Recent Changes

What's new in 2.6.0?

• A lookahead limiter has been added, see synth.limiter.active and other related limiter settings
• Support for 24bit and 32bit audio has been added, see fluid_synth_write_s24() and fluid_synth_write_s32()
• Added fluid_voice_set_callback() for voice lifecycle notifications

What's new in 2.5.4?

• By default, fluidsynth now auto selects the TCP port for the shell server, see shell.port

What's new in 2.5.0?

• FLUID_MOD_SIN is now deprecated, use the newly added fluid_mod_set_custom_mapping()
• A new mode for the custom IIR filter has been added: FLUID_IIR_BEANLAND
• fluid_mod_get_transform() now receives a const argument
• fluid_version_str() now returns a const char array
• An API for manipulating fluid_sfont_t specific default modulators has been added: fluid_sfont_get_default_mod() and fluid_sfont_set_default_mod()
• Support for specifying a synth-wide mode to interpret portamento time has been added, see fluid_portamento_time_mode, fluid_synth_get_portamento_time_mode(), fluid_synth_set_portamento_time_mode() and setting synth.portamento-time
• In all previous versions of fluidsynth, the synth's API mutex was unlocked too early when calls to fluid_synth_unset_program() and fluid_synth_alloc_voice() had been made; this race condition has been fixed

What's new in 2.4.5?

• In order to use the sdl3 audio driver, the downstream application is responsible for calling SDL_Init() and SDL_Quit(), just like it was practice for the sdl2 audio driver. Fluidsynth may raise a warning if this isn't done, see Creating the audio driver

What's new in 2.4.3?

• synth.note-cut has been introduced

What's new in 2.4.0?

• synth.device-id now has a default value of 16
• Default values of reverb and chorus settings have been tuned
• fluid_mod_get_transform() and fluid_mod_set_transform() for SoundFont 2.04 compliant modulators

What's new in 2.3.2?

• fluid_player_get_division() has been added

What's new in 2.2.7?

• Most getter functions of the MIDI event API are now const correct
• fluid_event_from_midi_event() has been added

What's new in 2.2.0?

• fluid_file_callbacks_t now uses long long as file-offset type (see fluid_long_long_t).This is a breaking change, which allows to load SoundFonts bigger than 2GiB on Windows. This change required to bump fluidsynth's SOVERSION.
• various fluid_event_*() functions that received a "value" argument of type short now receive an int argument in preparation for MIDI 2.0 support
• fluid_event_any_control_change() has been removed

• the sequencer has received a major revisal. For you that means:

  • the sequencer's queue no longer blocks the synthesizer thread, due to being busy arranging its events internally.
  • events that share the same tick was given a new, documented order, see fluid_sequencer_send_at().
  • the sequencer's scale can now be used for arbitrary tempo changes. Previously, the scale of the sequencer was limited to 1000. The only limitation now is >0.
  • there is now a dedicated event type for changing the sequencer's time scale, see fluid_event_scale().

• the dynamic-sample-loader has learned support to pin samples, see fluid_synth_pin_preset() and fluid_synth_unpin_preset()

• added getter and setter functions for individual effect groups
• support for UTF-8 filenames under Windows, see fluid_synth_sfload()
• MIDI Tempo of the fluid_player can now be overridden, see fluid_player_set_tempo()
• a per-tick callback has been added to the MIDI player, see fluid_player_set_tick_callback()
• WASAPI audio driver is now available on Windows
• the following drivers have gained support for new_fluid_audio_driver2():
  • DSound
  • WaveOut
  • WASAPI

What's new in 2.1.7?

• a regression introduced in 2.1.0 prevented chorus from being audible when fluid_synth_process() was used

What's new in 2.1.4?

• a regression introduced in 2.1.3 broke fluid_synth_start() for DLS presets

What's new in 2.1.1?

• requirements for explicit sequencer client unregistering have been relaxed: delete_fluid_sequencer() now correctly frees any registered sequencer clients (clients can still be explicitly unregistered)
• using the sequencer with the system timer as timing source has been deprecated

What's new in 2.1.0?

