Hardware Preprocessors

[SineVector241]

Hi. Is there a way to link hardware preprocessors such as Echo Cancellation, Denoiser and AutomaticGainControl?

For example I have a custom preprocessor classes that use the hardware version of those preprocessors which requires a session ID from the android recorder.

Example This is available via the net9.0-android TFM.

_echoCanceler = AcousticEchoCanceler.Create((int)_recorder.SessionId);

Can I use the default recorder available in MiniAudioEngine to link it up or would I need to create my own recorder/per platform audio engine and then link up the preprocessors?

[LSXPrime]

I don't have any knowledge about the audio API in Android, so I can't help with this specific issue. However, in this example, if the sessionId refers to the current device capture, you can acquire it from AudioEngine.Instance.Devices[CURRENT_DEVICE_INDEX].Id. This will return a pointer to an ma_device_id struct. This struct contains IDs for the current backend (for Android, this would be aaudio or opensl, represented as int or uint). You will need to manually parse this pointer to access the struct's contents yourself.

Regarding processing, you can add modifiers to the recorder component. These modifiers will process the audio data before it is passed to the encoder.

I've moved this to the Discussions section as it's not a bug report or feature request.

[SineVector241]

Thanks.

Usually with hardware preprocessors they handle everything before it even passes as a sample to the recorder (isn't specific to android, there's also a UWP driver preprocessor, iOS, etc...)

As for another question. I have noticed that you can only have 1 audio engine (I might be wrong here) which immediately starts the audio devices upon creation without being able to select a device beforehand and/or optionally declare that the device should start after creation which could be a problem since the default audio device may not support a certain format while the selected device does.

Another thing I noticed is that it doesn't seem like you can have multiple audio engines.
This doesn't completely fit well with what I am trying to do which is recording at 16bit Mono, compressing, sending over the network and playing back at 32bit IEEEFloat Stereo. I did a bit of digging through the code and it seems like you are managing a single static Instance that is shared when creating new components, visualisers, modifiers, etc...

This does make it a bit annoying since I can't explicitly select an audio engine to tell any newly created components to use that AudioEngine (Little bit similar to a state machine like OpenAL or OpenGL).

So the 2 main questions are:
Can I create multiple AudioEngines at different formats?
Can I select a device when creating an AudioEngine before it starts initialisation of the device and playing/recording?

This is a great project and does everything I need for my application except for the things I mentioned above which is preventing me from porting over due to how I handle audio devices and preprocessors both at the hardware and software level.

[LSXPrime]

Can I create multiple AudioEngines at different formats?

No, it's mainly working with one engine per instance, although we got Multiple Device support #26 in my TODO list, If multiple engines were needed or seemed beneficial, I would include it.

Can I select a device when creating an AudioEngine before it starts initialisation of the device and playing/recording?

I missed this while implementing the device switching feature, which can be done maybe with changing 2 code lines.

I am trying to do which is recording at 16bit Mono, compressing, sending over the network and playing back at 32bit IEEEFloat Stereo

I am actually doing something similar (Recording -> Compressing 32-Bit Float to 16-Bit S16 -> Send -> Processing on a server that needs 16-bit audio -> Receive -> Decompressing 16-Bit S16 to 16-Bit Float -> Playback) all on same AudioEngine.

[LSXPrime]

As for another question. I have noticed that you can only have 1 audio engine (I might be wrong here) which immediately starts the audio devices upon creation without being able to select a device beforehand and/or optionally declare that the device should start after creation which could be a problem since the default audio device may not support a certain format while the selected device does.

Another thing I noticed is that it doesn't seem like you can have multiple audio engines.
This doesn't completely fit well with what I am trying to do which is recording at 16bit Mono, compressing, sending over the network and playing back at 32bit IEEEFloat Stereo. I did a bit of digging through the code and it seems like you are managing a single static Instance that is shared when creating new components, visualisers, modifiers, etc...

This does make it a bit annoying since I can't explicitly select an audio engine to tell any newly created components to use that AudioEngine (Little bit similar to a state machine like OpenAL or OpenGL).

So the 2 main questions are:
Can I create multiple AudioEngines at different formats?
Can I select a device when creating an AudioEngine before it starts initialisation of the device and playing/recording?

If you're still interested in the library, there's a new update at #26.

[SineVector241]

As for another question. I have noticed that you can only have 1 audio engine (I might be wrong here) which immediately starts the audio devices upon creation without being able to select a device beforehand and/or optionally declare that the device should start after creation which could be a problem since the default audio device may not support a certain format while the selected device does.

Another thing I noticed is that it doesn't seem like you can have multiple audio engines.
This doesn't completely fit well with what I am trying to do which is recording at 16bit Mono, compressing, sending over the network and playing back at 32bit IEEEFloat Stereo. I did a bit of digging through the code and it seems like you are managing a single static Instance that is shared when creating new components, visualisers, modifiers, etc...

This does make it a bit annoying since I can't explicitly select an audio engine to tell any newly created components to use that AudioEngine (Little bit similar to a state machine like OpenAL or OpenGL).

So the 2 main questions are:
Can I create multiple AudioEngines at different formats?
Can I select a device when creating an AudioEngine before it starts initialisation of the device and playing/recording?

If you're still interested in the library, there's a new update at #26.

Thanks. I'll look into it in the future but probably not for a few months since I've already got my own implementation working that I don't really want to change too much now. Probably when I release a new update for my project I'll check again.