How this instrument works
Every pixel in a digital image stores a fixed number of bits — its color depth or bit depth — and a raster image is nothing more than a grid of those pixels, width times height of them. Multiply pixel count by bit depth and you get the total bits of raw image data; divide by eight for bytes and by a million for megabytes, and you have the uncompressed size a format like BMP or TIFF would need to store that image with no compression at all.
The compression ratio field lets this same formula estimate compressed formats too. A compression ratio of 1 means no compression, the raw pixel-data size; a typical JPEG might compress natural photos by a ratio of 10:1 or more, so entering 10 divides the raw size down to a rough estimate of the JPEG's size. Real compressors vary enormously with image content — a photo of noisy foliage compresses far less than a flat-color logo — so treat any compressed estimate as a ballpark, not a guarantee.
This formula deliberately ignores file-format overhead: headers, color palettes, row padding and metadata all add a small, mostly fixed number of bytes on top of the raw pixel data this calculator computes. For any image large enough to be worth sizing, that overhead is a rounding error next to the pixel data itself.
- Enter Width and Height in pixels — check an image's properties panel or export dialog if you don't already know them.
- Enter Bit depth in bits per pixel — 24 for standard true-color RGB, 8 for a 256-color palette, 32 to include an alpha channel.
- Leave Compression ratio at 1 for a raw, uncompressed estimate, or enter a higher number to roughly estimate a compressed format like JPEG.
- Read File size (MB) — the estimated storage this image occupies at those settings.
Worked example — a 1920x1080 24-bit image
Enter 1920 pixels for width, 1080 pixels for height, 24 bits for bit depth, and leave Compression ratio at 1 — a standard Full HD photo or screenshot with true-color, uncompressed storage. Multiplying gives 1920 x 1080 x 24 = 49,766,400 bits of raw pixel data.
Dividing by 8 x 1,000,000, File size (MB) reads 6.221 MB. That is roughly what this image would occupy saved as an uncompressed BMP or TIFF; saving the same photo as a JPEG instead typically shrinks it to somewhere between 500 KB and 2 MB, depending on how much detail the compressor can discard.
Questions
How big is an uncompressed 1920x1080 image file?
About 6.221 MB at 24-bit color. (1920 x 1080 x 24) / (8 x 1,000,000) = 6.2208 MB exactly, which the instrument displays rounded to 6.221 MB for a raw, uncompressed Full HD image — saving it as a compressed JPEG instead typically brings that down to well under 2 MB.
What bit depth should I use for a photo?
24 bits per pixel is standard true-color RGB, 8 bits each for red, green and blue, and covers the vast majority of photos and screenshots. Use 32 if the image also needs an alpha transparency channel, or 8 for a simple palette image limited to 256 colors, such as many icons and GIFs.
Why is my actual JPEG so much smaller than this raw estimate?
Because JPEG compression is doing real work: it analyzes the image and discards color and detail information the eye is unlikely to notice, often shrinking a photo to a tenth of its raw size or less. Enter a compression ratio, like 10, in this calculator to get a rough estimate of that compressed size instead of the raw, uncompressed figure.
Does this account for file headers and metadata?
No — it calculates the raw pixel-data size only. Real image files add a small amount of overhead for headers, color tables and metadata like EXIF camera data, but for any image with a meaningful pixel count that overhead is negligible next to the pixel data itself, typically well under 1% of the total file size.
Why does doubling the resolution more than double the file size?
It doesn't multiply by two — it multiplies by four, because both width and height double. Going from a 960x540 image to a 1920x1080 image quadruples the pixel count, and file size scales directly with pixel count, so the uncompressed file size quadruples too.