- all functions that take an input and output configuration are now consistently input, output. Please double check usages during ugrade.
- The miniaudio and RtAudio backends are no longer selected via cmake options. They are built as linkable libraries, and it is us to the LabSound user to pick one and use it, or to supply a custom backend. To make this easier, the joining of the AudioContext and a backend is now much simpler. The complex node hierarchy around hardware nodes has been reduced to an AudioDevice, from which a backend may be derived, an AudioRenderingNode that is supplied with, and owns an AudioDevice. The AudioRenderingNode is optionally supplied to the context to be pulled by a DestinationNode. Alternatively, the AudioRenderNode can be used directly for offline rendering via its offlineRender function.
- the final output of the graph should now be connected to a DestinationNode rather than the removed DeviceNode. This matches the WebAudio specification.
- see GetDefaultAudioDeviceConfiguration in ExamplesCommon to see one way to obtain input and output configurations. Set up now looks as follows:
AudioStreamConfig _inputConfig;
AudioStreamConfig _outputConfig;
auto config = GetDefaultAudioDeviceConfiguration(true);
_inputConfig = config.first;
_outputConfig = config.second;
std::shared_ptr<lab::AudioDevice_RtAudio> device(new lab::AudioDevice_RtAudio(_inputConfig, _outputConfig));
auto context = std::make_shared<lab::AudioContext>(false, true);
auto destinationNode = std::make_shared<lab::AudioDestinationNode>(*context.get(), device);
device->setDestinationNode(destinationNode);
context->setDestinationNode(destinationNode);- Note the explicit management of the Context, AudioDevice, and the DestinationNode. This separation means that the audio backend (RtAudio in this example), is no longer hidden in cmake settings and behind the scenes code, but up front declarative. It's possible to instantiate more than one kind of backend, and manage them on the fly. Due to the nature of the back end systems, it may not be possible to have more than one AudioDevice active at a time though, as some systems, such as RtAudio, use global variables and make other assumptions that break such usage.
- This separated management will be leveraged more highly in the future to make it much more straight forward to bind LabSound to other audio backends.
- rename DelayMode to DelayModel to match spec
- refactor parameter and setting initializaiton for audio nodes to refer to AudioParamDescriptors and AudioSettingDescriptors. This makes parameters and settings fully dynamic, and consistently iterable from external programs.
- HRTF panner now works for sample rates other than 44.1 khz.
- panners no longer buzz or pop at the end of a sample file
- the HRTF database is manually loaded now, so that the application has full control of when and how to load it. See ex_hrtf_spatialization for details.
- channel merger and splitter now function per the WebAudio spec
- WaveTable has been renamed to PeriodicWave to match the rename in the WebAudio spec.
- AudioNode now takes an AudioNodeDescriptor. Any custom nodes will need a tweak to match the new API. The interface for existing nodes hasn't changed. For reference when updating your own nodes, have a look at any of the built in nodes to see what the descriptor set up looks like.
- PannerNode no longer is responsible for managing the HRTF database
- HRTF database is owned by the AudioContext, and must be explicitly loaded.
- There is no default HRTF database
- Audio parameters constructor changed due to AudioParamDescriptor change.
Simplified channel management by automatically conforming signal processing pathways from the leaf nodes. Leaf node channel counts are propagated backwards towards the audio context such that intermediate nodes preserve channel counts, unless they have an explicit behavior of changing the channel count. For example, connecting a stereo source to a gain node will promote a mono gain node to stereo. A mono node connected to a stereo panner node will result in a stereo output.
SampledAudioNode has been rewritten to do it's own compositing when scheduled such that many instances play in an overlapping manner. This allows efficiently playing dozens of copies of the same sound simultaneously. The new algorithm requires less intermediate storage and has a much lower CPU cost. Detuning has been improved to allow a greater working range, and higher quality when pitching higher. SampledAudioNode no longer has an embedded stereo panning node, thus simplifying panning and compositing logic. Scheduling is now sample accurate, rather than quantum-accurate, and loop points and loop counts have been added.
