Releases: sfztools/sfizz
Releases · sfztools/sfizz
Release 1.2.3
Added
- Support for curvecc opcodes on v1 EGs (@PythonBlue)
- Support for
lotimer/hitimer
(@essej)
Changed
- Allow polyphonic aftertouch (cc 130) to respect the note number, which improves e.g. the choking logic (@essej)
Fixed
- Adjusted the gain compensation on pan stages (@cvde)
- Ensure that voices are cleaned up before being force-reused (@iv-m)
- Fixed a bug in the wavpack wrapper for multichannel files (@KKQ-KKQ)
- Fixed a bug in the way the tuning root was computed (@KKQ-KKQ)
- Fixed a bug where samples would be loaded in RAM twice when using
hint_ram_based=1
- Fixed a bug where quasi-simultaneous notes would not properly choke each other
- Don't send a note off event if a note was choked by a CC event.
1.2.2
Added
- WavPack support (#1129 by @lucianoiam)
Changed
- The project has been split into 2 repositories to separate
the library from the UI/plugins. (#1162) - Unclamp the cutoff and add OSC queries for the cutoff modifiers (#1100 by @PythonBlue)
- SIMDe update (#1161 #1163)
- Update cpuid, dr_libs and stb_vorbis libraries (#1164)
- Various CMake and CI changes/fixes, also because the repository split
Minimum CMake version required is now 3.15 for Windows, 3.13 for other systems - Configurable amount of Control Change messages via CMake'
MIDI_CC_COUNT
variable,
Config.h is now a generated file - Update Abseil to 20230125.3 (#1169)
Fixed
- Crash fix for reloading SFZ: needed to copy preloadCallCount in FilePool
copy constructors (#1170 by @essej) - Temporary fixes to build benchmarks and tests (see #1148)
- Fixed FlexEGs shape (#1176 #1177)
- Fixed build on Windows
- Fixed modulation matrix when CC modulations are per voice (#1173 #1179 by @PythonBlue)
- Fixed and reenabled tests
- Modifications to support universal x86_64/ARM builds on macOS (#1183 by @essej)
Removed
- VCPKG and MinGW CI builds
Release 1.2.1
- Mac OS Installer (#93 #1087)
- Updated bundled dependencies and added more options to use those
preinstalled in system (#1104 #1117 #1143 #1145 #1149) - CI: Various updates and fixes for GitHub Actions; removed Travis and Appveyor
- Add LFO frequency as an EG target (#1103)
- New
image_controls
opcode to add a background image on UI' Controls tab (#1156) - Disabled temporarily the embedded sample test
- Parse 0-valued note ons as note offs (#1072 #1073)
- Correct a bug with dynamic updates on offed EGs (#2 #1088)
- jack: Wait for CLI thread to finish on quit (#1097)
- Fix libsndfile build and add CI (#1112 #1090)
- Last keyswitches don't change the midi state (#1113 #1074)
- Add section suffix to man pages (#1124)
- HiDPI: Added a Zoom menu in Settings to apply an user' zoom ratio preference
(#1123 #1153) - Colored logo button on mouse hover (#1151)
- Centered, full-size background images container (#1152)
- XP style tooltips (#1156)
- Window size has been changed to be compatible with ARIA banks
(775x335 for image backgrounds) (#1140 #1156) - The CC knobs rotator and title label have a white color
on a semi-transparent background in the controls tab (#1156) - The info tab becomes a dark and semi-transparent overlay dialog above the
Home tab; both will display the same background image when available (#1156) - Handle LV2 scaleFactor supportedOption (#1153)
- Added suffix "-multi" to LV2 and VST3. (#1084)
- Fix LV2 manifest when using PSA (#1114)
- Fix 2 bugs in LV2 and VST3 for loads and stores (#1115 #1098)
1.2.0
The big stuff:
- The sfizz engine now handles multiple stereo outputs, through the
output
opcode. The VST3 and LV2 plugins both have a 16-out version (as 8 stereo outs) (#1002) - The
emscripten
branch allows to run sfizz through WASM for your web-oriented needs. An example with WebMidi is available at https://sfz.tools/sfizz-webaudio/ (requires a Chromium-based browser), with source code at https://github.com/sfztools/sfizz-webaudio. - sfizz can now parse embedded samples in base64 data (#1041)
Smaller things:
- Added support in the plugin GUI/parameters for the "sustain cancels release" behavior of LinuxSampler.
