AudioSession: Failed to set mode

[pax-k]

I'm using Whisper.rn together with my own Text-to-Speech library (Swift + Expo).

Flow:

1. Run transcribeRealtime()
2. Start speaking
3. Run stop()
4. My TTS library starts to playback the transcribed text

It works on the first run of the flow, but running it a second time, updateAudioSession() throws the error:

Error: Failed to set mode. Domain=NSOSStatusErrorDomain Code=-50 "(null)"

I suspect something happens when whisper.rn tries to restore the audio session a second time (after my TTS library sets its own session).

This is how whisper.rn is configured:

audioSessionOnStartIos: {
          category: AudioSessionIos.Category.Record,
          options: [
            AudioSessionIos.CategoryOption.InterruptSpokenAudioAndMixWithOthers,
            AudioSessionIos.CategoryOption.AllowBluetooth,
            AudioSessionIos.CategoryOption.AllowBluetoothA2DP,
          ],
          mode: AudioSessionIos.Mode.VoiceChat,
        },
        audioSessionOnStopIos: 'restore',

This is how my TTS lib sets its session:

self.audioSession = AVAudioSession.sharedInstance()
            try self.audioSession.setCategory(.playback, mode: .voicePrompt, options: [.duckOthers, .allowBluetooth, .allowBluetoothA2DP, .allowAirPlay])
            try self.audioSession.setActive(true, options: .notifyOthersOnDeactivation)

Do you have any clues? Thanks