diff --git a/sphinx/sphinx_to_html.py b/sphinx/sphinx_to_html.py
index 1107b30..7686109 100755
--- a/sphinx/sphinx_to_html.py
+++ b/sphinx/sphinx_to_html.py
@@ -28,10 +28,17 @@ if gitea_prefix.startswith(local_url):
     gitea_prefix = gitea_prefix[len(local_url):]
 
 if len(gitea_prefix):
-    org, repo, view, ref, branch = gitea_prefix.strip('/').split('/')[:5]
+    path_tokens = gitea_prefix.strip('/').split('/')
+    org, repo, view, ref, branch = path_tokens[:5]
 
     doc_url = f"{base_url}/{org}/{repo}/{view}/{ref}/{branch}"
-    image_url = f"{base_url}/{org}/{repo}/raw/{ref}/{branch}"
+    image_url = f"{base_url}/{org}/{repo}/media/{ref}/{branch}"
+
+    # Hardcoded exception for blender-manual, that has links relative
+    # to manual/ folder.
+    if len(path_tokens) > 5 and path_tokens[5] == 'manual':
+        doc_url += "/manual"
+        image_url += "/manual"
 else:
     doc_url = ""
     image_url = ""
@@ -48,10 +55,17 @@ with tempfile.TemporaryDirectory(dir=args.user_work_dir) as tmp_dir:
     page_contents = page_filepath.read_text()
 
     # Turn links into external links since internal links are not found and stripped.
-    def doc_link(matchobj):
+    def path_to_label(path):
+        path = path.removesuffix('/index')
+        return path.split('/')[-1].replace('_', ' ').replace('-', ' ').capitalize()
+    def doc_label_link(matchobj):
         return f"`{matchobj.group(1)}<{doc_url}/{matchobj.group(2).strip('/')}.rst>`_"
-    def ref_link(matchobj):
+    def doc_link(matchobj):
+        return f"`{path_to_label(matchobj.group(1))} <{doc_url}/{matchobj.group(1).strip('/')}.rst>`_"
+    def ref_label_link(matchobj):
         return f"`{matchobj.group(1)} <{placeholder_url}>`_"
+    def ref_link(matchobj):
+        return f"`{path_to_label(matchobj.group(1))} <{placeholder_url}>`_"
     def term_link(matchobj):
         return f"`{matchobj.group(1)} <{placeholder_url}>`_"
     def figure_link(matchobj):
@@ -59,12 +73,14 @@ with tempfile.TemporaryDirectory(dir=args.user_work_dir) as tmp_dir:
     def image_link(matchobj):
         return f"image:: {image_url}/{matchobj.group(1).strip('/')}"
 
-    page_contents = re.sub(":doc:`(.*)<(.+)>`", doc_link, page_contents)
-    page_contents = re.sub(":ref:`(.+)<(.+)>`", ref_link, page_contents)
-    page_contents = re.sub(":ref:`(.+)`", ref_link, page_contents)
-    page_contents = re.sub(":term:`(.+)`", term_link, page_contents)
-    page_contents = re.sub("figure:: (.+)", figure_link, page_contents)
-    page_contents = re.sub("image:: (.+)", image_link, page_contents)
+    page_contents = re.sub(":doc:`/(.+?)`", doc_link, page_contents)
+    page_contents = re.sub(":doc:`(.+?)<(.+?)>`", doc_label_link, page_contents)
+    page_contents = re.sub(":ref:`(.+?)<(.+?)>`", ref_label_link, page_contents)
+    page_contents = re.sub(":ref:`([\w\s-]+?)`", ref_link, page_contents)
+    page_contents = re.sub(":term:`(.+?)<(.+?)>`", term_link, page_contents)
+    page_contents = re.sub(":term:`([\w\s-]+?)`", term_link, page_contents)
+    page_contents = re.sub("figure:: (.+?)", figure_link, page_contents)
+    page_contents = re.sub("image:: (.+?)", image_link, page_contents)
 
     # Disable include directives and raw for security. They are already disabled
     # by docutils.py, this is just to be extra careful.