diff --git a/scripts/fix_translation_memory.py b/scripts/fix_translation_memory.py
new file mode 100644
index 0000000000..622d51cfd3
--- /dev/null
+++ b/scripts/fix_translation_memory.py
@@ -0,0 +1,64 @@
+##% Script to fix a corrupted Translation memory from existing po files
+
+import os
+import re
+import argparse
+from pathlib import Path
+from fuzzywuzzy import fuzz
+from fuzzywuzzy import process
+import xml.etree.ElementTree as ET
+from xml.sax.saxutils import unescape
+
+
+def load_existing_xmtm(path: Path) -> ET.Element:
+    """Load existing xmtm file and return the root element"""
+    tree = ET.parse(path)
+    return tree.getroot()
+
+def load_existing_po(path: Path) -> dict:
+    """Load existing po file and return a dictionary of msgid and msgstr"""
+    content = path.read_text(encoding="utf-8")
+    content = "".join(content.splitlines()[16:])
+    return dict(re.findall(r'[^#]msgid.?\"+\s?([\s|\S]+?)\"*?msgstr.?\"([\s|\S]+?)\"?#', content))
+
+
+def main(tmx_source_path: Path, tmx_target_path: Path, i18n_path: Path):
+
+    po_content = {}
+    for file in i18n_path.rglob("cura.po"):
+        print(os.path.join(i18n_path, file))
+        po_content[file.relative_to(i18n_path).parts[0].replace("_", "-")] = load_existing_po(Path(os.path.join(i18n_path, file)))
+
+    root = load_existing_xmtm(tmx_source_path)
+    root_old = ET.ElementTree(root)
+    root_old.write("old.tmx", encoding="utf-8", xml_declaration=True)
+    for tu in root.iter("tu"):
+        if "cura.pot" not in [t.text for t in tu.findall("prop") if t.attrib["type"] == "x-smartling-file"]:
+            continue
+        tuvs = tu.findall("tuv")
+        key_source = tuvs[0].find("seg").text
+        key_lang = tuvs[1].attrib["{http://www.w3.org/XML/1998/namespace}lang"]
+        if key_lang in po_content and key_source in po_content[key_lang]:
+            tuvs[1].find("seg").text = po_content[key_lang][key_source]
+        else:
+            fuzz_match_ratio = [fuzz.ratio(k, key_source) for k in po_content[key_lang].keys()]
+            fuzz_max_ratio = max(fuzz_match_ratio)
+            fuzz_match_key = list(po_content[key_lang].keys())[fuzz_match_ratio.index(fuzz_max_ratio)]
+            if fuzz_max_ratio > 90:
+                fuzz_match_po_value = po_content[key_lang][fuzz_match_key]
+                tuvs[1].find("seg").text = fuzz_match_po_value
+                # print(f"[{key_lang}] Fuzz match: {key_source} -> {fuzz_match_key} -> {fuzz_match_po_value} with a ratio of {fuzz_max_ratio}")
+            else:
+                # print(f"[{key_lang}] No match for: {key_source} -> {tuvs[1].find('seg').text} -> highest ratio: {fuzz_max_ratio}")
+                print(f"[{key_lang}] {key_source} == {fuzz_match_key} [{fuzz_max_ratio}]")
+    fixed_root = ET.ElementTree(root)
+    fixed_root.write(tmx_target_path, encoding="utf-8", xml_declaration=True)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Fix a corrupted Translation memory from existing po files")
+    parser.add_argument("tmx_source_path", type=Path, help="Path to the source TMX file")
+    parser.add_argument("tmx_target_path", type=Path, help="Path to the target TMX file")
+    parser.add_argument("i18n_path", type=Path, help="Path to the i18n folder")
+    args = parser.parse_args()
+    main(args.tmx_source_path, args.tmx_target_path, args.i18n_path)