IRCAM / FORUMNET

IRCAM Lab Created by

the IRCAM Analysis Synthesis Team
the IRCAM Valorisation Dept.
Chapitre 1

INTRODUCTION

1- Installation

2- Autorization

3- Starting and Configurations

Section 1

INSTALLATION

INSTALLATION

1. Open the Disk Image 1. Open the Disk Image “IRCAMLab TS.dmg” by double clicking on the icon

2. Slide the App

3. You’re Done !

2. Inside the Disk Image, slide the IrcamLab TS.app in the Applications folder
Alias.

3. Double click or Open the IrcamLab TS.app located in your Applications folder

2
Section 2

STARTING and SETTINGS

MENU 1. Open IRCAM Lab T.S

1. Open T.S

2. Audio Configuration

2. Select Menu > Preference

a) Select the Audio tab

b) Select your audio output and your audio

input

c) (optional) you can also set a different

sample rate and audio buffer size. 512
samples (and above) should be the default
and adviced minimum size.

3. To configure MIDI see MIDI REMOTE section

3
4
Chapitre 2

USER MANUAL

1- Import Audio

2- Screen fonctions

3- Transpose & Stretch

4- Remix

5- Transport
Section 1

IMPORT AUDIO

MENU You have 3 different ways to open an audio file in T.S. Files must be in either .wav
1. Load a file or .aiff formats.

2. Record Audio Live 1- LOAD A FILE

3. Record audio thru Sound Flower a- cmd O: open an audio file (.wav) or open files in the file menu
b- Drag & drop: an audio file or a project (.tsproj) directly into the application
c- Click on browse in the middle of the application and select the desired file

2- RECORD AUDIO
a- Select the audio entry in the menu “preference”
b- Select “record” in the menu “file”
c- Click on “Click to start recording”

6
3- RECORD THRU SOUND FLOWER via a HOST

a- Sélectionner «Sound Flower» dans les préférences Audio Input

b- Sélectionner Sound Flower Output in your Host Audio output
c- Open Record in menu File and Click for recording.

7
Section 2

SAVE PRESETS / PROJECTS

MENU 1- Save / Load preset

1. Save / Load Presets You can save the parameters configuration (transpose, stretch, f0 max...) in a
presets file for a later reuse. The option is accessible through the File Menu > Save
2. Project File
Presets or File Menu > Load Presets

2- Project files
IRCAM Lab TS can also save a project file (file extension .tsproj) wich
will contain the engine parameters (preset) and also the audio file
absolute path for a later reuse.
Once you open the project, TS will reload the file and the parameters
automatically.
If Ircam Lab TS cannot find the audio file it will ask you for a possible location.

8
Section 3

DISPLAY / ZOOM

MENU T.S lets you display frequency and spectral contents in one window.
1. Spectra and Waveform

2. Bottom bar functions

3. Other view settings

4. Zoom and Scrolling

Chan Select Mix View Color Select Spectral Adjustment view

On/ Off
a- Frequency analyzer display (blue)
b-Data positions display (note, velocity, etc.)

9
2. Bottom bar functions 3. Other view settings
Chan. Select: Lets you see the channels selected Fonction On/Off :
Mix View: Enables a visual mix between the spectrum and the
a- Follow playback
wave form
Color Select: 6 different color maps to visualize the spectrum b- Logarithmic view (sonogram)
Spectral Adjustment: Ajust the energy and sensitivity of the c- Show Keyboard overlay
spectrum
4. Zoom and Scrolling
On/off Music Datas: Lets you know the cursur’s note, frequency,
energy and MIDI information Click and drag your mouse in the time ruler view to perform a
On/off Analyser View: Opens the analyzer window, includes a zoom in/out and an horizontal scrolling. The vertical mouse
resizing option moves will zoom in when you pull down the mouse and zoom out
when you push it up. Accordingly you can slide left or right the
mouse to navigate horizontally through the waveform.

mouse cursor in a bubble.

On/Off Analyser view: show/hide the instantaneous spectrum.

10
Section 4

TRANSPOSE & STRETCH Fonctions

MENU These modules are the main transposition functions for time-stretching and
1. Fonctions customizable user modes.

