OpenCV real time streaming video capture is slow. How to get synced with real time?

Question

I'm using Python OpenCV to receive rtsp images in real time.

But there's a problem with the video being shown being too late and cutting off.

I'm looking for a way to sync in real time..

(For your information, even if there is no recognized part of the object below, it appears at the same speed...)

• cv2 speed

class Camera:
    def __init__(self, source, PT_PATH):
        self.capture = cv2.VideoCapture(source)
        if not PT_PATH == "original":
            self.model = load_model(PT_PATH)
        else:
            self.model = "original"
        self.streaming_state = False

    def start(self):
        self.streaming_state = True

    def stop(self):
        self.streaming_state = False
        self.capture.release()
        print("memory check : ",self.capture.isOpened())


    def read_frames(self):
        while self.streaming_state:
            # print("capture 상태 : ",self.capture.isOpened())
            success, capture_img = self.capture.read()
            if not success:
                print("not read frame")
                break
            
            if not self.model == "original":
                pre_process_image, model_to_device = preprocess_frame(capture_img=capture_img, model=self.model)
                post_process_image = postprocess_frame(pre_process_image, capture_img, model_to_device)
                yield (b'--frame\r\n'
                        b'Content-Type: image/jpeg\r\n\r\n' + post_process_image + b'\r\n')
            else:
                capture_img = httpProcessing(capture_img=capture_img)
                yield (b'--frame\r\n'
                        b'Content-Type: image/jpeg\r\n\r\n' + capture_img + b'\r\n')

Questions:

What am I doing wrong here?

Why is it so slow?

How do I sync it to real-time speeds?