diff --git a/doc/translations/extract.py b/doc/translations/extract.py
new file mode 100644
index 0000000000..c794490f1f
--- /dev/null
+++ b/doc/translations/extract.py
@@ -0,0 +1,260 @@
+#!/usr/bin/env python3
+
+import argparse
+import os
+import re
+import shutil
+from collections import OrderedDict
+
+EXTRACT_TAGS = ["description", "brief_description", "member", "constant", "theme_item", "link"]
+HEADER = """\
+# LANGUAGE translation of the Godot Engine class reference
+# Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur.
+# Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md).
+# This file is distributed under the same license as the Godot source code.
+#
+# FIRST AUTHOR <EMAIL@ADDRESS>, YEAR.
+#
+#, fuzzy
+msgid ""
+msgstr ""
+"Project-Id-Version: Godot Engine class reference\\n"
+"Content-Type: text/plain; charset=UTF-8\\n"
+"Content-Transfer-Encoding: 8-bit\\n"
+
+"""
+
+## <xml-line-number-hack from="https://stackoverflow.com/a/36430270/10846399">
+import sys
+
+sys.modules["_elementtree"] = None
+import xml.etree.ElementTree as ET
+
+## override the parser to get the line number
+class LineNumberingParser(ET.XMLParser):
+    def _start(self, *args, **kwargs):
+        ## Here we assume the default XML parser which is expat
+        ## and copy its element position attributes into output Elements
+        element = super(self.__class__, self)._start(*args, **kwargs)
+        element._start_line_number = self.parser.CurrentLineNumber
+        element._start_column_number = self.parser.CurrentColumnNumber
+        element._start_byte_index = self.parser.CurrentByteIndex
+        return element
+
+    def _end(self, *args, **kwargs):
+        element = super(self.__class__, self)._end(*args, **kwargs)
+        element._end_line_number = self.parser.CurrentLineNumber
+        element._end_column_number = self.parser.CurrentColumnNumber
+        element._end_byte_index = self.parser.CurrentByteIndex
+        return element
+
+
+## </xml-line-number-hack>
+
+
+class Desc:
+    def __init__(self, line_no, msg, desc_list=None):
+        ## line_no   : the line number where the desc is
+        ## msg       : the description string
+        ## desc_list : the DescList it belongs to
+        self.line_no = line_no
+        self.msg = msg
+        self.desc_list = desc_list
+
+
+class DescList:
+    def __init__(self, doc, path):
+        ## doc  : root xml element of the document
+        ## path : file path of the xml document
+        ## list : list of Desc objects for this document
+        self.doc = doc
+        self.path = path
+        self.list = []
+
+
+def print_error(error):
+    print("ERROR: {}".format(error))
+
+
+## build classes with xml elements recursively
+def _collect_classes_dir(path, classes):
+    if not os.path.isdir(path):
+        print_error("Invalid directory path: {}".format(path))
+        exit(1)
+    for _dir in map(lambda dir: os.path.join(path, dir), os.listdir(path)):
+        if os.path.isdir(_dir):
+            _collect_classes_dir(_dir, classes)
+        elif os.path.isfile(_dir):
+            if not _dir.endswith(".xml"):
+                # print("Got non-.xml file '{}', skipping.".format(path))
+                continue
+            _collect_classes_file(_dir, classes)
+
+
+## opens a file and parse xml add to classes
+def _collect_classes_file(path, classes):
+    if not os.path.isfile(path) or not path.endswith(".xml"):
+        print_error("Invalid xml file path: {}".format(path))
+        exit(1)
+    print("Collecting file: {}".format(os.path.basename(path)))
+
+    try:
+        tree = ET.parse(path, parser=LineNumberingParser())
+    except ET.ParseError as e:
+        print_error("Parse error reading file '{}': {}".format(path, e))
+        exit(1)
+
+    doc = tree.getroot()
+
+    if "name" in doc.attrib:
+        if "version" not in doc.attrib:
+            print_error("Version missing from 'doc', file: {}".format(path))
+
+        name = doc.attrib["name"]
+        if name in classes:
+            print_error("Duplicate class {} at path {}".format(name, path))
+            exit(1)
+        classes[name] = DescList(doc, path)
+    else:
+        print_error("Unknown XML file {}, skipping".format(path))
+
+
+## regions are list of tuples with size 3 (start_index, end_index, indent)
+## indication in string where the codeblock starts, ends, and it's indent
+## if i inside the region returns the indent, else returns -1
+def _get_xml_indent(i, regions):
+    for region in regions:
+        if region[0] < i < region[1]:
+            return region[2]
+    return -1
+
+
+## find and build all regions of codeblock which we need later
+def _make_codeblock_regions(desc, path=""):
+    code_block_end = False
+    code_block_index = 0
