flutter_soloud

• use of dr_mp3.h instead of minimp3 for streaming MP3
• fix crash when calling addAudioDataStream before play #306

• fix seek a buffer stream actually seeks at half the wanted position #296

• fixed stuttering with MP3 streams #301
• fixed a Web bug when compiling with WASM in release mode.

• fix win: fix build error #292

• Added support for mp3 streams
• Added support for Vorbis streams
• Added web_radio.dart example to demonstrate how to receive an audio stream (ie, an icecast stream) and then add the audio chunks to BufferStream
• Add BufferType.auto to auto detect ogg/opus/vorbis/mp3 streams
• Deprecate BufferType.opus in favor of BufferType.auto
• Get TAGs info also while streaming and not only by sending chunks of an audio file. For MP3s, the TAGs are obtained from ID3V2 or passing icy-metaint (obtained from the header of the online stream) before adding audio chunks to the BufferStream. New metadata is notified by the onMetadata callback of setBufferStream

• feat: allow to specify NO_OPUS_OGG_LIBS in Android build config #282. Thanks to @ekuleshov
• fix: 16KB memory page size for Android because of Play Console warning #283
• fix pub points

• macos fix: abseil not found when using this plugin with some other native plugins like firebase #280. Thanks to @jochy.

• ios fix: abseil not found when using this plugin with some other native plugins like firebase #271

• fix 16 KB native library alignment on Android #248
• fix build issue on iOS and macOS #265 #266
• fix: the seek method was causing inconsistent behavior when using BufferStream in BufferingType.released mode. It is now supported in BufferingType.preserved mode.

• OGG is now also supported using readSamplesFrom* methods.
• fix getPosition and buffering for released buffer.

• fix #104, #245, #249. It is now possible to use a 3rd party plugin like audio_session to manage audio context.
• new audio context example in example/lib/audio_context/audio_context.dart.
• fix GetPosition returned value for buffer streams.
• fix Web hot reload/restart #258 and #259.

• added getStreamTimeConsumed() to get the time consumed by a buffer stream of kind BufferingType.released. Since the position of this kind of stream is always 0, this method is useful to know the time already played.
• fix pause/unpause on audio stream buffering.
• added buffer_stream/simple_noise_stream.dart example.

• fix: Loading the same AudioSource twice (in parallel) crashes #247
• fix win: force cmake to build the plugin in release mode even if building in debug

• fix: Setting pan doesn't cancel pan oscillation #239

• fix: disposeSource crash on Android #240

• fix: adding audio data to an unended BufferStream throws error #235

• docs: clarify docs regarding semitones and shift parameters of the pitchShiftFilter #233 by @bemain
• fix: readSamplesFrom*() works again
• more accurate samples average in readSamplesFrom*

• fix: passing a group handle to seek() throws error #228
• fix: listPlaybackDevices fails to retrieve devices when the device prefix contains Chinese characters #227 by @WHYBBE

• fix: when the speed is changed, after seeking, the getPosition() returns the wrong position #223

• fix building issue with XCode 16.3 #217

• fix listPlaybackDevices on Web #214
• log maxActiveVoiceCountReached exception with Level.INFO #212

• enhanced documentation clarity and organization by moving it to the dedicated flutter_soloud_docs repo. Powered by docs.page from Invertase and can be viewed here.
• Web fix: Uncaught (in promise) TypeError #208
• fix: error when loading very short MP3 files #181

• fix: Sounds seemingly "backed up in a queue" when playing too many at once #204

• when calling AudioData.getAudioData is now possible to check if the audio data is the same as before. Useful to visualize waveforms. This is because AudioData.getAudioData returns the current data in the buffer and if it is called before the buffer has been updated, it will return the previous data.
• better FFT data for a better visualization.
• added resetBufferStream method to SoLoud. It happens that when playing a stream, maybe from the web, it is necessary to change it to another source. The player continues to play the already added audio data to the buffer. This method can be used to reset the buffer and start with the new audio data.

• it's now possible to choose to not link opus and ogg libraries (see NO_OPUS_OGG_LIBS.md). Fix for #191 and #192.

• fixed crash when trying to play a sound after deactivating its active filter #189

• fix: error while calling listPlaybackDevices() #186.
• Android example folder recreated.

