Kindly Morrow
ESP32-CAM WiFi Camera Dev Board with OV2640 Sensor
Attach a 2MP camera and WiFi radio to your next project for under $10. The ESP32-CAM streams MJPEG video, captures stills to MicroSD, and runs your own firmware, all on a board smaller than a matchbook.
Built around the ESP32-S with a dual-core Xtensa LX6 processor, this board pairs the OV2640 image sensor directly to the chip's camera interface (DVP) for low-latency frame capture. The onboard MicroSD slot handles local storage up to 32GB, so footage survives WiFi dropouts. Flash it over serial with Arduino IDE, PlatformIO, or MicroPython.
Things to build with this
- Use the OV2640's built-in edge detection registers and the ESP32's ULP coprocessor to trigger a full-resolution JPEG capture and HTTP POST only when pixel-level motion is detected, no cloud service or external PIR needed
- Stream MJPEG frames over WiFi to a Python script running OpenCV, then send recognized face embeddings back to the ESP32 to toggle a GPIO-controlled door relay, all processing split between the board and a local machine
- Mount the board on a timelapse rig and use deep sleep between captures, waking on an RTC alarm, writing each JPEG directly to MicroSD with an incrementing filename. Weeks of footage on a single LiPo cell.
Key Features
- ESP32-S dual-core Xtensa LX6 at 240MHz with 4MB flash
- OV2640 2MP sensor: up to 1600x1200 JPEG stills, MJPEG video stream over WiFi
- MicroSD slot: local image and video storage up to 32GB (FAT32)
- 802.11 b/g/n WiFi + Bluetooth 4.2 onboard, no external module needed
- DVP camera interface with direct DMA transfer for efficient frame buffering
- Built-in LED flash (GPIO 4) programmable as a strobe or fill light
- Compact 40mm x 27mm footprint with external antenna connector for better range
Frequently Asked Questions
How do I program it? There's no USB port on the board.
You need a USB-to-TTL serial adapter (FTDI or CH340-based, 3.3V logic). Connect TX, RX, GND, and 5V, then pull GPIO 0 to GND before powering on to enter flash mode. The ESP32-CAM requires 5V at decent current (500mA minimum) during programming. A lot of upload failures trace back to an underpowered adapter.
Can this actually run facial recognition on-device?
Yes, with limits. Espressif's esp-who framework runs face detection and recognition directly on the ESP32-S using a lightweight CNN model. Accuracy is reasonable at close range with good lighting. For anything more complex, stream frames to a local server and do inference there.
What's the maximum video resolution it can stream over WiFi?
The OV2640 can output up to 1600x1200 for stills, but MJPEG streaming over WiFi is more practical at SVGA (800x600) or VGA (640x480). Higher resolutions saturate the bandwidth and cause frame drops. Most people run 640x480 at 10-15fps for reliable streaming.
Does it work with MicroPython?
It does, though camera support in MicroPython on the ESP32-CAM depends on the firmware build. Use a camera-enabled MicroPython build (search for esp32-cam micropython firmware). The standard MicroPython binary from micropython.org does not include camera drivers.
Why we stock this
Curated by Kindly Morrow. We test and vet every product before it hits the store. If we wouldn't use it in our own builds, we don't sell it.
Things to build with this
Fun projects to try once you get your hands on it.
Weather station that texts you
Hook up a temperature sensor and have it send you a Telegram message when it drops below freezing. 20 lines of MicroPython.
Garage door opener you control from bed
Wire a relay to your garage door motor and trigger it from your phone via Home Assistant. No cloud, no subscription.
Plant watering system that knows when to water
Soil moisture sensor + a small pump. Runs on a schedule or when the soil gets dry. Your plants stop dying.
Desk presence sensor for smart lighting
mmWave or PIR sensor detects when you sit down and turns on your desk lights. Turns them off when you leave. Zero effort after setup.
