Sample JXA script for running speech recognition on audio. Shows how you can supply code blocks to ObjC methods in JXA (lines 33-37).

SpeechRecognizer.scpt

ObjC.import("AVFoundation");
ObjC.import('Speech');
ObjC.import("objc");
	
function recordForDuration(duration, destination) {
	const settings = $.NSMutableDictionary.alloc.init;
	settings.setValueForKey($.kAudioFormatAppleIMA4, $.AVFormatIDKey);
    
	// Some macOS versions fail to link $.AVAudioFormat, so we manually get the class
	const format = $.objc_getClass("AVAudioFormat").alloc.initWithSettings(settings);
	const recorder = $.objc_getClass("AVAudioRecorder").alloc.initWithURLFormatError(destination, format, null);

	// Start recording & wait until done
	recorder.recordForDuration(duration);
	while (recorder.isRecording) {
		$.NSRunLoop.currentRunLoop.runUntilDate($.NSDate.dateWithTimeIntervalSinceNow(0.1));
	}
	recorder.stop;
}

(() => {
	const outputURL = $.NSURL.alloc.initFileURLWithPath($.NSTemporaryDirectory().stringByAppendingString("temp.mp4"));
	
	// Record audio to temp file
	recordForDuration(5, outputURL);
  
  	// Setup speech recognition
	const recognizer = $.SFSpeechRecognizer.alloc.init;
	const request = $.SFSpeechURLRecognitionRequest.alloc.initWithURL(outputURL);
	
	// Perform recognition and store result
	let finalResult = "";
	const task = recognizer.recognitionTaskWithRequestResultHandler(request, (result, error) => {
		if (result.js) {
			finalResult = result.bestTranscription.formattedString.js;
		}
	})
	
	while (task.state != $.SFSpeechRecognitionTaskStateCompleted && task.state != $.SFSpeechRecognitionTaskStateCanceling) {
		$.NSRunLoop.currentRunLoop.runUntilDate($.NSDate.dateWithTimeIntervalSinceNow(0.1));
	}
	
	// Remove temp file
	$.NSFileManager.defaultManager.removeItemAtURLError(outputURL, $.nil);
	
	return finalResult;
})()