- The internal midi state is not reset when loading/reloading. This means in particular that controls changed on an instrument will be kept if you edit the underlying SFZ file (#1002)
- Updated dependent libraries (#1018)
- Support extended CCs 140 and 141 for "keydelta" (#1003)
- Negative values for lokey will not disable the region anymore (#1019)
- Updated the GUI option panel for the plugins (#1022) and some hints on the tab buttons (#1057)
- Choke groups now act through CC switches (#1024)
- Corrected a bug when reloading some files (#1028)
- Added better support for files that are fully loaded in memory and not simply preloaded (#1031)
- Corrected mistakes in the shelving filters (#1027)
- Corrected a crash in some VST3 hosts (#1035)
- Properly read exponential notation in floats (#1036)
- sfizz is now clamping flex EG levels (#1037)
sw_last
is now properly affected byoctave_offset
andnote_offset
(#1039)- Corrected a bug where long-tailed effects would be cut off (#1050)
- Corrected a bug in the band rejection filters (#1052). The
apf_1p
filter still misbehaves and is disabled for now. - Numpad input is now properly handled (#1053)
- In the LV2 plugin, plugin-side automation is now enabled by default (which means all SFZ control changes are naturally bound to incoming MIDI control changes and not controlled by a plugin parameter). You can disable this at build time and show the relevant parameters using
SFIZZ_LV2_PSA
. Too many hosts expect their plugins to respond to midi messages for things like pedal and volume. (#1054) - Added support for
loprog/hiprog
(#1058) - The LV2 plugin now has a string output parameter with the CC used by the instrument (#1060)
API changes:
sfizz_enable_logging
andsfizz_disable_logging
(and their C++ counterparts) are deprecated, replaced bysfizz_get_callback_breakdown
which needs to be called at the end of a callback by the host (#1034). All file writing is handled by the host for logging.- Added
sfizz_send_program_change
and its C++ counterpart. - Added
sfizz_set_sustain_cancels_release
and its C++ counterpart.
1.1.1
- Roll back VSTGUI to 4.10.2 to avoid corruption of the UI in some scenarios
1.1.0
New supports and plugins:
- Added a Puredata plugin (#869 #873 #874)
- Added support for
hint_sustain_cancels_release
, which when activated reproduces a LinuxSampler behavior by sustaining already released notes (#898). A GUI option should come later. - Added support for
amp_veltrack_oncc
,pitch_veltrack_oncc
, andfil_veltrack_oncc
(#938) - Dynamic EGs are supported for flex EGs (#933 #932 #930)
- Support
_curvecc
for extended CCs (#993) - Support
lorand
andhirand
on CC-triggered regions (#997) - The engine now parses
oscillator=auto
and only enables oscillators on short samples with wavetable tags in the WAV metadata (#954)
Bug fixes and other improvements:
- Added a GUI volume meter (#859)
- Implement support for VST3 keyswitch and midi note names (#877 #876)
- Fixed a bug which could prevent parameters to be automated in Ardour (#893)
- Fixed a bug where high definition CC, crossfade or aftertouch events could be ignored (#855 #896)
- Resetting all controllers will now reset them to their default, instead of 0 (#905)
- Filter resonance can now be negative (#919)
- Corrected a bug for the LV2 plugin when used in the Ardour DAW (#924)
- Respect the editor set for sfz files on MacOS (#925)
- The JACK and PureData clients can use custom importers (#926)
- Fix
off_by
behavior with long-releasing notes (#972) - Polyphony groups can have negative indices (#934)
- Parse
lokey=-1
ashikey=-1
(#941) - Improved the preloading for monolithic wave files (#935)
- Since the default build of the LV2 plugin is not statically linked against libsndfile, the distribution license is changed to the ISC license (#943)
- The DecentSampler importer now better translates the loop modes (#981)
- Even when plugin-side automation is disabled (through
SFIZZ_LV2_PSA
), sustain and sostenu CC events will still go through to the plugin to improve the user experience on hosts that do not support automatic mapping of control ports to midi events through the LV2midi:binding
extension (#963). - Fixed a bug that could hang the background thread in the VST plugin (#984 #865 #864)
- The current keyswitch is properly saved and restored with the plugin state (#961)
- It is now possible to enter values for CCs on the plugin GUI by double clicking on the CC knob (#945). A high-resolution input option is also available on a right click.