• refrain from using fluid_synth_set_sample_rate()
• synth.sample-rate is no real-time setting anymore, see note about fluid_synth_set_sample_rate()
• new reverb engine
• chorus is now stereophonic
• smallest allowed chorus speed is now 0.1 Hz (previously 0.29 Hz)
• the following audio drivers were added:
  • opensles
  • oboe
  • sdl2
  • waveout

What's new in 2.0.8?

• fluid_sample_set_sound_data() caused broken sound when copying sample data

What's new in 2.0.7?

• fluid_free() has been added to allow proper deallocation by programming languages other than C/C++

What's new in 2.0.6?

• the MIDI player did not emit any audio when calling fluid_player_play() after fluid_player_stop()

What's new in 2.0.5?

• fluid_synth_process() omitted audio samples when called with arbitrary sample counts that were not a multiple of fluid_synth_get_internal_bufsize()
• fluid_synth_sfunload() was not releasing sample buffers of SoundFont3 files if synth.dynamic-sample-loading was set to FALSE

What's new in 2.0.3?

• fix incorrect behaviour of fluid_sample_set_sound_data()
• add missing getters for midi events:
  • fluid_midi_event_get_text()
  • fluid_midi_event_get_lyrics()

What's new in 2.0.2?

• fluid_synth_error() has been deprecated, use fluid_set_log_function() to interfere log messages

What's new in 2.0.0?

FluidSynths major version was bumped. The API was reworked, deprecated functions were removed.

Important changes that may not result in a compilation error but may cause your app to misbehave:

• all public fluid_settings_* functions that return an integer which is not meant to be interpreted as bool consistently return either FLUID_OK or FLUID_FAILED
• fluid_settings_setstr() cannot be used to set integer (toggle) settings with "yes" or "no" values anymore. Use fluid_settings_setint() instead, for example: fluid_settings_setint(settings, "synth.reverb.active", 0) instead of fluid_settings_setstr(settings, "synth.reverb.active", "no")
• explicit client unregistering is required for fluid_sequencer_register_client() and fluid_sequencer_register_fluidsynth() (since fluidsynth 2.1.1 not required anymore, but still recommend)
• all public functions consistently receive signed integers for soundfont ids, bank and program numbers
• use unique device names for the audio.portaudio.device setting
• fluid_synth_process() received a new more flexible implementation, but now requires zeroed-out sample buffers

Other changes in FluidSynth 2.0.0 concerning developers:

• all public delete_* functions return void and are safe when called with NULL
• the shell command handler was decoupled internally, as a consequence the param list of new_fluid_server() and new_fluid_cmd_handler() was adapted
• fluid_settings_set* functions no longer silently register unknown settings but return an error instead
• reverb: roomsize is now limited to an upper threshold of 1.0 to avoid exponential volume increase
• rename fluid_mod_new() and fluid_mod_delete() to match naming conventions: new_fluid_mod() and delete_fluid_mod()
• rename chorus getters to match naming conventions: fluid_synth_get_chorus_speed() and fluid_synth_get_chorus_depth()
• fluid_synth_remove_sfont() returns FLUID_OK or FLUID_FAILED
• introduce a separate data type for sequencer client IDs: fluid_seq_id_t
• fluid_get_userconf() has been implemented for Windows

New Features and API additions:

• add midi.autoconnect a setting for automatically connecting fluidsynth to available MIDI input ports
• add synth.overflow.important and synth.overflow.important-channels settings to take midi channels during overflow calculation into account that are considered to be "important"
• add synth.dynamic-sample-loading a setting for enabling on demand sample loading
• add support for polyphonic key pressure events, see fluid_event_key_pressure() and fluid_synth_key_pressure()
• add fluid_synth_add_default_mod() and fluid_synth_remove_default_mod() for manipulating default modulators
• add individual reverb setters: fluid_synth_set_reverb_roomsize(), fluid_synth_set_reverb_damp(), fluid_synth_set_reverb_width(), fluid_synth_set_reverb_level()
• add individual chorus setters: fluid_synth_set_chorus_nr(), fluid_synth_set_chorus_level(), fluid_synth_set_chorus_speed(), fluid_synth_set_chorus_depth(), fluid_synth_set_chorus_type()
• add realtime settings for synth.reverb.damp and synth.chorus.depth parameters
• add seek support to midi-player, see fluid_player_seek()
• expose functions to manipulate the ladspa effects unit (see ladspa.h)
• add support for text and lyrics midi events, see fluid_midi_event_set_lyrics() and fluid_midi_event_set_text()

