work on granular compression in revoice patchs.
Previous version (0.2) of Sonic Field has many enhancements including but not limited to:
1) Granular synthesis
2) RBJ IIR Filers (high,low,high-shelf,low-shelf,peaking)
3) Cleaner implementation of resonance synthesis
4) Better error handling - fix the line number bug
5) System hang on error in paralleliser bug fixed
6) Envelope follower.
7) Fix to numerous bugs in shaped filters
8) Addition of shaped butterworth and bessel bandpass filters
9) Improved FM signal modulation and addition of fm signal generator
10) Addition of shaped power
Thursday, 6 December 2012
Compression-102: Loudness Wars And Extreme Compression
Compression 102, how do we boost loudness via dynamic range compression to catch people's attention.
This is not about making audio sound better; it is about making it sound more striking. It is about capturing people's attention and hooking them into a radio program, or the background in a music store. It is about making sales of music, or in advert breaks. This is a fascinating area, though it is important to realise that we are not attempting to make stuff sound better in a classic sense, just louder.
Here is the video - I have put the compression patch used at the bottom of this post:
http://youtu.be/l_75rzdaxkg
"temp/step-1.wav" ReadFile ^left ^right
{
(
(?signal Reverse,1,50)Follow Reverse,
(?signal ,1,50)Follow,
)Mix !shape
(?shape,(2,?shape MaxValue)/)NumericVolume !shape
(0.04,>shape)DirectMix Clip !shape
(>signal,?shape)Multiply Normalise !signal
(
(?signal Normalise,25,25)Follow,
(?signal Normalise Reverse,25,25)Follow Reverse,
)Mix !shape
(?shape,(1,?shape MaxValue)/)NumericVolume !shape
(0.3,>shape dbs+24)DirectMix !shape
(
>signal,
>shape
)Divide Normalise !signal
}!envelope-compress
(?left ,2000,6)BesselHighPass !signal-h
(?left ,2000,6)BesselLowPass !signal
(>signal,150,1,3)RBJPeaking !signal
?envelope-compress Invoke
(
(30 Silence,(>signal dbs+6 WaveLimit,2000,6)BesselLowPass)Concatenate,
>signal-h dbs+6 WaveLimit
)Mix Normalise dbs+3 WaveLimit !left-new
(?right,2000,6)BesselHighPass !signal-h
(?right,2000,6)BesselLowPass !signal
(>signal,150,1,3)RBJPeaking !signal
?envelope-compress Invoke
(
(30 Silence,(>signal dbs+6 WaveLimit,2000,6)BesselLowPass)Concatenate,
>signal-h dbs+6 WaveLimit
)Mix Normalise dbs+3 WaveLimit !right-new
((>left-new,>right-new)StereoMonitor,"temp/tone.wav")WriteFile32
Wednesday, 5 December 2012
Compression 101
I have recently put up a video introducing the basics of audio compression (dynamic range compression) for voice:
The patch used in this video is this one (remember AGPL):
2 SetSwapLimit
"temp/input.wav" ReadFile ^signal
?signal !orig
[Low peak cut]
(?signal,150,3)ButterworthLowPass !lower
(?signal,150,3)ButterworthHighPass !upper
(>lower,60,2)ButterworthHighPass !lower
(?lower Normalise,100,100)Follow !shape
(?shape,(1,?shape MaxValue)/)NumericVolume !shape
(0.1,>shape dbs+48)DirectMix !shape
(
>lower,
>shape
)Divide !lower
(
>lower,
>upper
)Mix Normalise !signal
[ De-ess ]
(?signal,4000,5)ButterworthLowPass !lower
(?signal,4000,5)ButterworthHighPass !upper
(?upper Normalise,100,100)Follow !shape
(?shape,(1,?shape MaxValue)/)NumericVolume !shape
(0.1,>shape dbs+48)DirectMix !shape
(
>upper,
>shape
)Divide !upper
(
>lower,
>upper dbs+3
)Mix Normalise !signal
{
(
(?signal Reverse,1,50)Follow Reverse,