- Added a text interface to
sfizz_jack
(#973) - CC-triggered voices can be cut off by polyphony groups (#917)
- Allow building on MacOS 10.9 (#990)
- Improved the X11 runloop (#986)
- Added a polyphony option to
sfizz_render
(#992) - Corrected a bug which prevented VST3 or LV2 plugins to be loaded by the Ardour DAW (#995)
- Corrected a bug where octave and note offset were not reset upon loading a new file (#994)
- Corrected a bug where CC-triggered voices could generate note-offs when cut by a polyphony group (#998)
- Added man pages for
sfizz_render
andsfizz_jack
(#999) - Various build system improvements for different platforms (#957 #958 #964 #967 #968 #989 #978 #942 #921 #920 #902 #849 #850 #856 #1001)
1.0.0
We would like to thank to our contributors at https://opencollective.com/sfztools, as well as everyone who is regularly testing and pushing sfizz to the limits! Your support to this project in any form is much appreciated.
You can download builds catered for your Linux distribution over at OBS. The Mac, Windows and MOD packages are below in the release assets, as well as the source packages.
Headlines:
- SFZ v1 is virtually supported except for a handful of opcodes ! Please check
https://sfz.tools/sfizz/development/status/ for the up-to-date status of opcode support. - It is now possible to build sfizz without relying on libsndfile, using a set of libraries. This is now the
default build mode. Building with libsndfile can be enabled at configure time. - The library and plugins can now load DecentSampler files, and could accomodate other formats.
- CCs, keyswitch range, key ranges and active keyswitch are now displayed in the editor for all plugins.
There has been a lot of UI work to make it more practical to use. - There is an OSC interface in the library, which allows to have introspection into the currently
loaded file, the state of CCs/keyswitches, and also set some parameters for loaded regions.
API/ABI changes for the library:
- The ABI is broken, leading to a major version change.
- Thread-safety now has to be handled externally. The API has been updated to provide a concurrency spec.
- The current version of the internal oversampling factor was too unstable and has been disabled for the time
being. The API and plugin parameters are staying for compatibility. A more robust implementation should come soon. - Added documentation in the API calls to make clear that sample-accurate triggering of voice requires
messages to be sent in order. The VST plugin in particular did not respect this and has been updated. - A new
bpmTempo
method has been added to pass tempo as beats-per-minute rather than seconds per beat.
The previous version is deprecated. - There are new HD (float) versions of API calls for note and CC events.
aftertouch
is renamed aschannel_aftertouch
(for C) andchannelAftertouch
(for C++) throughout the API
to be more consistent withpolyAftertouch
.
UI:
- UI improvements and tweaks, adding themability in particular (#826 #824 #822 #821 #819 #818 #816 #812 #799 #743 #691 #674 #673 #672 #670 #662 #661 #656)
- Added an info panel in the plugin UI (#793 #792 #791 #789 #788 #787 #782 #779 #773)
- Improve font handling (#669 #622)
- Improve file handling in the UI (#645 #659 #658) and enable the use of a "default user directory" or environment variable.
This also allows the plugin to "find" missing sfz files in a different environment (#644 #581 #572 #532) - Support DecentSampler and a generic importer for other formats (#725 #715 #680)
- Show the current keyswitch on the editor and the virtual keybnoard (#665 #657)
- Added a button to reset the scala file to default (#660)
- The plugins now display an image through the
image
opcode (#770)
Plugins:
- Remove automation from un-automatable parameters in the VST plugin (#814)
- Order VST events (#810 #803)
- Added the option to build a VST2 version of sfizz if you have the SDK available (#708)
- The C++ standard is set to 14 if you build the UI on LV2, or if you build the VST version. This will likely
increase to 17 shortly due to the dependency on VSTGUI. The ui-less LV2 plugin and the library remain C++11
compatible. Note that windows builds already use C++17 on all targets. (#795) - Fix corruptions and crashes that happened in the LV2 plugin (#794 #668 #557)
- The internal controllers of sfizz are now presented as parameters in the LV2 plugin. Experimental midi binding
is also active on them. By default the "direct" midi input is disabled and you need to map to the parameters.
