What DPI Should You Scan Photos At?
Scan DPI is how many samples the sensor takes per inch of the original. 300, 600, or 1200 is a capture setting, not a label you paste on later. Here is which number to pick, the pixel sizes you get, and the one mistake that adds no detail.
The short answer
What DPI to scan photos is three jobs, not one number. Match the original and what you will do with the file.
- 300 DPI — same-size reprints of paper prints, and documents you may reprint or run through OCR. A 4 × 6 in print at 300 is 1200 × 1800 px, which is what a minilab wants for another 4 × 6.
- 600 DPI — photos you may enlarge, restore, crop, or keep as an archive master. Twice the samples of 300, so a later 8 × 10 from a 4 × 6 still has 300 samples per printed inch.
- 1200–3200 DPI — 35 mm slides and negatives, because the original is tiny. A 36 × 24 mm frame at 300 DPI is only about 425 × 283 px. You need a dense sample of a small piece of film, not a photo-print habit.
Pick optical DPI, not an interpolated marketing number. If the scanner offers 4800 but the sensor is 600 optical, 4800 is software guessing. You cannot recover grain that the CCD never saw. For a box of family prints you will reprint at the same size, 300 is enough. For the one faded portrait you might print large, 600. For the sleeve of slides, start at 2400 if the hardware will do it, 1200 if that is the ceiling.
Why scan DPI is different from the DPI tag
On a scanner, DPI is a real sampling density. You are asking a sensor to take N samples per inch of the physical original. Scan a 4 × 6 in print at 300 and you should get about 1200 × 1800 px. Scan it at 600 and you get about 2400 × 3600 px — actually more information, because the sampler walked a finer grid. That is capture. The file that comes out also carries a density tag (JFIF, EXIF XResolution, PNG pHYs, TIFF XResolution), which is only a label for how those pixels should be spaced on paper later. The two numbers ought to match. Often they do not.
Relabelling is the opposite operation. Opening a 300 DPI scan and writing 600 into the tag does not walk the grid again. The pixel count stays 1200 × 1800. Print software will now claim a 2 × 3 in print. No extra grain, no extra face detail, no archive master. That rewrite is useful when the tag is wrong — a 600 optical scan saved as 72 — and useless as a substitute for scanning. For dots versus pixels in the file, see DPI vs PPI. For why raising the tag adds no pixels, see Increase Image DPI.
The one mistake: treating “increase DPI” in an editor as a rescan. Scanning at 600 genuinely captures more. Relabelling a 300 scan to 600 does not. If the scan was truly 200 DPI and you need 300, the honest move is to put the print back on the glass.
Pixel sizes you get
Pixels = inches × scan DPI. For film, convert millimetres first: inches = mm ÷ 25.4, then multiply by DPI and round to the nearest integer. The 4 × 6 row is the one most family archives actually produce.
| Original | 300 DPI | 600 DPI | 1200 DPI |
|---|---|---|---|
| 4 × 6 in print | 1200 × 1800 | 2400 × 3600 | 4800 × 7200 |
| 5 × 7 in print | 1500 × 2100 | 3000 × 4200 | 6000 × 8400 |
| 8 × 10 in print | 2400 × 3000 | 4800 × 6000 | 9600 × 12000 |
A 35 mm frame is 36 × 24 mm (about 1.417 × 0.945 in). At photo-print DPIs it is undersampled; film scanners are specified at 1200, 2400, or 3200 for that reason. Width = 36 ÷ 25.4 × DPI; height = 24 ÷ 25.4 × DPI.
| Original | 1200 DPI | 2400 DPI | 3200 DPI |
|---|---|---|---|
| 35 mm frame (36 × 24 mm) | 1701 × 1134 | 3402 × 2268 | 4535 × 3024 |
Read the 4 × 6 @ 600 cell against a later print: 2400 × 3600 px at 300 DPI on paper is an 8 × 12 in enlargement — or an 8 × 10 after a modest crop — without inventing pixels. The same print scanned at 300 and then “bumped” to 600 in software is still 1200 × 1800, a 4 × 6 at 300. That is why the scan setting has to be chosen before you put the lid down.
File size and formats
Uncompressed 8-bit RGB is width × height × 3 bytes. A 4 × 6 in print scanned at 600 DPI is 2400 × 3600 × 3 = 25,920,000 bytes, about 24.7 MB. JPEG of that scan is a fraction — often 2–8 MB at a quality you would keep — because the encoder throws away what the eye is less likely to miss. TIFF and PNG are larger: an uncompressed TIFF sits near that 24.7 MB; LZW TIFF and PNG come in under it but still dwarf a JPEG. 16-bit (48-bit colour) roughly doubles the TIFF.
Scan masters to TIFF when you can: no JPEG generation loss, and scanner software usually writes the optical DPI into XResolution / YResolution. JPEG is fine for sharing and for reprints you will not edit again. If a scanner TIFF already has the right pixels but the tag is 72, 96, or missing, rewrite the tag — do not re-encode the image. The TIFF DPI Changer is built for those scanner TIFFs.
Disk fills up at 1200 on paper. A 4 × 6 @ 1200 is 4800 × 7200, about 99 MB uncompressed. That is why 1200 is the wrong default for a shoebox of prints and the right default for a 36 mm strip of film.
When the scanner saved the wrong DPI tag
Consumer scanner software is famous for this: you set 600 DPI, the CCD samples 600, and the JPEG lands tagged 72 — a leftover screen default — or with no density field at all. Windows then reports 72. A layout program places the picture as if it were a poster. Nothing is wrong with the pixels. The capture record was overwritten on export.
Prove it by counting. A 4 × 6 in original that is 2400 × 3600 px was sampled at 600, whatever the Properties panel says. A 1200 × 1800 file from the same print was sampled at 300. If the pixel count matches the optical setting and the tag does not, fix the tag losslessly with the Image DPI Changer on the homepage. Check the current value first with the DPI Checker (JPEG and PNG). You are restoring the capture record, not pretending you scanned twice.
If the pixel count is the 72-DPI-sized one — a 4 × 6 that is 432 × 648 px — you actually scanned at 72, or someone downsampled after. Relabelling that file to 300 does not invent a reprint. Put the print back on the glass at 300 or 600.
Settings checklist
- Colour 24-bit. Photographs want 8 bits per channel (24-bit RGB). 8-bit greyscale is for text and line art. 48-bit colour is for archives that will be graded later and for disks that can take the weight; it is not a sharpness upgrade on a 300 DPI scan.
- No auto-sharpen. Scanner sharpening bakes a halo you cannot take off. Capture flat; sharpen a copy when you print. Unsharp already in the master is the other common archive regret, after scanning at 72.
- Descreen only for printed halftones. Magazines, newspapers, and books are dots. Descreen (a blur matched to the screen frequency) is for those. A photographic print is continuous tone. Descreen on a family snapshot just softens it.
- Clean the glass. Dust at 600 DPI is a white asteroid on a sky. Wipe the platen, and the print. For film, use a blower on the holder. Scanning twice because of a hair costs more time than a cloth.
Leave auto-crop off if it eats borders or deckle edges you wanted. Preview once, set the bed region, then scan. Optical DPI first; format second; tag last — and only if the exporter lied about the number you already captured.