Build a Cloud-Connected Photo Frame with Raspberry Pi, Google Drive, rclone, fbi, and Python

I have been shooting with a digital SLR since 2003, and my total shot count is now at 388,524. I wanted a way to display those photos as a slideshow at home — without the hassle of manually copying files to an SD card every week.

The Problem with Commercial Photo Frames

Most digital photo frames read images from an SD card or USB drive, in the 7–10 inch range. They work fine, but constantly re-copying photos is tedious enough to make you stop using it.

What I wanted was a photo frame that pulls photos automatically from cloud storage like Google Drive. No such product seemed to exist, so I built one.

Hardware

Raspberry Pi 4

A photo frame runs 24/7, so low power consumption matters. Raspberry Pi 4 fits well.

Raspberry Pi 4 Model B 4GB (Japan authorized distributor)

Amazonで見る

Mobile Monitor

A 15.6-inch FullHD IPS mobile monitor. Choosing an IPS panel avoids the narrow viewing angles common in off-the-shelf photo frames. Mobile monitors are compact and flexible to position.

15.6-inch Mobile Monitor with Touch Panel

Amazonで見る

microSD Card and Power Supply

SanDisk High Endurance microSDHC 32GB for Dashcam and Action Cameras

Amazonで見る

Miuzei Raspberry Pi 4 USB-C Power Supply 5.1V 3A with Switch (PSE certified)

Amazonで見る

System Architecture

OS: Raspberry Pi OS

Write Raspberry Pi OS to the SD card using Raspberry Pi Imager. Since no desktop is needed for a photo frame, the Lite (CLI-only) version is sufficient.

Image Display: fbi

fbi renders images directly to the framebuffer from the command line. An earlier version used a GUI on Ubuntu but proved unstable, so I switched to this CLI-based approach.

Cloud Storage Sync: rclone

rclone downloads photos from Google Drive on a schedule. It supports over 70 cloud storage providers including Dropbox and S3.

The flow is: rclone downloads photos → fbi displays them as a slideshow.

Setup

Installing Raspberry Pi OS

Use Raspberry Pi Imager to write the image to the SD card.

Either 32-bit or 64-bit works for this use case.

Installing fbi

sudo apt-get -y install fbi

Granting Permissions

By default only root can use fbi to display images. Add the pi user to the video group:

pi@raspi3-photo:~ $ ls -Fla /dev/fb0
crw-rw---- 1 root video 29, 0 Apr  7 10:33 /dev/fb0

sudo usermod -aG video pi

Installing rclone

curl https://rclone.org/install.sh | sudo bash

After installing, run rclone config to authenticate with Google Drive.

Slideshow Program

Algorithm

1. Use rclone ls to list photos in the designated cloud storage folder
2. Start downloading images at random once an initial batch is available
3. Display downloaded images with fbi
4. Continue fetching more images in the background
5. Advance to a random new image at the configured interval
6. Use weighted random selection — recently shown images are deprioritized
7. Run rclone ls weekly to merge newly added photos into the database
8. Keep the database in memory (not a JSON file) for faster merging

Language: Python

The slideshow controller is written in Python. The refactored object-oriented version reduced startup time by 1700% compared to the original.

For Python learning, Python 3 Bootcamp by Silicon Valley Engineer (Udemy) was helpful.

Source Code

Published on GitHub:

GitHub - yukishita/photoView4: Cloud-connected slideshow program

Usage

Add this to crontab to start the slideshow at boot:

@reboot /home/pi/photoView4/photoView4.sh

Demonstration

The improved v4 demo. Startup to first image display is dramatically faster than the original:

Original Version (for reference)

The original downloads the full file list before starting, which makes startup slow:

Tips: Scheduled Display Power Off

Running the display 24/7 wastes electricity. Use vcgencmd in crontab to schedule power:

0  0 * * * /usr/bin/vcgencmd display_power 0
30 6 * * * /usr/bin/vcgencmd display_power 1

This turns the display off at midnight and back on at 6:30 AM. A motion or light sensor can make this even more automatic.

Summary

Raspberry Pi 4 was used here, but Raspberry Pi Zero 2 W should also work. For an always-on device, choosing a lower-power board reduces running costs.