• BufferStream now supports 2 type of buffering:
  • BufferingType.preserved (default): preserve the data already in the buffer while playing.
  • BufferingType.released: free the memory of the already played data for longer playback.
• breaking change: splitted [maxBufferSize] to [maxBufferSizeBytes] and [maxBufferSizeDuration] in SoLoud.setBufferStream. This gives the user a way to choose the maximum buffer size using bytes or time.
• breaking change: removed initialized getter in favor of isInitialized
• removed deprecated timeout parameter in SoLoud.init.
• removed deprecated filter_params.dart.
• fixed biquad resonant filter frequency default parameter #179
• fix: on some unclear conditions isInitialized returns false on MacOS after the engine starts with no error #177
• fix: Call loadMem will crash the application #174.

• fix: clicks and pops when changing waveform frequency #156.
• added Limiter and Compressor filters (see example/lib/filters/).
• added BufferStream #148. Now it's possible to add audio data and listen to it. It provides a customizable buffering length which automatically pauses the playing handle if there is not enough data, for example, when receiving audio data from the web. It also provides a callback that allows you to know when the buffering is started and stopped. The audio data can of of the following formats:
  • s8 signed 8 bit
  • s16le signed 16 bit little endian
  • s32le signed 32 bit little endian
  • f32le float 32 bit little endian
  • opus Opus codec compressed audio with Ogg container. Useful for streaming from the Web (ie using OpenAI APIs).
• fixed Web Worker initialization non-fatal error that could occur on Web.
• fixed sound distortion using a single pitchShift filter and changing the relative play speed #154.
• fixed the use of LoadMode.disk on the Web platform, which in some cases caused the allInstancesFinished event to not be emitted.
• improved performance on Web, MacOS and iOS.
• getting wave and FFT samples is now simpler and faster.
• To avoid future incompatibilities when using other WASM compiled plugins, it is now necessary to add a new script to index.html:

  <script src="assets/packages/flutter_soloud/web/libflutter_soloud_plugin.js" defer></script>
  <script src="assets/packages/flutter_soloud/web/init_module.dart.js" defer></script>
  

• added listPlaybackDevices to get all the OS output devices available.
• added deviceId parameter to the init() method. You can choose which device is delegated to output the audio.
• added changeDevice method to change the output playback device on-the-fly.
• fix: now throws when loading a file that might be corrupt #145.

• fixed a bug that caused an error when loading a sound more than twice.

• added readSamplesFrom*() methods to read N audio data within a time range from a file or memory #75. Example in example/lib/wave_data/wave_data.dart.

• fixed waveform generation, which somehow oscillates frequencies after some time #129.
• fixed iOS compilation by rising minimum iOS version to 13 #128.
• fixed iOS compilation on the new MacOS 15 with XCode 16 #130.

• added audio_data example.
• added compatibility for Web platform in the pubspec.
• bug fix when loading multiple audio files asynchronously.
• better error message when something goes wrong loading a file.

• bug fix when loading multiple audio files asynchronously.

• added bool isActive and int index getters to filters.
• added a timeStretch() method to the single pitchshift filter.
• fixed building error on Windows.
• updated examples.

• added support for the Web platform.
• added getPan(), setPan() and setPanAbsolute().
• added loadMem() to read the given audio file bytes buffer (not RAW data). Useful for the Web platform.
• fixed getFilterParamNames().
• added AudioData class to manage audio samples.
• added player initialization parameters: sample rate, buffer size, number of channels (mono, stereo, quad, 5.1, 7.1).
• added voice groups.
• it's now possible to set filters not only globally, but also to single audio sources (not on the web platform).
• fade and oscillate filter parameters.
• experimental capture feature removed.
• now accessing the filter has been simplified with the use of SoLoud.filters and AudioSource.filters to use global and single sound filters.

• Fixed wrong exception raised by setVolume() when a handle is no more valid.

• Fix init error on hot restart.

• A giant leap forward from the previous version (many thanks to Filip Hráček).

• Major changes to API. There are quick fixes (dart fix) to automatically rename many changed APIs.

• SoLoud methods now throw instead of returning a PlayerErrors object.

• added getActiveVoiceCount() to get concurrent sounds that are playing at the moment.

• added countAudioSource() to get concurrent sounds that are playing a specific audio source.

• added getVoiceCount() to get the number of voices the application has told SoLoud to play.

• added getMaxActiveVoiceCount() to get the current maximum active voice count.

• added setMaxActiveVoiceCount() to set the current maximum active voice count.

• added setProtectVoice() and getProtectVoice() to get/set the protect voice flag.

• SoLoud.activeSounds is now an Iterable instead of a List.

• All time-related parameters and return values are now Duration type. Before, they were double.