The WaveShaperNode has been rewritten to minimize CPU overhead. Unnecessary object orient complexity has been deleted.
Threading and thread safety has been improved. The update service thread is nearly empty at this point, as graph modifications are now managed by the main update. Communications with other threads now occurs through concurrent lockfree queues. Many mutexes are now gone, as well as rare termination races.
- autoconfigure channel counts
- rename BPMDelay.h to BPMDelayNode.h
- rewrite WaveShaper for efficiency
- rewrite SampledAudioNode for efficiency and sample accurate timing
- rewrite resampler for SampledAudioNode for quality and speed
- use concurrent queues to eliminate mutexes in audio callback
- minimize use of maintenance thread
- bulletproof asynchronous connects and disconnects of node wiring
- improved offline context interface
- recorder node can run as a pull node indepdently of device context
- improved miniaudio backend
- numerous Android fixes (by xioxin)
- New platforms: Linux, Raspberry Pi, iOS, Android, in addition to macOS and Windows.
- various fixes to regressions in SampledAudioNode
- add accessors to expose node parameters for integration into DAWs
- various platform fixes for iOS and Android
- Github Actions now managing the CI
This 1.0 release finalizes the API with functionality that allows full introspection of all nodes and attributes. To go with this 1.0, LabSoundGraphToy is also released at https://github.com/LabSound/LabSoundGraphToy to enable exploration of the library, its nodes, and its functionality.
- Much thread safety introduced
- ADSRNode now has a gate param that can be used to schedule note on and off
- Profiling system added
- Node scheduling reworked
- Audio popping mitigated
- RecorderNode defers mix to down to save
- RecorderNode create bus from recording
- SfxrNode exposes presets as settings
- SampledAudioNode has scheduled voices
- AudioNodes all take an AudioContext& during construction
- Offline rendering works consistently with online rendering, including automatic pull nodes
- Automatic pull nodes work reliably
- Updated use of data types to conform to LabSound conventions
- New GranulationNode
- fixed almost all of the compiler warnings
- removed bang in favor of an onStart callback that allows rescheduling in the processing call
- Removed many custom backends in favor of RtAudio and miniaudio
- Consolidated demos into a single header
- clang format applied to sources
- cloning method added to AudioBus
- Added Input and Output configuration structs to allow selection of devices
- names added for params and settings for use in tools
- Bus added as available Setting type
- ADSRNode timing fixed
- AnalyserNode can render output to any desired output buffer size by binning and interpolation
- AudioDevice nodes added in major backend refactoring
- ConvolverNode - new implementation
- DiodeNode is now derived from AudioNode
- OscillatorNode - new implementation
- SuperSaw is now scheduled
-
AudioSettings. AudioParams already exist for rate varying values on Audio Nodes. AudioParams are registered on nodes by name, and can be looked up by name. Non-rate varying parameters such as filter modes, buffer sizes, and so on could not be similarly looked up by name. AudioSettings are introduced in this release for non-rate varying parameters, such as filter modes, buffer sizes and so on. Settings for each AudioNode are documented in its corresponding header file. A vector of param or setting names can be fetched from an AudioNode and used to populate a user interface. Each fetched AudioSetting can be queried as to whether its last set type was a Uint32 or a Float, in order that user interfaces can self configure appropriately.
-
Added some nullchecks to AudioChannel memory management as pointed out by harmonicvision
- clean up how AudioNodes are connected to one another. Formerly, the connect(...) method was called directly on an AudioNode, taking the graph lock and effecting changes to the AudioContext itself. This is no longer the case. AudioContext is now the arbiter of all graph changes.
- Numerous bug and maintenance fixes.
- A cmake build system.
- Several breaking API changes related to AudioContext creation and PannerNode initialization.
- Maintenance vs2015