2. Transpose

3. Stretch
Transpose : +/- 24 half tones
4. x100 Fx
Formant : +/- 2400 Cts

Strech : x 0.50 to x 10

OTHER FUNCTIONS
1/2 Tone: Transpose by ½ or 1/100 tone
Tape Mode: Just like an analogue tape deck
Formant: On/Off for the Formant function, the default value is 0000
30% - x100: Lets you choose from 3 stretching presets +/- 30% - Normal - X100
Link: Lets you synchronize the Transpose and Formant parameters
MIDI Remote: cmd m Provides you with external MIDI controls of each of the
parameters Transpose, Formant, and Stretch.

11
Section 5

REMIX Fonctions

MENU 1. Remix Functions

1. Remix Functions
This module, unique at IRCAM, makes it
2. Error & Relax possible to mix individually the transient,
sinusoïdal and noise components of an
audio signal.

a- Transient

b- Sinusoïdal

c- Noise

2. Additional Error & Relax Parameters

The Relax parameter knob controls the time distance for
morphing between transient factor and stationary factor. Usually if
you wish to remove the transients from an audio signal you set
this value to the max duration of the transients in.

The Error parameter knob controls the desired noise missclassification probability
(from 0 to 1). For the default value (0.1) about 10% of the noise peaks are
classified as sinusoids.

12
Section 6

TRANSPORT

MENU The Transport module includes the following functions (from left to right):
1. Transport fonctions
Return Rewind

Stop Stop playback

Play Start to play

Loop On/Off Activate/

Deactivate the loop

Record (Real-Time) (this mode allows you to record an audio file real-time while
you modify the different parameters)

F0Max Maximum fondamental frequency adjustment knob

Settings (Setup the quality / options of the SuperVP audio engine)

Sonogram Settings (to configure the visualization options of the sonogram)

Export Audio (Simple Bounce)

13
Chapitre 3

EXPERT MODE

1- Create a Loop

2- Export Audio and Loop

3- MIDI Remote
Section 1

CREATE A LOOP

CREATE A LOOP 1. Click and drag to select a loop region

1. Click and Drag to select a loop region
Simply click on the waveform and drag the mouse to create a loop region
2. Activate the loop mode

3. Reset loop region

You can also :

- click on a border of the selection to extend or schrink an end point.

- click inside the grey border and drag the entire region.

15
2. Activate the loop mode

Clicking the Loop button in the toolbar will activate the loop
region.

3. Reset the loop region

Double click on the time ruler section (where the time is

displayed) to erase and deactivate the current loop region.

16
Section 2

EXPORT AUDIO

EXPORT AUDIO 1. Configure the export

1. Configure the export

2. Bounce Audio

3. Recording a Real Time Performance

Click on Menu > Preferences

- Select the Export tab in the Audio Settings window

- Select either AIFF or WAV file format for export

- Select the bit depth (16 or 24 or 32 bit available)

2. Bounce Audio

The Bounce audio button, In the toolbar will bounce directly the source
transformations offline statically. This is the faster way to export your work if

17
only intend to apply a constant transformation. The start and end
of the region to export can be set using the loop region.

3. Recording a Real Time Performance

The real-time record button will allow to record on the fly

the transformations you dynamically apply to your sound.

Just press the button when you want to record, play, change the
sliders and when you’re finished with it then a dialog will ask you
where to put the file and how to name it. This is as simple as
that !

18
Section 3

MIDI REMOTE

MIDI REMOTE 1. Connect your midi remote interface

1. Connect your midi remote interface
First verify that your interface or your midi remote interface is currently
2. Open MIDI tab in the preferences panel connected to your computer either by USB or Firewire depending on
3. Monitoring MIDI messages your model.

4. MIDI Learn 2. Open the MIDI tab in the Preferences panel

19
Ensure that your midi interface appears correctly in the MIDI Input a «red overlay» should appear once it is being edited and move
Ports list. If not, reconnect your interface, relaunch IrcamLab TS one of your controls in your midi remote interface. The «midi
and open the MIDI preferences panel again. controller» number should appear on top of the overlay and
indicate you that it is has been received correctly. Switch and
Once it is visible, do not forget to activate it by click on the «off»
click to the other sliders to configure them.
button. It should after change to a blue «on» button once it is
connected to the software One a slider has been correctly mapped to a midi control then
you can see a clear «green overlay» on the slider.
3. Monitoring MIDI messages
Press Command-M again to finish the midi mapping and start
You can monitor the incoming midi channel messages from your
moving your controls : the corresponding sliders should move
interface and ensure the communication is properly working in
accordingl in IrcamLab TS.
the Midi Input Monitor black screen. If you wish to see the raw
midi messages, just press the «Show raw data» button, usefull to
verify the MIDI protocol information really passed in.