A compile-time option is available to re-enable the midi input. (#776) - Fix crashes in the VST plugin (#709 #702 #681 #625)
- Enable checking if file changed even when the transport is not running (#712 #689)
- The Mac builds are now signed (#550 #552)
- Update VST to 3.7.2 (#798)
Library:
- Small files with only zeroes are considered as
*silence
(#831) - The library uses an Hermite interpolation by default (#828 #829)
- Fix an error where sample indices could wrap for massive pitch modulations (#825 #751)
- Add high-definition versions of the API calls (#820)
- Corrected a bug where regions with long release envelopes would not stop on sample ending (#811)
- Support polyphonic aftertouch events, triggers and targets (#809 #764)
- Support monophonic aftertouch events, triggers and targets (#765 #630 )
- Corrected a regression where sfizz would stop loading the SFZ file if some sample files cannot be loaded (#806)
- Update abseil and ghc::filesystem (#780), the tunings library from Surge (#785), catch2 (#711)
- It is now possible to change some parameters of the regions after loading the file through the OSC interface (#761)
- Improve the choking logic to match the spec and other implementations (#778)
- Added modulation targets for amplitude, pan, width, position, pitch and volume (#760)
- Support CC modifiers for
end
,loop_start
andloop_end
(#752) - Support CC modifiers for LFO targets (#753)
- Support SFZ v1 extended CCs (#747)
- Implement the sostenuto logic (#739)
- Refactored opcode reading and storing (#748 #727 #721 #722 #720 #700 #559)
- Fix the parser to support sharp (#) symbols in file names (#746)
- Improved the ADSR transitions from decay to sustain, and from release to 0 (#734 #373)
- Reset the smoothers if the playhead moved (#731 #710 #382)
- Corrected a bug where the first sample was ignored on playback (#732 #394)
- Added CC modulation for
depth
targets (#730) - If the sustain level is low enough the envelope is set to free-run (#723, #676)
- Fix loop crossfades when the loop starts early in the sample (#718)
- Fix random generators to match ARIA's behavior (#719)
- Fix corruption appearing in some sample libraries (#717 #695)
- Smoothers are now linear (#713 #714 #703)
- Filter/EQ parameters are clamped even with modulation (#701 #699)
- Implement v1 LFOs (#693)
- Support
count
andloop_count
(#654) - Fix a bug where the sample rate was not propagated properly to the flex EGs (#683)
- Delay envelope start by the voice delay (#655 #432)
- Support
delay_oncc
(#653) - Support
sw_vel
(#650) - Fix
note_polyphony
choking (#634 #529) - Respect the
end
opcode (#618) - Fix allocations that may happen on the RT thread (#621)
- Fix the polyphony manager when the engine polyphony is changed after a file is loaded (#619)
- Avoid reloading invalid files (#614)
- Allow parsing of some badly formed files files (#600)
- Match ARIA's LFOs (#613)
- Added windowed sinc interpolation for
sample_quality
from 3 to 10 (#609) - Improve the Hermite interpolation (#597)
- Update AtomicQueue to correct a bug (#583)
- Support
lfoN_beats
(#553) - Add region and voice introspection to the OSC interface (#555)
- Enable fast-math on MSVC (#567)
- Fix crashes when building with libsndfile (#542 #543)
- Properly pad and align the wav files upon reading (#538)
- Add an alternative to libsndfile (#482 #534 #535 #536)
- Handle gracefully having a different
sw_default
in a region (#531) - Support disjoint keyswitch ranges, as well as
sw_lolast
andsw_hilast
(#526)
0.5.1
- Corrected race conditions that appeared with the new thread and file pools (#507 #508 #514 #521)
- Take the internal oversampling factor into account for loop points, and solve an issue where loop points specified in sfz files were not taken into account (#506)
- Fix an implementation error for the internal hash function when applied on a single byte (#512)
- Knobs are linear in the AU plugin (#517)
- Fix a crash in VSTGUI (#520)
- Fix the resource path in the LV2 plugin under windows (#524)
- Add MacOS make install rules (#525)
0.5.0
Big stuff:
- Added basic support for Flex EGs (#388) as modulation sources (targets to come)
- Added basic support for LFOs (#338) as modulation sources (targets to come)
- EGs and LFOs can now target EQs and filters (#424)
- A new GUI has been added and is common to the LV2 and VST plugin (#397 #404 #419 #489 #492 #495 #496 #497); still mostly work in progress, more to come!
