diff --git a/.gitignore b/.gitignore
index 5bcf497..f45a78b 100644
--- a/.gitignore
+++ b/.gitignore
@@ -22,6 +22,7 @@ var/
 wheels/
 pip-wheel-metadata/
 share/python-wheels/
+share/tessdata/
 *.egg-info/
 .installed.cfg
 *.egg
diff --git a/install/common/download-tessdata.py b/install/common/download-tessdata.py
new file mode 100644
index 0000000..cf2b366
--- /dev/null
+++ b/install/common/download-tessdata.py
@@ -0,0 +1,92 @@
+import hashlib
+import io
+import json
+import logging
+import pathlib
+import subprocess
+import sys
+import tarfile
+import urllib.request
+
+logger = logging.getLogger(__name__)
+
+TESSDATA_RELEASES_URL = (
+    "https://api.github.com/repos/tesseract-ocr/tessdata_fast/releases/latest"
+)
+TESSDATA_ARCHIVE_URL = "https://github.com/tesseract-ocr/tessdata_fast/archive/{tessdata_version}/tessdata_fast-{tessdata_version}.tar.gz"
+TESSDATA_CHECKSUM = "d0e3bb6f3b4e75748680524a1d116f2bfb145618f8ceed55b279d15098a530f9"
+
+
+def git_root():
+    """Get the root directory of the Git repo."""
+    # FIXME: Use a Git Python binding for this.
+    # FIXME: Make this work if called outside the repo.
+    cmd = ["git", "rev-parse", "--show-toplevel"]
+    path = (
+        subprocess.run(cmd, check=True, stdout=subprocess.PIPE)
+        .stdout.decode()
+        .strip("\n")
+    )
+    return pathlib.Path(path)
+
+
+def main():
+    logging.basicConfig(
+        level=logging.DEBUG,
+        format="%(asctime)s - %(levelname)s - %(message)s",
+        datefmt="%Y-%m-%d %H:%M:%S",
+    )
+
+    share_dir = git_root() / "share"
+    tessdata_dir = share_dir / "tessdata"
+
+    # Get the list of OCR languages that Dangerzone supports.
+    with open(share_dir / "ocr-languages.json") as f:
+        langs_short = sorted(json.loads(f.read()).values())
+
+    # Check if these languages have already been downloaded.
+    if tessdata_dir.exists():
+        expected_files = {f"{lang}.traineddata" for lang in langs_short}
+        files = {f.name for f in tessdata_dir.iterdir()}
+        if files == expected_files:
+            logger.info("Skipping tessdata download, language data already exists")
+            return
+        else:
+            logger.info(f"Found {tessdata_dir} but contents do not match")
+            return 1
+
+    # Get latest release of Tesseract data.
+    logger.info("Getting latest tessdata release")
+    with urllib.request.urlopen(TESSDATA_RELEASES_URL) as f:
+        resp = f.read()
+        releases = json.loads(resp)
+        tag = releases["tag_name"]
+
+    # Get latest release of Tesseract data.
+    logger.info(f"Downloading tessdata release {tag}")
+    archive_url = TESSDATA_ARCHIVE_URL.format(tessdata_version=tag)
+    with urllib.request.urlopen(archive_url) as f:
+        archive = f.read()
+        digest = hashlib.sha256(archive).hexdigest()
+        if digest != TESSDATA_CHECKSUM:
+            raise RuntimeError(f"Checksum mismatch {digest} != {TESSDATA_CHECKSUM}")
+
+    # Extract the languages models from the tessdata archive.
+    logger.info(f"Extracting tessdata archive into {tessdata_dir}")
+    with tarfile.open(fileobj=io.BytesIO(archive)) as t:
+        for lang in langs_short:
+            member = f"tessdata_fast-{tag}/{lang}.traineddata"
+            logger.info(f"Extracting {member}")
+            # NOTE: We want `filter="data"` because it ignores ownership info, as
+            # recorded in the tarfile. This filter will become the default in Python
+            # 3.14. See:
+            #
+            # https://docs.python.org/3/library/tarfile.html#tarfile-extraction-filter
+            t.extract(member=member, path=share_dir, filter="data")
+
+    tessdata_dl_dir = share_dir / f"tessdata_fast-{tag}"
+    tessdata_dl_dir.rename(tessdata_dir)
+
+
+if __name__ == "__main__":
+    sys.exit(main())