(?signal ,1,50)Follow,
)Mix !shape
(?shape,(2,?shape MaxValue)/)NumericVolume !shape
(0.04,>shape)DirectMix Clip !shape
(>signal,?shape)Multiply Normalise !signal
(
(?signal Normalise,25,25)Follow,
(?signal Normalise Reverse,25,25)Follow Reverse,
)Mix !shape
(?shape,(1,?shape MaxValue)/)NumericVolume !shape
(0.3,>shape dbs+24)DirectMix !shape
(
>signal,
>shape
)Divide Normalise !signal
}!envelope-compress
{
(?signal,50,10)Granulate !grains
Bunch !output
(
>grains,
{
^grain ^time
(
(?time,(?time,?grain length)+,?signal)Cut !sample
(?grain Rectify MaxValue !v,?gate-point)Gt,
{
(
(?v,?slam-point)Gt,
{
(>grain,(?slam-point,?v)/)NumericVolume !grain
},
{
[ Do nothing ]
}
)Choose Invoke
}
,
{
(>grain,(1,(?gate-point,?v)/)/)NumericVolume !grain
}
)Choose Invoke
((>grain swapout,>time),>output)AddEnd !output
}
)InvokeAll
>output MixAt Normalise !signal
}!compress
{
0.02 !gate-point
0.50 !slam-point
[ Do in pairs to avoid running out of memory ]
(
(
?compress,
)DoAll Mix Swapout,
(
?compress,
)DoAll Mix Swapout,
(
?compress,
)DoAll Mix Swapout,
(
?compress,
)DoAll Mix Swapout,
)Mix Normalise !signal
}!granular-compress
?envelope-compress Invoke
[ Reverb ]
(15 Silence,?signal)Concatenate !note-left
?signal !note-right
(>note-left ,1000,0.25)RBJLowPass !note-left
(>note-right,1000,0.25)RBJLowPass !note-right
(
1,2,
{
!outer
(
1,4,
{
!t
{((30,(?t,(24,?outer)*)*)+ Prime Silence,?note-left )Concatenate} Do !nlTask
{((45,(?t,(15,?outer)*)*)+ Prime Silence,?note-right)Concatenate} Do !nrTask
{((>nlTask Done, -15)Volume,(700,(?t,3)*)-,1)ButterworthLowPass} Do !nlTask
{((>nrTask Done, -15)Volume,(700,(?t,3)*)-,1)ButterworthLowPass} Do !nrTask
{(>nlTask Done, ?note-left )Mix Normalise} Do !nlTask
{(>nrTask Done, ?note-right)Mix Normalise} Do !nrTask
?nlTask Done !note-left
?nrTask Done !note-right
}
)Repeat
}
)Repeat
(
>note-left dbs-12,
?signal
)Mix Normalise !left
(
>note-right dbs-12,
>signal
)Mix Normalise !right
(>left ,3000,2,6)RBJPeaking Normalise !left
(>right,3000,2,6)RBJPeaking Normalise !right
(>left, 150,2,12)RBJPeaking Normalise !left
(>right,150,2,12)RBJPeaking Normalise !right
>left !signal ?granular-compress Invoke
>signal dbs+6 WaveLimit !left
>right !signal ?granular-compress Invoke
>signal dbs+6 WaveLimit !right
(
(
>left,
>right
),"temp/step-1.wav"
)WriteFile32
Friday, 30 November 2012
Speaking Voice Processing
Sonic Field started life as an audio processing project - back to its roots!
The following script uses envelope followers to remove background rumble from the recording without taking out the lows from the voice. It also has a de-esser which is much the same thing but for high frequencies. Some light reverb is use via a very simple algorithm which folds the signal back on its self. This reverb' is seeded using a bit of Haas effect form the mono code.
The script also does some gating to remove noise between the words and compression of the over all signal to make the volume even through out. It does have a nasty habit of not catching very short audio spikes which end up left in the signal. I need to figure out how to fix this, but it is so rare that a manual tweak in Audacity does the trick.
(>left, 150,2,12)RBJPeaking Normalise !left
(>right,150,2,12)RBJPeaking Normalise !right
The above filters then make the voice sound deeper. The values would need tweaking for a female voice.
