How to organize streaming video from ESP32 so that it can be watched from anywhere with access to the Internet?

Question

Do it yourself

Hi all!
Straight to the point.
There is ESP32 + Relay (simple relay) + OV2640 (camera. or OV7670) - not yet decided on the camera.
There is an MQTT server
. There will be a mobile application for searching and configuring an ESP32-based device.
All this is implemented, programmed and everything works wonderfully. BUT!
We can’t figure out how to implement a video stream to a mobile application from a device.
At the moment, the OV2640 camera is working successfully. Fortunately, this means that on the local network we were able to capture the image returned from the device. In fact, it's not even a video, and not the finished picture in ASCII format. The code was found on the Internet.
As far as I understand, the device itself should stream the image (for example, RTSP) to the server, and we already take it from it in the mobile application. Now the opposite is happening ... we break into the device and get what the protection server sends to the microcontroller.
In general, I wildly apologize for the mistakes, not understanding the technologies and everything else, but I'm here because I need help. How to make it possible to receive an image (video in priority) from the camera on the ESP32 without losing the ability to control the junk via MQTT subscriptions.
In a lower priority (if you can’t pick up the video), then I would like to get a picture on request from the application. That is, we press the "Get image" button and a picture arrives from the camera.
Please send. In which direction to dig? Maybe someone has experience.
Conditions:
1. ESP32 (with camera and relay) in the local network via WiFi
2. MQTT server to the Internet
3. Mobile application anywhere where there is Internet
4. Do not touch the ports on the router (no forwarding and the like). do not touch the router at all, it works in stock mode.
Thanks everyone in advance

Answer 1

RTSP implies encoding (H263 compatible), but ESP32 cannot encode in hardware, and the resources of the processor are also not enough.
From this it follows that the ESP32 + OV2640 does not. You need a hardware codec and already stream video from it.

Answer 2

Google IP CCTV Camera. Their immeasurable number. Wired - wireless, street - home, with different picture quality, with various add-ons. functions. Many have mobile apps. And the described system is not for video at all.

Answer 3

Wow, here's something similar implemented.
You can try to rewrite and optimize
https://youtu.be/dA6Y_tvaMjQ
But as the previous speakers said, esp32 is rather weak for this, but it will do for sending a photo every 2-3 seconds.