• Added new (experimental) AudioSource.allInstancesFinished stream. This can be used to more easily await times when it's safe to dispose the sound. For example:

  final source = soloud.loadAsset('...');
  // Wait for the first time all the instances of the sound are finished
  // (finished playing or were stopped with soloud.stop()).
  source.allInstancesFinished.first.then(
    // Dispose of the sound.
    (_) => soloud.disposeSound(source)
  );
  soloud.play(source);
  

• added looping and loopingStartAt properties to SoLoud.play() and SoLoud.play3d().

• added SoLoud.getLooping() to retrieve the looping state of a sound.

• added SoLoud.getLoopPoint() and SoLoud.setLoopPoint() to get and set the looping start position of a sound.

• New methods SoLoud.loadAsset() and SoLoud.loadUrl() to load audio from assets and URLs, respectively.

• added mode property to SoLoud.loadFile() and SoloudTools.loadFrom* to prevent to load the whole audio data into memory:

  • LoadMode.memory by default. Means less CPU, more memory allocated.
  • LoadMode.disk means more CPU, less memory allocated. Lags can occur while seeking MP3s, especially when using a slider.

• Switched from print() logging to using the standard package:logging. See README.md to learn how to capture log messages and how to filter them.

• The capture feature is on experimental stage to be fine-tuned in the near future. All methods related to audio capture have been extracted to a separate class. So now, there are two classes:

  • SoLoud for playing audio
  • SoLoudCapture for capturing audio

• The Web platform is a work in progress, stay tuned!

• Switched LICENSE from Apache-2.0 to MIT.

• getLoopPoint now returns Duration.
• Major changes to API docs and README.
• Renamed SoLoud.disposeSound to SoLoud.disposeSource. Quick fix available.
• Renamed SoLoud.disposeAllSound to SoLoud.disposeAllSources. Quick fix available.
• Removed unused AudioSource.keys property.
• Switched LICENSE from Apache-2.0 to MIT.

• some little fixes.

• added getActiveVoiceCount() to get concurrent sounds that are playing at the moment.

• added countAudioSource() to get concurrent sounds that are playing a specific audio source.

• added getVoiceCount() to get the number of voices the application has told SoLoud to play.

• added getMaxActiveVoiceCount() to get the current maximum active voice count.

• added setMaxActiveVoiceCount() to set the current maximum active voice count.

• added setProtectVoice() and getProtectVoice() to get/set the protect voice flag.

• All time-related parameters and return values are now Duration type. Before, they were double.

• Fixed velocity computation bug in example/.

• Renamed SoundEvent to SoundEventType. Quick fix available.

• SoundProps.soundEvents is now a Stream, not a StreamController

• SoundProps.soundEvents stream is now closed automatically when SoLoud.disposeSound() is called.

• SoLoud.activeSounds is now an Iterable instead of a List (therefore, it cannot be modified from outside the package).

• Renamed SoLoud.getFxParams to SoLoud.getFilterParameter. This mimics the C++ API name. Quick fix available.

• Renamed SoLoud.setFxParams to SoLoud.setFilterParameter. This mimics the C++ API name. Quick fix available.

• Renamed SoundProps to AudioSource. Quick fix available.

• Added new (experimental) AudioSource.allInstancesFinished stream. This can be used to more easily await times when it's safe to dispose the sound. For example:

  final source = soloud.loadAsset('...');
  // Wait for the first time all the instances of the sound are finished
  // (finished playing or were stopped with soloud.stop()).
  source.allInstancesFinished.first.then(
    // Dispose of the sound.
    (_) => soloud.disposeSound(source)
  );
  soloud.play(source);
  

• Deprecated shutdown(). Replaced with the synchronous deinit(). Quick fix available.

• Renamed initialize() to init(), in order to come closer to the original C++ API, and also to have a symmetry (init/deinit). Quick fix available.

NOTE: This version is much more breaking than the ones before it. It might be worth it to first upgrade your code to 2.0.0-pre.1, use the quick fixes to rename the methods, and only then upgrade to 2.0.0-pre.2 and beyond.

• SoLoud methods now throw instead of returning a PlayerErrors object. This is a massive breaking change, but it makes the package API more idiomatic and easier to use.