• complete rewrite of the soundfont loader API, settingsee sfont.h

  • support for 24 bit audio samples, see fluid_sample_set_sound_data()
  • expose new_fluid_defsfloader() to support loading soundfonts from memory, see fluid_sfloader_set_callbacks() and fluidsynth_sfload_mem.c
  • remove these structs from the public API and provide proper getter and setter functions instead:
  • struct _fluid_sfloader_t
  • struct _fluid_sample_t
  • struct _fluid_sfont_t
  • struct _fluid_preset_t

• add an additional general-purpose IIR filter, see fluid_synth_set_custom_filter()

• add a custom sinusoidal modulator mapping function, see FLUID_MOD_SIN
• implement polymono support according to MIDI specs:
  • add basic channel support, see fluid_synth_reset_basic_channel(), fluid_synth_set_basic_channel(), fluid_synth_get_basic_channel()
  • implement MIDI modes Omni On, Omni Off, Poly, Mono, see fluid_basic_channel_modes
  • implement portamento control, see fluid_synth_set_portamento_mode(), fluid_synth_get_portamento_mode()
  • implement legato control, see fluid_synth_set_legato_mode(), fluid_synth_get_legato_mode()
  • implement breath control, see fluid_synth_set_breath_mode(), fluid_synth_get_breath_mode()

API cleanups:

• the ramsfont has been removed, because it is unmaintained and believed to be unused; please get in touch with the mailing list if you still need it
• remove deprecated fluid_synth_get_channel_info() in favour of fluid_synth_get_program() and fluid_synth_get_channel_preset()
• remove deprecated fluid_settings_getstr()
• remove deprecated fluid_synth_set_midi_router(), instead supply the midi-router instance when creating a command handler with new_fluid_cmd_handler()
• remove deprecated fluid_get_hinstance() and fluid_set_hinstance() (dsound driver now uses the desktop window)
• remove deprecated fluid_synth_create_key_tuning(), use fluid_synth_activate_key_tuning(synth, bank, prog, name, pitch, FALSE) instead
• remove deprecated fluid_synth_create_octave_tuning(), use fluid_synth_activate_octave_tuning(synth, bank, prog, name, pitch, FALSE) instead
• remove deprecated fluid_synth_select_tuning(), use fluid_synth_activate_tuning(synth, chan, bank, prog, FALSE) instead
• remove deprecated fluid_synth_reset_tuning(), use fluid_synth_deactivate_tuning(synth, chan, FALSE) instead
• remove deprecated FLUID_HINT_INTEGER
• remove deprecated fluid_synth_set_gen2() as there doesn't seem to be a use case for absolute generator values
• remove deprecated "synth.parallel-render" setting
• remove obsolete "audio.[out|in]put-channels" settings
• remove unimplemented "synth.dump" setting
• remove fluid_cmd_handler_register() and fluid_cmd_handler_unregister() from public API, as they seem to be unused downstream
• remove misspelled FLUID_SEQ_PITCHWHHELSENS macro
• remove struct _fluid_mod_t from public API, use the getters and setters of mod.h instead
• remove struct _fluid_gen_t, fluid_gen_set_default_values() and enum fluid_gen_flags from public API
• remove macros fluid_sfont_get_id() and fluid_sample_refcount() from public API
• remove FLUID_NUM_MOD macro from public API
• remove the following deprecated enum values from public API:
  • GEN_LAST
  • LAST_LOG_LEVEL
  • FLUID_SEQ_LASTEVENT
  • FLUID_MIDI_ROUTER_RULE_COUNT