4. MIDI Learn

Close the Preferences panel if it is open and press Command+M

on your keyboard. This will «light on» the remotely controllable
sliders by midi.

A «purple overlay» will mean that no controls are currently

mapped to the fonction. To control it with MIDI, click on the slider,

20
Chapitre 4

IrcamLab TS Settings

1. Audio Transformation Settings

2. Sonogram Visualization Settings

Section 1

AUDIO TRANSFORMATION SETTING

AUDIO SETTINGS 1. Introduction

1. Introduction

2. Quality

3. Source

4. Transient Preservation

The Transformation settings are available through the engine wheel icon in the
IrcamLab TS document’s panel. It will configure internally the IRCAM SuperVP
(super phase vocoder) engine to process the current document. SuperVP is a well
known engine in the professional audio world to be one of the most precise and
accurate engine when it comes to transpose, stretch an audio source or process
vocal identities. Several parameters can be changed that will dramatically change
the result of the transformation

2. Quality

22
You have the option to choose between 3 different quality options 8 different source options are available in IRCAM Lab TS wich
: Low, Medium or High, the last being the default. should give an hint to the engine and apply automatically different
configurations to suit your transformation.
- «Low» : apply the default quality transformation.
a) «Polyphonic» and «Polyphonic dense» options are best suited
- «Medium» : will activate a «source position preservation» wich
for rich sources with multiple instruments or a polyphonic
will preserve the stereo or the multichannel configuration of the
content (ex: a single track in your music library). For the expert,
source signal during the transformation.
the «polyphonic» option will apply a window size of ~0.088
- «High» : activate the «source position preservation» and the seconds and the «dense» one a window size of ~0.12 seconds.
«complete sinusoidal models» for sinusoidal remixing option
b) The «Monophonic» options are best suited for monophonic
(see noise/sinus/transient remix panel) wich is an enhanced
sources.
model compared the «Medium» and the «Low» option
c) The «Single Voice» options will activate a special «waveform
3. Source
preservation» option especially suited for voice allows to process
the separate voiced and unvoiced parts of the signal separately
and so, focus on the relevant voiced part.

For both different ranges are available to match as closely as

possible the fundamental frequency of the current audio source :

- low 50 -150 : if the source have a fundamental frequency

between 50 and 150 Hertz. Once activated the F0 max rotary
slider will be set to 150 Hertz. A window of ~0.09 seconds will be
used for the transformation.

- mid 150-450 : if the source have a fundamental frequency

between 150 and 450 Hertz. Once activated the F0 max rotary

23
 slider will be set to 450 Hertz. A window of ~0.033 seconds will
be used for the transformation.

- high 450-1500 : if the source have a fundamental frequency

between 450 and 1500 Hertz. Once activated the F0 max rotary
slider will be set to 150 Hertz. A window of ~0.11 seconds will
be used for the transformation.

NB: by fundamental frequency, we mean, for example in a song

(especially in accapella), the most lowest note, expressed in
frequency) that you can ear.

4. Transient Preservation The «Transients Preservation» option allows to manage the way
transients are rendered when processing a signal.
Transients can be attack or decay transients. Signals can have
IRCAM Lab TS provides an rather simple way to transform your
transients, and noise. These are two different components, which
audio source. For the definitive approach to transform the sound
can be confused at the analysis. In order to evaluate the
and a fully configurable tool on your transformation, IRCAM
respective noise and transient rate in the signal, a statistic model
suggest you to look at the AudioSculpt software package or for
is used.
top and precise vocal gender / transposition transformations to
look at the Flux IrcamTools TraX.

24
Section 2

SONOGRAM VISUALIZATION SETTINGS

SONOGRAM SETTINGS 1. Introduction

1. Introduction

2. Analysis types

3. Window Size

4. Window Step

5. FFT size and Oversampling

6. Settings

Sonograms are visual representations of the spectral contents of a sound : the

distribution of the signal energy over frequency, and how this energy distribution
evolves in time. This section gives an overview of the various types of sonograms
available in IRCAM Lab TS.