  Before:

  final ret = await SoLoud.play(sound);
  if (ret.error != PlayerErrors.noError) {
    print('Oh no! ${ret.error}');
  } else {
    print('Playing sound with new handle: ${ret.newHandle}');
  }
  

  After:

  try {
    final handle = await SoLoud.play(sound);
    print('Playing sound with new handle: $handle');
  } on SoLoudException catch (e) {
    print('Oh no! $e');
  }
  

• added looping and loopingStartAt properties to SoLoud.play() and SoLoud.play3d().
• added SoLoud.getLooping() to retrieve the looping state of a sound.
• added SoLoud.getLoopPoint() and SoLoud.setLoopPoint() to get and set the looping start position of a sound.
• New methods SoLoud.loadAsset() and SoLoud.loadUrl() to load audio from assets and URLs, respectively. These replace the old SoloudTools.loadFrom* methods (which are now deprecated).
  • The new methods also correctly invalidate the temporary files (for example, when an asset changes between versions of the app, we don't want to play the old file).
• Rename SoloudTools to SoLoudTools for consistency. (Quick fix available.)
• Rename SoLoudTools.initSounds to SoLoudTools.createNotes for clarity. (Quick fix available.)

• added bool SoLoud.getVisualizationEnabled() to get the current state of the visualization.
• added mode property to SoLoud.loadFile() and SoloudTools.loadFrom* to prevent loading the whole audio data into memory:
  • LoadMode.memory by default. Means less CPU, more memory allocated.
  • LoadMode.disk means more CPU, less memory allocated. Lags can occur while seeking MP3s, especially when using a slider.
• Switched from print() logging to using the standard package:logging. See README.md to learn how to capture log messages and how to filter them.
• Renamed SoLoud.startIsolate() to SoLoud.initialize()
• Renamed SoLoud.stopIsolate() to SoLoud.shutdown()
• Removed SoLoud.initEngine() (it shouldn't be called manually)
• None of the renaming changes are strictly breaking (yet). The old method names still exist as aliases to the new names, and are merely marked @deprecated. There is a quick fix (dart fix) to automatically rename them.
• The singleton SoLoud instance is now accessible through SoLoud.instance. Accessing it through SoLoud() is now deprecated.
  • This change cannot be automated through a Quick Fix. You will need to manually replace SoLoud() with SoLoud.instance in your code.
• All methods related to audio capture have been extracted to a separate class. So now, there are two classes:
  • SoLoud for playing audio
  • SoLoudCapture for capturing audio
• The SoLoud class is now an interface class. This means you can implement it (e.g. for mocking in tests) but you can't extend it. This reduces the fragile base class problem and makes the API easier to evolve.
• Added a new, more usable way of finding out whether the audio engine is initialized and ready to use:
  • SoLoud.initialized (returns a future, safe to check during initialization)
    • This is a much easier way to check engine readiness than subscribing to SoLoud.audioEvents and waiting for the isolateStarted event.
  • SoLoud.isInitialized (returns synchronously)
  • previous methods to check readiness (isPlayerInited and isIsolateRunning()) are now deprecated
• SoLoud.initialize() can now be safely called during engine shutdown. It will wait for the engine to shut down before re-initializing it. Same for SoLoud.shutdown(), which will wait for the engine to initialize before shutting it down, to avoid various race conditions.
• Sound handles and sound hashes are now typed: SoundHandle and SoundHash instead of raw integers. This prevents erroneously passing a sound handle as a sound hash, for example. This is a breaking API change, but, in practice, shouldn't be much of a problem, since these objects were always meant as identifiers (to be taken from some API calls and put into others).
• SoundProps.handle renamed to SoundProps.handles (because it's a Set) and also disallowed modifying it from outside the package.
• All fields of SoundProps marked final. This is a breaking change but unlikely to have an effect (as most users hopefully don't assign to these fields).

• updated mp3, flac and wav decoders
• updated miniaudio to 0.11.21
• fixed the doppler effect in 3D audio example

fixed compilation on Windows

• fixed compilation on iOS and macOS

• waveform example page updated with sound FXs
• added sound FXs
  • biquadResonantFilter
  • eqFilter
  • echoFilter
  • lofiFilter
  • flangerFilter
  • bassboostFilter
  • waveShaperFilter
  • robotizeFilter
  • freeverbFilter

• added waveform generator
• added a test page for waveform
• added some tests in tests dir
• miniaudio updated to v0.11.18

• SoLoud().loadFile now can return PlayerErrors.fileAlreadyLoaded when a sound has already been loaded previously. It still returns the SoundProps sound. It's not a breaking error.
• added Soloud().disposeAllSound to stop and dispose all active sounds

breaking change: Soloud().stopSound has been renamed to Soloud().disposeSound

added load sound tools:

• SoloudLoadingTool.loadFromAssets()
• SoloudLoadingTool.loadFromFile()
• SoloudLoadingTool.loadFromUrl()

added also a spin around example

• added 3D audio with example