Here is a video where I used the script to process my voice:
http://youtu.be/7qhoXL7EKYs
Here is the script:
"temp/input.wav" ReadFile ^signal
?signal !orig
[Low peak cut]
(?signal,150,3)ButterworthLowPass !lower
(?signal,150,3)ButterworthHighPass !upper
(>lower,60,2)ButterworthHighPass !lower
(?lower Normalise,100,100)Follow !shape
(?shape,(1,?shape MaxValue)/)NumericVolume !shape
(0.1,>shape dbs+48)DirectMix !shape
(
>lower,
>shape
)Divide !lower
(
>lower,
>upper
)Mix Normalise !signal
[ De-ess ]
(?signal,4000,5)ButterworthLowPass !lower
(?signal,4000,5)ButterworthHighPass !upper
(?upper Normalise,100,100)Follow !shape
(?shape,(1,?shape MaxValue)/)NumericVolume !shape
(0.1,>shape dbs+48)DirectMix !shape
(
>upper,
>shape
)Divide !upper
(
>lower,
>upper dbs+3
)Mix Normalise !signal
(
(?signal Reverse,1,50)Follow Reverse,
(?signal ,1,50)Follow,
)Mix !shape
(?shape,(2,?shape MaxValue)/)NumericVolume !shape
(0.04,>shape)DirectMix Clip !shape
(>signal,?shape)Multiply Normalise !signal
(
(?signal Normalise,25,25)Follow,
(?signal Normalise Reverse,25,25)Follow Reverse,
)Mix !shape
(?shape,(1,?shape MaxValue)/)NumericVolume !shape
(0.3,>shape dbs+24)DirectMix !shape
(
>signal,
>shape
)Divide Normalise !signal
[ Reverb ]
(15 Silence,?signal)Concatenate !note-left
?signal !note-right
(>note-left ,1000,0.25)RBJLowPass !note-left
(>note-right,1000,0.25)RBJLowPass !note-right
(
1,2,
{
!outer
(
1,4,
{
!t
{((30,(?t,(24,?outer)*)*)+ Prime Silence,?note-left )Concatenate} Do !nlTask
{((45,(?t,(15,?outer)*)*)+ Prime Silence,?note-right)Concatenate} Do !nrTask
{((>nlTask Done, -15)Volume,(700,(?t,3)*)-,1)ButterworthLowPass} Do !nlTask
{((>nrTask Done, -15)Volume,(700,(?t,3)*)-,1)ButterworthLowPass} Do !nrTask
{(>nlTask Done, ?note-left )Mix Normalise} Do !nlTask
{(>nrTask Done, ?note-right)Mix Normalise} Do !nrTask
?nlTask Done !note-left
?nrTask Done !note-right
}
)Repeat
}
)Repeat
(
>note-left dbs-20,
?signal
)Mix Normalise !left
(
>note-right dbs-20,
>signal
)Mix Normalise !right
(>left ,3000,2,6)RBJPeaking Normalise !left
(>right,3000,2,6)RBJPeaking Normalise !right
(>left, 150,2,12)RBJPeaking Normalise !left
(>right,150,2,12)RBJPeaking Normalise !right
(
(
>left,
>right
)StereoMonitor,"temp/tone.wav"
)WriteFile32
Tuesday, 27 November 2012
Working With The Sonic Field Console
Having A Console Makes A Huge Difference To Productivity And Creativity
Up until now the productivity of Sonic
Field has been severely limited by the need to re-run a patch every
time it is edited; this make playing around with sound a very slow
process indeed. The addition of the Console processor opens up a
whole new arena of rapid sound development from Sonic Field. It also
opens up lots of possibilities for fun.
The implementation of Console
at the time of writing this is about as simple as it is possible to
be. In agile style, I got something working and will use it to find
out ways to improve it.
How It Works
At any point in a patch the Processor
Console can be used. This will start interacting with the console. It
simply reads from Standard Input. This works OK with the Eclipse
console which Sonic Field gets when run under Eclipse. The console
accepts lines of input. If a line consist of just:
%end%
then the console processor stops and
the next processor in the patch is run.
The Console accumulates a patch. Each
line of input is added to the patch. When an line consisting of :
%
and nothing else is entered, the
current accumulated patch is parsed and run. This allows for complex
patches to be either typed or pasted into the console window and then
run. The patch runs in the context (same store values) as the patch
which contained the original Console processor. However, there is a
difference.
When a patch launched from the command
line (or Eclipse) causes an error Sonic Field shuts down. Patches
entered in the Console to not cause Sonic Field to shut down in this
way. If you make a mistake in the Console Sonic Field will tell you
something about the mistake and then just start reading from Standard
Input again.