+    code_block_regions = []
+    while not code_block_end:
+        code_block_index = desc.find("[codeblock]", code_block_index)
+        if code_block_index < 0:
+            break
+        xml_indent = 0
+        while True:
+            ## [codeblock] always have a trailing new line and some tabs
+            ## those tabs are belongs to xml indentations not code indent
+            if desc[code_block_index + len("[codeblock]\n") + xml_indent] == "\t":
+                xml_indent += 1
+            else:
+                break
+        end_index = desc.find("[/codeblock]", code_block_index)
+        if end_index < 0:
+            print_error("Non terminating codeblock: {}".format(path))
+            exit(1)
+        code_block_regions.append((code_block_index, end_index, xml_indent))
+        code_block_index += 1
+    return code_block_regions
+
+
+def _strip_and_split_desc(desc, code_block_regions):
+    desc_strip = ""  ## a stripped desc msg
+    total_indent = 0  ## code indent = total indent - xml indent
+    for i in range(len(desc)):
+        c = desc[i]
+        if c == "\n":
+            c = "\\n"
+        if c == '"':
+            c = '\\"'
+        if c == "\\":
+            c = "\\\\"  ## <element \> is invalid for msgmerge
+        if c == "\t":
+            xml_indent = _get_xml_indent(i, code_block_regions)
+            if xml_indent >= 0:
+                total_indent += 1
+                if xml_indent < total_indent:
+                    c = "\\t"
+                else:
+                    continue
+            else:
+                continue
+        desc_strip += c
+        if c == "\\n":
+            total_indent = 0
+    return desc_strip
+
+
+## make catlog strings from xml elements
+def _make_translation_catalog(classes):
+    unique_msgs = OrderedDict()
+    for class_name in classes:
+        desc_list = classes[class_name]
+        for elem in desc_list.doc.iter():
+            if elem.tag in EXTRACT_TAGS:
+                if not elem.text or len(elem.text) == 0:
+                    continue
+                line_no = elem._start_line_number if elem.text[0] != "\n" else elem._start_line_number + 1
+                desc_str = elem.text.strip()
+                code_block_regions = _make_codeblock_regions(desc_str, desc_list.path)
+                desc_msg = _strip_and_split_desc(desc_str, code_block_regions)
+                desc_obj = Desc(line_no, desc_msg, desc_list)
+                desc_list.list.append(desc_obj)
+
+                if desc_msg not in unique_msgs:
+                    unique_msgs[desc_msg] = [desc_obj]
+                else:
+                    unique_msgs[desc_msg].append(desc_obj)
+    return unique_msgs
+
+
+## generate the catlog file
+def _generate_translation_catalog_file(unique_msgs, output):
+    with open(output, "w", encoding="utf8") as f:
+        f.write(HEADER)
+        for msg in unique_msgs:
+            if len(msg) == 0:
+                continue  ## ignore
+
+            f.write("#:")
+            desc_list = unique_msgs[msg]
+            for desc in desc_list:
+                path = desc.desc_list.path.replace("\\", "/")
+                if path.startswith("./"):
+                    path = path[2:]
+                f.write(" {}:{}".format(path, desc.line_no))
+            f.write("\n")
+
+            f.write('msgid "{}"\n'.format(msg))
+            f.write('msgstr ""\n\n')
+
+    ## TODO: what if 'nt'?
+    if os.name == "posix":
+        print("Wrapping template at 79 characters for compatibility with Weblate.")
+        os.system("msgmerge -w79 {0} {0} > {0}.wrap".format(output))
+        shutil.move("{}.wrap".format(output), output)
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--path", "-p", default=".", help="The directory containing XML files to collect.")
+    parser.add_argument("--output", "-o", default="translation_catlog.pot", help="The path to the output file.")
+    args = parser.parse_args()
+
+    output = os.path.abspath(args.output)
+    if not os.path.isdir(os.path.dirname(output)) or not output.endswith(".pot"):
+        print_error("Invalid output path: {}".format(output))
+        exit(1)
+    if not os.path.isdir(args.path):
+        print_error("Invalid working directory path: {}".format(args.path))
+        exit(1)
+
+    os.chdir(args.path)
+    print("Current working dir: {}\n".format(os.getcwd()))
+
+    classes = OrderedDict()  ## dictionary of key=class_name, value=DescList objects
+    _collect_classes_dir(".", classes)
+    classes = OrderedDict(sorted(classes.items(), key=lambda kv: kv[0].lower()))
+    unique_msgs = _make_translation_catalog(classes)
+    _generate_translation_catalog_file(unique_msgs, output)
+
+
+if __name__ == "__main__":
+    main()