Sonograms include 2 types of time/frequency representations:

25
a) Representations of the temporal evolution, (FFT, Reassigned
Spectrum) of the frequency and amplitude of the energy of all
components – sinusoids and noise – of the signal.

b) Temporal representations of the spectral envelope of the signal

(LPC, True Envelope, Discrete Cepstrum). The spectral
envelope is a representation that characterises the spectral
color of the signal. In other words it is a rough characterises the
rough energy distribution neglecting details of individual
sinusoidal components or noise.
The Fast Fourier Transform decomposes the spectrum into a sum
of short sections of windowed sinusoids, and represents the
distribution of the energy over these components. Energy
variations are represented with a contrast scale.
2. Analysis Types

b) Reassigned Spectrum

a) FFT (Fast Fourier Transform)

26
The energy distribution in the spectrum is also calculated with an analysis but it is well know that it is well suited only for analysis of
FFT, but the energy of the individual sinusoidal components, as unvoiced or noise signals. For these kind of signals, the LPC
well as that of the transients is reassigned to increase the analysis reveals the correct filter whenever the filter order is
frequency and time resolution. The reassigned spectrogram is sufficiently high (LPC Order). If applied to voiced signals and even
especially suited for precise analysis of sinusoidal components if the correct filter order is specified the LPC analysis generally
and transients. suffers from systematic errors and the True Envelope method
discussed below should be prefered.

d) True Envelope
c) LPC (Linear Predictive Coding)

The true envelope analysis is one of the most efficient spectral

The LPC analysis is related to a source-filter model of speech
envelope estimation techniques. It is designed to improve the
signals. It estimates the filter of a source filter model using an
results of the LPC. The main advantage is the fact that an
auto regressive filter with a filter order specified by the user. The
appropriate order of the True Envelope analysis can be derived
filter is then displayed as a spectral envelope. This set of
from the fundamental frequency of the sound signal. The true
coefficients is a simplified representation of the resonator filter of
envelope analysis will neglect spectral details with precision
the sound producing system, notably the vocal tract filter for
smaller then the given max frequency. Accordingly, the true
speech. The LPC analysis has been widely used for speech
envelope order is parameterized by the maximum fundamental
27
frequency of a given sound signal. To improve the precision, the 3. Window Size
can setup the time varying fundamental frequency of a sound
signal (Max Fund. Freq) that will then provide optimal analysis
resolution.

e) Discrete Cepstrum
On the left, the Window Size zone displays three text prompts :

• number of samples  : the actual size of the window,

according to the sampling rate of the signal.

• seconds : the duration represented by window, since a

sampling rate corresponds to 1 second.

• fundamental frequency  : the lowest frequency the

analysis can detect, and its frequency resolution.

Each of these three parameters are interdependant. They can be

Like the LPC or True Envelope, the Discrete Cepstrum analysis is calculated according to one of the two others and the sampling
a method to estimate the filter part of a source filter model. It is a rate of the signal. Hence, modifying a parameters implies a
precursor of the true envelope method that is much slower and modification of the others.
does not support the convenient order selection that is used for
3 types of analysis windows are configurable, which have
selecting the order of the true envelope algorithm. It remains to
different windowing curves  : Blackman, Hamming and Hanning.
be present in IrcamLab TS to provide compatibility with the other
The type of curve can improve the quality of the analysis,
Ircam software AudioSculpt but should be replaced by the True
depending on the type of signal.
Envelope.
4. Window Step

28
 sampling frequency. rWith an oversampling factor of 1, AS will
use the closest to the window size power of 2 as an FFT size.
With a 2 oversampling factor, the number of bins is multiplied by
2, etc.

The Window Step parameters allow to determine the interval 6. Extends across sounds limits options
between each successive window. Indeed, for a precise analysis,
This option allows to locate the center of the window on the first
it is necessary to overlap windows.
sound sample, so that the sonogram is displayed from T0 to the
This interval is a factor which can be calculated either  with the end of the signal.
ratio between the window size (adaptive) or an integer adaptative
As the left part of the first window and right part of the last
oversampling with a percentage of the window size (manual)
window are located before the beginning and after the end of the
signal, these parts will be replaced by 0s in the analysis.

5. FFT size and oversampling • If the option is selected, and if the signal shows a
steep beginning or ending, transients and clicks may appear, and
will be preserved when processing treatments.

• If the option is not selected, the attack and release

transients will not appear, and will be replaced by the analysis
window slope
In an FFT, the analysis window is divided into frequency "bins", or
bands. This number of bins is proportional to the Window Size. 7. Settings
The bigger the FFT size, the more bins we get, and the higher the
IrcamLab TS comes with several different settings for tuning
frequency resolution is.
representations for specific instruments (window size, step, etc.).
The Oversampling menu allows to increase the frequency You can select one of the proposed configuration or select one
resolution of the analysis. The signal is sampled with a higher close to the instrument you wish to analyse and modify it.

29