One Small Issue:
At the time of writing there is a
slight issue in that if there is a layout problem with code inside a
code block which is entered into the Console Sonic Field gets the
code block tracking system a bif stuffed up. If, after a patch layout
problem, you keep getting 'unmatched brace' errors, then enter
}
%
An Example:
Below is a slightly abridged example
form the Eclipse console of me working on a bass/rhythm line in Sonic
Field
SFPL:
[3]1 >> Sonic Field Thread Factory:
SFPL:
[3]1 >> ---------------------------
SFPL:
[3]1 >> Found CPUs: 8
SFPL:
[3]1 >> Initial Max Threads: 16
Parsing:
scripts/play/FastBass.sfpl size: 866
SFPL:
[532]1 >> Sonic Field Version 0.2 Starting Processing
SFPL:
[532]1 >> ============================================
?play
invoke
%
(
(0,2.0),
((?bt,2)*,0),
((?bt,4)*,2)
)NumericShape
!qEnv
%
?play
invoke
%
{
((?bt,4)*,?pitch)SinWave
MakeSawTooth dbs+3 Saturate Normalise !signal
(?qEnv,4)DirectRelength
!qsEnv
(?qsEnv,?qsEnv,?qsEnv,?qsEnv,?qsEnv)Concatenate
!qsEnv
(0,?signal
Length,>qsEnv)Cut !qsEnv
(?pitchEnv,4)DirectRelength
!psEnv
(?psEnv,?psEnv,?psEnv,?psEnv,?psEnv)Concatenate
!psEnv
(0,?signal
Length,>psEnv)Cut !psEnv
(>signal,?psEnv,?qsEnv)ShapedLadderLowPass
Normalise !signal
>signal
Normalise !signal
(>signal,2)Power
Normalise !signal
(>signal,?pitch,1,24)RBJPeaking
Normalise !signal Monitor
}!play
%
?play
invoke
%
"A2"
Note !pitch
?play
invoke
%
com.nerdscentral.sfpl.SFPL_RuntimeException:
Runtime Exception at key word Invoke
operand: '109.99999987361397'
at line: 0, colum 30 in Console
at
com.nerdscentral.sfpl.SFPL_Runner.Invoke(SFPL_Runner.java:103)
at
com.nerdscentral.audio.io.SF_Console.Interpret(SF_Console.java:57)
at
com.nerdscentral.sfpl.SFPL_Runner.Invoke(SFPL_Runner.java:87)
at
com.nerdscentral.sfpl.SFPL_Runner.Invoke(SFPL_Runner.java:64)
at
com.nerdscentral.sfpl.RenderRunnerImp.runParser(RenderRunnerImp.java:163)
at
com.nerdscentral.sfpl.RenderRunnerImp.render(RenderRunnerImp.java:65)
at
com.nerdscentral.sfpl.RenderRunner.main(RenderRunner.java:25)
Caused
by: com.nerdscentral.sfpl.SFPL_RuntimeException:
Failed to resolve type: class java.lang.Double
at
com.nerdscentral.sfpl.Caster.describeSFPLType(Caster.java:55)
at
com.nerdscentral.sfpl.Caster.throwConversion(Caster.java:49)
at
com.nerdscentral.sfpl.Caster.makeBunch(Caster.java:68)
at
com.nerdscentral.sfpl.SFPL_Parser$cInvoke.Interpret(SFPL_Parser.java:1166)
at
com.nerdscentral.sfpl.SFPL_Runner.Invoke(SFPL_Runner.java:87)
...
