The most common question about photo mosaics is also the hardest to answer in a single number: how many photos do you actually need? The short version is at least 300. The longer version depends on how big the grid is, how much repetition you can live with, and how varied your photos are.
The Math Behind Tile Repetition
A mosaic is a grid of cells, each one holding a single photo. How many cells there are depends on the resolution, and a typical high-quality mosaic runs 5,000 to 15,000. Since almost nobody hands over that many unique photos, the algorithm reuses tiles. The rate is just division:
Average repetition = total cells / number of photos provided
| Photos provided | 5,000-cell mosaic | 10,000-cell mosaic | 15,000-cell mosaic |
|---|---|---|---|
| 100 | 50x each | 100x each | 150x each |
| 300 | 17x each | 33x each | 50x each |
| 500 | 10x each | 20x each | 30x each |
| 1,000 | 5x each | 10x each | 15x each |
| 2,000 | 2.5x each | 5x each | 7.5x each |
Good algorithms space the duplicates out so the same photo never lands next to itself. But past roughly 50 repeats each, the patterns start to show no matter how cleverly they’re spaced.
Quality Tiers
Under 200 Photos: Minimum Viable
Possible, but compromised. Every photo turns up dozens of times. It reads fine from across the room, but up close the repetition is obvious, and some areas go patchy where the algorithm couldn’t find a good color match and had to settle.
300 to 500 Photos: Standard
The sweet spot for most people. Repetition is there but not distracting, and the algorithm has enough variety to handle most of the target’s color regions. This is the floor most professional services will accept.
500 to 1,000 Photos: High Quality
Noticeably better. More color coverage means more accurate matching, and less repetition means more of the picture feels one of a kind. If you’re printing big, this range earns its keep.
1,000+ Photos: Premium
Exceptional. Each tile shows up fewer than ten times, and every section feels distinct. This is where the gap between a good algorithm and a great one becomes visible, because the software finally has real choices to make instead of reusing whatever’s left.
Why Variety Matters More Than Quantity
A thousand photos of your dog on the same brown couch will lose to three hundred taken across different places, seasons, and light. Here’s why.
The algorithm has to match every color region in your target. If the target is a face against a blue sky, it needs:
- Skin tones (warm beiges, pinks, browns)
- Blues (the sky)
- Darks (hair, shadows)
- Lights (highlights, teeth, eyes)
If all thousand of your photos are warm indoor shots, the face will come out beautifully and the sky won’t. The algorithm recycles the handful of bluish photos it has, or fakes it with poor matches, and the sky goes muddy.
Tips for Building a Better Tile Library
- Mix environments. Indoor, outdoor, beach, city, park. Different settings hand you different palettes for free.
- Mix lighting. Sun, overcast, golden hour, flash. Each shifts the dominant tones.
- Include close-ups. Food, flowers, textures, objects. They tend to be saturated and uniform, which makes them gold for matching.
- Don’t over-curate. That “bad” photo with the weird cast might be the exact green the algorithm needs for a patch of leaves. Let it decide what’s useful.
- Skip near-duplicates. Ten frames from one burst add count, not variety. Keep the best and move on.
What About the Target Image?
The target matters too. High-contrast targets, with strong light and dark, are forgiving, because the algorithm has clear bins to sort tiles into. Soft, pastel, evenly lit targets are demanding: the color matching has to be more precise, so you need more tile variety to pull it off.
Faces are usually easier than landscapes. Skin sits in a narrow band of color, so even a modest library tends to hold enough warm tones to fill it convincingly.