ChimeSDK Audio Only - not working/can't figure out

I'm using the chime sdk + transcribe integration and it seems to work fine when a user joins with video, however if I don't call start video input, and only call start audio, the transcriptions never work. I can't figure out why that is, and I've run the system on and off just commenting out startvideoinput and that is always enough to cause this change in behavior. Maybe I'm missing a setting/configuration that is starting video has to be sent?

Here's my client code (minus some extraneous application specific code):

if (audioDevices.length > 0){
   meetingSession.audioVideo.startAudioInput(audioDevice);
   console.log(audioDevice)
   console.log("Audio device found, joining")
 } else {
   console.log("No audio devices found, unable to join");
   leaveConferenceCall();
   return;
 }
 
 const audioElement = $('#conf_audio')[0];
 meetingSession.audioVideo.bindAudioElement(audioElement);
 meetingSession.audioVideo.chooseVideoInputQuality(480, 270, 10);
 document.querySelector('#conferenceVideos').show();
 
 if (videoDevices.length > 0 && videoDevice != 'none'){
   await meetingSession.audioVideo.startVideoInput(videoDevice);
   const element = getVideoPanel('self');
   meetingSession.audioVideo.startVideoPreviewForVideoInput(element);
 } else {
   console.log("No video devices provided, joining with audio only");
 }

 const video_update_observer = {
   videoTileDidUpdate: (tileState) => {
     if (!tileState.boundAttendeeId || tileState.localTile ) {
       return;
     }
     meetingSession.audioVideo.bindVideoElement(
       tileState.tileId,
       getVideoPanel(tileState.tileId)
     );
   }
 }

 meetingSession.audioVideo.addObserver(video_update_observer);

 const video_remove_observer = {
   videoTileWasRemoved: (tileId) => {
     if (tileId in videoPanels) {
       meetingSession.audioVideo.unbindVideoElement(tileId);
       videoPanels[tileId].remove();
       delete videoPanels[tileId];
     }
   }
 }

 meetingSession.audioVideo.transcriptionController?.subscribeToTranscriptEvent(transcriptEventHandler);

 meetingSession.audioVideo.addObserver(video_remove_observer);
 
 if (videoDevice != 'none'){
   meetingSession.audioVideo.startLocalVideoTile();
 }

 meetingSession.audioVideo.start();

If I either set the video device to 'none' or comment out the start video call, joining with just audio doesn't work and no transcript events are produced (beyond the initial start event). With video, works fine.