6 more
"A2"
Note !pitch
%
?play
println
%
com.nerdscentral.sfpl.SFPL_Runner@c6824d9
?play
println
%
com.nerdscentral.sfpl.SFPL_Runner@c6824d9
?play
invoke
%
{
((?bt,4)*,?pitch)SinWave
MakeSawTooth dbs+3 Saturate Normalise !signal
(?qEnv,4)DirectRelength
!qsEnv
(?qsEnv,?qsEnv,?qsEnv,?qsEnv,?qsEnv)Concatenate
!qsEnv
(0,?signal
Length,>qsEnv)Cut !qsEnv
(?pitchEnv,4)DirectRelength
!psEnv
(?psEnv,?psEnv,?psEnv,?psEnv,?psEnv)Concatenate
!psEnv
(0,?signal
Length,>psEnv)Cut !psEnv
(>signal,?psEnv,?qsEnv)ShapedLadderLowPass
Normalise !signal
>signal
Normalise !signal
(>signal,2)Power
Normalise !signal
(>signal,?pitch,1,24)RBJPeaking
Normalise dbs+3 Saturate Normalise !signal
?bt
WhiteNoise MakeTriangle Normalise !signalb
(
>signal,>signalb
)Concatenate
!signal Monitor
}!play
%
?play
invoke
%
?play
invoke
%
{
((?bt,4)*,?pitch)SinWave
MakeSawTooth dbs+3 Saturate Normalise !signal
(?qEnv,4)DirectRelength
!qsEnv
(?qsEnv,?qsEnv,?qsEnv,?qsEnv,?qsEnv)Concatenate
!qsEnv
(0,?signal
Length,>qsEnv)Cut !qsEnv
(?pitchEnv,4)DirectRelength
!psEnv
(?psEnv,?psEnv,?psEnv,?psEnv,?psEnv)Concatenate
!psEnv
(0,?signal
Length,>psEnv)Cut !psEnv
(>signal,?psEnv,?qsEnv)ShapedLadderLowPass
Normalise !signal
>signal
Normalise !signal
(>signal,2)Power
Normalise !signal
?bt
WhiteNoise MakeSawTooth !signalb
(?signalb,(?pitch,4)*,24,1.5)RBJLowShelf
!signalb
(
((0,0),(100,1),(?signalb
Length,0))NumericShape,
>signalb
)Multiply
Normalise !signalb
(
>signal,>signalb
)Concatenate
!signal
(>signal,?pitch,1,24)RBJPeaking
Normalise dbs+3 Saturate Normalise !signal Monitor
}!play
%
?play
invoke
%
{
((?bt,4)*,?pitch)SinWave
MakeSawTooth dbs+3 Saturate Normalise !signal
(?qEnv,4)DirectRelength
!qsEnv
(?qsEnv,?qsEnv,?qsEnv,?qsEnv,?qsEnv)Concatenate
!qsEnv
(0,?signal
Length,>qsEnv)Cut !qsEnv
(?pitchEnv,4)DirectRelength
!psEnv
(?psEnv,?psEnv,?psEnv,?psEnv,?psEnv)Concatenate
!psEnv
(0,?signal
Length,>psEnv)Cut !psEnv
(>signal,?psEnv,?qsEnv)ShapedLadderLowPass
Normalise !signal
>signal
Normalise !signal
(>signal,2)Power
Normalise !signal
?bt
WhiteNoise MakeSawTooth !signalb
(?signalb,(?pitch,4)*,1.5,24)RBJLowShelf
!signalb
(
((0,0),(100,1),(?signalb
Length,0))NumericShape,
>signalb
)Multiply
Normalise !signalb
(
>signal,>signalb
)Concatenate
!signal
(>signal,?pitch,1,24)RBJPeaking
Normalise dbs+3 Saturate Normalise !signal Monitor
}!play
%
?play
invoke
%
{
((?bt,4)*,?pitch)SinWave
MakeSawTooth dbs+3 Saturate Normalise !signal
(?qEnv,4)DirectRelength
!qsEnv
(?qsEnv,?qsEnv,?qsEnv,?qsEnv,?qsEnv)Concatenate
!qsEnv
(0,?signal
Length,>qsEnv)Cut !qsEnv
(?pitchEnv,4)DirectRelength
!psEnv
(?psEnv,?psEnv,?psEnv,?psEnv,?psEnv)Concatenate
!psEnv
(0,?signal
Length,>psEnv)Cut !psEnv
(>signal,?psEnv,?qsEnv)ShapedLadderLowPass
Normalise !signal
>signal
Normalise !signal
(>signal,2)Power
Normalise !signal
?bt
WhiteNoise MakeSawTooth !signalb
(?signalb,(?pitch,2)*,1.5,24)RBJLowShelf
!signalb
(
((0,0),(100,1),(?signalb
Length,0))NumericShape,
>signalb
)Multiply
Normalise !signalb
(
>signal,>signalb
dbs+12
)Concatenate
!signal
(>signal,?pitch,1,24)RBJPeaking
Normalise dbs+3 Saturate Normalise !signal Monitor
}!play
%
?play
invoke
%
?signal
monitor
%
((?signal),"temp/tone.wav")WriteFile32
%
Subscribe to:
Posts (Atom)