Convert TIFF to JPG on Mac

Because TIFF’s default is to store every pixel, in full, with no compression at all, and a scanner produces a great many pixels.

Letter paper at 300 dpi is 2,550 × 3,300 pixels. At three bytes per pixel that is 25,245,000 bytes of raw payload before any headers, for one page of an invoice. The uncompressed TIFF our fixture produced came to 25,249,952 bytes, which is the payload plus 4,952 bytes of tags. Scan a 40-page contract at 300 dpi in colour and you have a gigabyte of paper.

Here is the same page written seven ways:

Three things fall out of that table. LZW is the setting that matters: it is lossless, it is understood by everything that reads TIFF, and it removed 82% of the file. If your scanner is producing uncompressed TIFFs, turning LZW on costs you nothing and there is no reason not to.

PackBits, on this file, did nothing at all. The sips -s formatOptions packbits output was byte-identical to the uncompressed one, and tiffinfo reports Compression Scheme: None on it. So the option was not applied rather than applied and ineffective. PackBits is a simple run-length scheme that struggles with scanner noise in any case, but do not expect sips to give it to you.

Zip beats LZW by 19% and JPEG-in-TIFF beats both, but a lossy TIFF is a strange object: you have paid TIFF’s overhead and given up the lossless guarantee that was the reason to use it. If lossy is acceptable, go to JPEG properly and take the full 92.7%.

This is the section to read before you convert anything you cannot re-scan. TIFF supports multiple image directories in one file, which is why fax software and document scanners have used it for decades. Most image tools pretend that feature does not exist.

The fixture is a three-page 13,311,448-byte TIFF. Pillow reports n_frames = 3. Here is what each route does with it:

ImageMagick is the correct tool here and it is free. It notices the extra directories and numbers the outputs for you:

brew install imagemagick

# one JPEG per page, numbered
magick contract.tif -quality 85 page.jpg

# keep the page order readable past nine pages
magick contract.tif -quality 85 page-%02d.jpg

# or collapse the pages into a single PDF, which is usually what you wanted
magick contract.tif contract.pdf

Our own failure is worth being precise about, because the error message is misleading. It is not a broken install. We tested a single-page TIFF written by the same tool, which converted normally to 2,588,914 bytes, then a two-page TIFF written by a completely different library, which failed the same way. Multi-frame input is the trigger, the same limitation that applies to animated GIF and animated WebP. If your TIFFs are scanned documents with more than one page, use ImageMagick for the split and keep us for the batch afterwards.

If the destination is a document rather than a set of images, converting a multi-page TIFF to PDF and then compressing that is a shorter road: compressing a scanned PDF without wrecking the OCR covers what that costs.

For single-page TIFFs, all the usual routes work and none of them need an install.

# a folder of single-page TIFFs, default quality (equivalent to 75)
mkdir -p jpg
sips -s format jpeg *.tif --out jpg/

# archival-leaning quality, and cap the long edge for web use
sips -s format jpeg -s formatOptions 92 scan.tif --out scan.jpg
sips -s format jpeg -Z 2400 *.tif --out jpg/

Both .tif and .tiff work, and sips identifies the format from the file contents rather than the name. Create the output directory first: --out will not make one and will not complain about its absence.

Preview covers the graphical case. Open the TIFF, File → Export As…, choose JPEG, set Quality. For a folder, open them all in one window, select the thumbnails and use Export Selected Images….

Smol converted the 4,507,252-byte LZW scan to 2,588,914 bytes, a 42.6% cut, and the Adobe RGB photograph from 653,924 to 141,346. Larger than sips in both cases because it converts at a higher internal fidelity, and it does the whole folder from a Finder right-click without a shell loop.

Yes, through every native route we tested, and this matters more for TIFF than for any other input. Scanners and repro work routinely tag files Adobe RGB or a device profile, and Adobe RGB numbers displayed as sRGB come out visibly flat and wrong.

The last row is the mistake to avoid. -strip is often recommended as a file-size trick, and it removes the ICC profile along with the metadata. The pixel values do not change, so nothing looks broken in the file listing. Every viewer downstream simply assumes sRGB and renders the colours incorrectly. If you want the metadata gone without the colour damage, strip the metadata as a separate, deliberate operation and leave the profile alone.

Resolution survives as well, which is the other thing archivists care about. A 300 dpi TIFF came out as a JPEG reporting dpiWidth: 300.000 through both sips and Smol. That tag is what makes the page print at 8.5 × 11 inches instead of at 35 × 46. Check it with sips -g dpiWidth -g dpiHeight out.jpg before you delete anything.

For a multi-page TIFF, use ImageMagick. We fail on those outright, and sips is worse than failing because it succeeds with one page and says nothing about the others. magick contract.tif page-%02d.jpg is the right answer and it is free.

For one file, use Preview or sips. Two clicks or one command, already installed, and the built-in encoder produced the smaller JPEG on every fixture here.

For a real archive, do not convert at all. If the TIFF is the preservation master, it should stay a TIFF, with LZW on and the profile intact. Generate JPEG derivatives for distribution and keep the master. Converting the master and deleting it is a one-way decision that a 92.7% size saving does not justify.

Smol earns the $29 on volume: several hundred single-page scans that all have to come out as JPEG at one setting, with profiles preserved and resolution tags intact, from the Finder right-click menu, on a machine where client documents are not allowed to be uploaded. Sixty images converted in 2.6 seconds against 12.55 seconds for the equivalent shell loop in our batch measurements. One purchase, no subscription, fully offline.

Related: PNG to JPG, JPG to PNG, WebP to JPG, compressing a scanned PDF, and every format Smol reads and writes.