Voice manager and TTS: select the audio format

[lolodomo]

The algorythm behind getBestMatch is not really clear to me. Can someone explain it ? Is it right that the idea was to give priority to WAV format ? As the TTS and sinks audio formats are in not ordered sets (Set), I am not really sure the result is the one expected.

I have now a use case where MP3 format is used by VoiceRSS but due to the low bitrate of the produced file, the audio sink is not able to render it. Maybe we should have the minimum MP3 bitrate accepted by any audio sink + the MP3 bitrate used by any TTS service in order to avoid selecting MP3 format if the 2 values ae not compatible. Another solution would be to let the user select/force an audio format when running TTS. Any thoughts ?

[lolodomo]

I would like to know if it was intentional to have the methods getBestMatch and getPreferredFormat more or less duplicated here https://github.com/eclipse/smarthome/blob/master/bundles/core/org.eclipse.smarthome.core.audio/src/main/java/org/eclipse/smarthome/core/audio/AudioFormat.java#L250 and here https://github.com/eclipse/smarthome/blob/master/bundles/core/org.eclipse.smarthome.core.voice/src/main/java/org/eclipse/smarthome/core/voice/internal/VoiceManagerImpl.java#L302

There are few differences:

• in AudioFormat, getBestMatch returns null if there is no preferred format found for source, while not in VoiceManagerImpl
• in AudioFormat, getPreferredFormat uses by default 16bit, 16KHz, big endian, while 16bit, 44,1KHz, big endian in VoiceManagerImpl

Would it make sense to have only one instance for these 2 methods in AudioFormat ? For me, the correct getBestMatch method is the one in VoiceManagerImpl. For getPreferredFormat, we could imagine to have defaultBitDepth and defaultFrequency as parameters of the method ?

[lolodomo]

To solve the problem identified by my use case (audio sink incompatible with low MP3 bitrate), I think the solution could be:

• to set the bitrate (64 kbps) in the MP3 format in the VoiceRSS binding (as it is what the service delivers)
• for audio sinks that don't support certain MP3 bitrates, define several supported MP3 formats, each one with an accepted bitrate, for example MP3 96 kbps, MP3 128 kbps, MP3 192 kbps, MP3 256 kbps and MP3 320 kbps but not MP3 64 kbps. Doing that, the voice manager will not find a matching format for VoiceRSS source and audio sinks that don't support MP3 64 kbps.

What do you think about that ?