- Provided build systems to use sfizz with the VCV-Rack SDK and the DISTRHO Plugin Framework
New features:
- Added support for
sustain_lo
(#327) - Audio files are now read incrementally, improving the availability under load (#294)
- A new output port for active voices has been added in the LV2 plugin (#321)
- Added support for effect types
reverb
,disto
,gate
andcomp
- The voice stealing is now configurable using
hint_stealing
, with possible values asfirst
,oldest
(default), andenvelope_and_age
. The latter is the more CPU-consuming version which requires to follow the envelope of each voice to kill low-volume ones preferably. Note that the voice stealing continue to kill all voices started on the same event by default (i.e. all layers of the same note). (#344 #384 #353) sfizz
now internally uses a modulation matrix to link all modulation sources (CC, LFOs, and EGs) and targets (#335 #351 #386)- Added support for
off_time
and complete support foroff_mode
. The voice stealing logic was improved to take into accountpolyphony
,note_polyphony
andgroup_polyphony
properly (#349 #352 #393 #413 #414 #467). Note that this support is also available for the engine polyphony. In this case, some additional voice will take over for the release duration (#477). - The wavetable quality has been improved (#347)
- Support for
offset_cc
(#385) sfizz_render
now supports a--quality
switch, which acts like thesample_quality=
opcode (#408)pitch_keycenter=sample
is now taken into account (#362)- Support
oscillator_detunecc
(#434) - Support basic FM synthesis for oscillator regions (#436)
- CC 7, 10 and 11 are now linked by default to pan, volume and expression (#475)
- Support
hint_ram_loading
for loading the whole samples in RAM (#353) - Support for
loop_crossfade
(#464) - All phase-related opcodes in sfizz now use the 0->1 convention, as does ARIA/Sforzando, instead of the 0->360 convention (#499)
Issues:
- Loading probable wavetable files, or wav files containing wavetable metadata now sets
oscillator=on
on the region (#431) - The default
sample_quality
was put back to 1 for live playing and 2 for freewheeling (#405) - Fix an unwanted copy in the realtime thread (#334)
- Improve the filter shortcut path (#336)
- Fix the default
ampeg_attack
andampeg_release
to avoid clicks (#437) - Corrected a race condition in freewheeling mode (#500)
- Fixed a potential non-realtime operation in the realtime thread (#498)
- Fix a bug when using a larger internal oversampling for regions with an
offset
value (#469) - Fix an issue when loops occured more than once in a block (#462)
- Increase the range of the clamping on amplitude (#468) and pitch (#474)
- Fix CC modulations with their source depth set to 0 (#475)
- Fix an overshoot for crazily large cutoff values (#478); cutoffs are now clamped
- Improve the file loading logic to keep files in memory for a short while in case they get reused (#425)
- Fix the MIDNAM output for the case where extended CCs are used (#420)
- Fixed a bug where release voices where not ignored on self-mask search (#348)
- Improved the release logic in many cases (#324 #414 #413)
- Set the level of the
*noise
generator to match ARIA's (#429) - Support for
atom:Blank
atoms in the LV2 plugin (#363) - Fixed
amp_veltrack
behavior (#371) - Fix the ADSRH envelope release rate (#376)
- Fixed an error for files where the loop spans the entire file (#378)
- Fixed
sustain_cc
behavior (#377) - Match the default volumes with ARIA (#381)
- Properly set the
loop_mode
for release regions (#379) - Regions with
end=0
are now properly disabled (#380) - Fix
fil_random
to be bipolar (#452) - The
sequence
order now properly starts at 1 (#451) - Fix an issue on Flush to Zero on some ARM platforms (#455)
- Fix
pitch_veltrack
(#461) - Opcode values now properly stop at the
<
character (#439) - Fix various build errors and issues on all platforms (#345 #401 #400 #399 #417 #447 #449 #443 #453 #456 #459 #471 #484 #487 #488 #491)
- The file dialog initial directory is now the root of the current loaded file (#428)
- Existing and known CC values are now correctly taken into account for modulations (#421)
- Fix various performance regressions and improved performance especially on ARM builds (#410 #412 #415 #426)
API changes:
- Added API support for setting the playback state, time position and signature (#354)
- The API documentation on the sfizz's website has been streamlined alot !
