Copyright (C) 2004 by Predrag Viceic viceic@net2000.ch
This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
- INSTALLATION
*In order to generate makefiles and compile you should execute
qmake freecycle.pro
make
in the freecycle dir.
- FOR CONDITIONAL COMPILES PLEASE EDIT src/cond.pri
- You should install Soundtouch 1.3.0 or you will have namespace problems when compiling Freecycle.
- Before compiling Portmidi, you should change the following method in in pmlinuxalsa.c in order to enable runtime midi interface discovery (and to avoid a memoryleak):
----------8<-------------8<-----------8<-----------
void pm_linuxalsa_term(void)
{
snd_seq_close(seq);
//added
free(descriptors);
descriptors=NULL;
pm_descriptor_index=0;
pm_descriptor_max=0;
//end added
}
----------8<-------------8<-----------8<-----------
- In order to link libinstpatch with freecycle you have to add
#ifdef __cplusplus
extern "C" {
#endif
at the begining of libinstpatch.h.in and
#ifdef __cplusplus
}
#endif
at the end. be sure to change the .in file as it's used by the configure script to generate libinstpatch.h
- MANUAL
The manual is available as download at the savannah repository. It's a result of the effort of one person, Mat (@podulator.com), who is the most fervent alpha tester and idea contributor of this project.
COMMAND LINE SWITCHES
--ueber : Launches Freecycle with UeberStyle
KNOWN BUGS:
- Many, but works pretty well with small waves (~30 sec)
- >1 LASPA audio outputs can not be routed to the same audio channel
- if timestretch segfaults, try unlocking all locked beatlines.
- don't timestretch if you don't have to.
- ISSUES
Mandrake10 users which experience compile problems with ueberstyle.cpp should either:
- activate QMotifStyle support in their qt install or
- disable Ueberstyle in Freecycle by removing it in main.cpp and src.pro
- FAQ
-
Q: Could You tell me what is this popup menu with this kind of time signatures
for? How to use it.
A: When the sound signal gets analysed, the bpm computation starts based on
beatline positions. If there are at least 4 locked beatlines, only the locked
beatlines serve as the basis for the bpm computation, If there are less than
4 beatlines locked, the unlocked (green) beatlines are used. The value
obtained is then multiplied by the values in the combo. That gives you the
possibilitiy to change the detected 75BPM to i.e. 150 BPM. The computed value
is then used for:
- Time stretching
- measure marker computation (the bar between two loop points, when dragged)
- bpm lines computation (activated by the clock icon and at least one locked beatline)
- midi file generation
A:The detected bpm value is multiplied by the value in the combo. i.e. 150BPM x 1/2 =75BPM
Q: Is it possible to stretch one beat in a whole loop?
- For now, time stretching works on the whole sound file.
- What are the different analisys methods best for?
- Well that's a short question with a huge answer. :)
The first (default) algorithm (time domain) detects beats by comparing the instant energy (nb of samples configured as instant energy buffer size in Configuration->Configure Amplitude Domain) with the average energy (nb of samples configured as avergage energy buffer size in Configuration->Configure Amplitude Domain). If the difference exceeds some treshold computed internaly and based on the variance of the signal, then the beatline is put at that position (well it's a little bit more complicated because Freecycle then does zero-crossing, but you should have got the point..)
The second algorithm (frequency domain) does the same thing but the signal is splitted in multiple bands and the algorithms compare the instant energy of each frequency band with the average energy of that band. That gives you the positibility to beatmatch the signal with continuous amplitude but varying frequency (flute playing legato notes). Configuration->Configure FFT-> Beatmatching FFT gives you the parameters for this algorithm. Window is equivalent to the [instant energy buffer size]. Average energy is hardcoded as 44032 frames.
If you want to know more about the above two algorithms please read the following [http://www.yov408.com/beat/BeatDetectionAlgorithms.pdf] as it served as the basis for these two algorithms.
The aubio algorithms are well known among DSP experts, which I'm not. Please search on the web if you want some precisions about those. Feel free to post your findings on freecycle-users :)
Q: What is the fft configuration and configure amplitude domain menus for and how to set it for the best (depending on cpu and memory) program performance?
- Generaly speaking slower->better and faster->less precise. Experiment! The default values are ok on a 2Ghz CPU and 1Gb RAM. But this depends on the size of your soundfile, so if you keep those relatively small it should be ok even on less recent computers.
You can see the effect of fft configuration parameters in the spectrogram view (View->Spectrum). Spectrogram FFT parameters are used for drawing the spectrum. If you link those parameters to the Beatmnatching FFT, you cans see what the Frequency domain analysis algorithm "sees".
Enjoy,
Predrag Viceic