0.4.0
Big stuff:
- Added support for polynomial resamples and
sample_quality
opcodes (#238 #267). The engine now defaults to a value of2
for this opcode, which is more intensive than the original linear interpolation resampler but provides a better quality. Added support for better resampling algorithms also in the wavetables viaoscillator_quality
(#287). - Support
_curvecc
and_stepcc
opcodes (#166 #155 #77) as well as_smoothcc
opcodes (#181 #48 #22 #153 #297 #285) - Added support and API for Scala tuning files in the engine and the plugins (#253 #268 #282)
Other new features:
- Added support for unison oscillators (#161)
- Support for the
polyphony
opcode at all levels (#171 #275), as well asnote_polyphony
. Thegroup=
polyphony is also more flexible and can be defined anywhere. - Added support for
offset_cc
(#170 #159) - Added support for
direction=reverse
(#185 #179) - Added support to label the keys using a
label_key
opcode. This is not really standard yet, but it is now integrated in the LV2 plugin to advertise the names in the MIDNAM file and possibly change their labels in hosts that support it. (#174 #154) - Added support for block comments
/* */
in the parser (#196 #195) - Added a
sfizz_render
client in tree; you can build it with the make targetsfizz_render
if theSFIZZ_RENDER
CMake variable is set toON
. (#200 #201 #206) - Add support to integrate sfizz in DPF plugins (#216)
- Added an AudioUnit target (#224)
- Added support for the
set_hdcc
opcodes and overall added the ability to support floating-point CCs from the API (#233 #232 #244) - Added support for FLAC loops (#242 #229)
- Support the
mapPath
feature of the LV2 specifications, for tentatively better portability in plugin states (#303) - New instances of the sfizz LV2 plugin will now load a default
*sine
instrument (#283)
Issues:
- Solved some issues with DSmolken's drumkits related to the ampeg envelope (#172)
- An exception problem was thrown if an sfz file was deleted (#182 #184)
- Properly bundle the
dylib
for macOS (#188) - Improved the filter stability (#198 #199 #210)
- Handle
USE_LIBCPP
properly on configure (#203) - Fix the handling of loop markers if sample
end=
is present (#202 #204) - Handle note on with 0 velocity as note offs in the jack client (#208 #211)
- Solved an issue with super short files (#215)
- Corrected a stack smashing bug in the LV2 plugin (#226)
- Fixed some parsing issues with
$variables
(#230) - Properly advertise the VST plugin parameters (#241)
- Process
$
expansions in#include
(#247) - Change the default build type to
RelWithDebInfo
(#249) - Improve the note stealing algorithm (#214); note that this is still very much a work in progress since many heuristics are in play here. Feel free to report misbehavior regarding note stealing as we improve this regularly.
- Corrected a bug with SFZ v1
velcurve
(#263) - Properly support the
off_by=-1
opcode to correctly reset the value. (#235) - Corrected some errors with null-terminated atoms in the LV2 plugin (#269)
- Ignore garbage values following e.g. a key number in opcode values (as in
key=64Garbage
->key=64
) (#263) ampeg_****_onccXX
modifiers now properly consider multiple CC modifiers (#300 #167)- Add headers and group sources in the CMake project for integration with e.g. Qt (#312)
- Trigger on CC does not require disabling the key triggering through e.g.
key=-1
(#315) - Support flat notes parsed as string values (#291 #289)
- Improved handling of
release_key
(#298); still not perfect, if the region spans multiple key and multiple notes happened with the pedal down, only a single voice will start. - Properly read the LV2 option list until the end (#323, by @atsushieno)
- Corrected a parsing issue when
$variables
were part of an opcode name (#328) - Various other plumbing changes
API additions:
- Added API calls to set
$variable
define values prior to loading an SFZ file (#168 #119 #130) - Added API calls to get key labels and cc labels defined by
label_key
andlabel_cc
(#174) - Added an API call to load an sfz file as an
std::string
orconst char*
(#217) - Added API calls for Scala files and tunings (#253)
- Added high-definition floating point CC API calls (#244)
- Added API calls to change the default resampling quality (#267 #238)