diff --git a/anno/anno/render.py b/anno/anno/render.py
index ee01eb1..6df14c7 100644
--- a/anno/anno/render.py
+++ b/anno/anno/render.py
@@ -10,9 +10,12 @@
 
 # -----------------------------------------------------------------------------
 
+
 def render_markdown(value):
     extra_args = ['--katex']
-    output = pypandoc.convert_text(value, to='html5', format='md',
+    output = pypandoc.convert_text(value,
+                                   to='html5',
+                                   format='md',
                                    extra_args=extra_args)
     return output
 
@@ -32,21 +35,72 @@ def make_pdf(note):
                           outputfile=note.pdf_fname)
 
 
-def parse_frontmatter(text):
-    if not text.startswith('---'):
-        raise ValueError('Files must contain frontmatter with title and date.')
-    try:
-        # Credit: https://github.com/eyeseast/python-frontmatter/
-        FM_BOUNDARY = re.compile(r'^-{3,}\s*$', re.MULTILINE)
-        _, fm, content = FM_BOUNDARY.split(text, 2)
-        parts = [p for p in fm.split('\n') if p]
-        meta = {x[0].strip(): x[1].strip()
-                for x in [x.split(':', 1) for x in parts]}
-        return meta
-    except ValueError:
-        raise ValueError('Error parsing frontmatter.')
-
-
 def jinja2_filter_date_to_string(date_str):
     date = datetime.datetime.strptime(date_str, '%Y-%m-%d')
     return date.strftime(c.datefmt)
+
+
+# This function and regular expressions are taken from markdown2:
+#
+#     https://github.com/trentm/python-markdown2/blob/
+#       4d2fc792abd7fbf8ddec937812857f13fded61cf/lib/markdown2.py#L453
+#
+# We aren't using this library because it's not clear from the docs how to just
+# extract the metadata.
+#
+# -----------------------------------------------------------------------------
+
+_meta_data_pattern = re.compile(
+    r'^(?:---[\ \t]*\n)?'
+    r'(.*:\s+>\n\s+[\S\s]+?)'
+    r'(?=\n\w+\s*:\s*\w+\n|\Z)|([\S\w]+\s*:(?! >)'
+    r'[ \t]*.*\n?)(?:---[\ \t]*\n)?',
+    re.MULTILINE)
+_key_val_pat = re.compile(r"[\S\w]+\s*:(?! >)[ \t]*.*\n?", re.MULTILINE)
+# this allows key: >
+#                   value
+#                   continues over multiple lines
+_key_val_block_pat = re.compile(
+    "(.*:\s+>\n\s+[\S\s]+?)(?=\n\w+\s*:\s*\w+\n|\Z)", re.MULTILINE)
+_meta_data_fence_pattern = re.compile(r'^---[\ \t]*\n', re.MULTILINE)
+_meta_data_newline = re.compile("^\n", re.MULTILINE)
+
+
+def _extract_metadata(text):
+
+    # GWG: Sometimes extra `\r` characters are injected in the text. I am not
+    #      sure why.
+    text = text.replace('\r', '')
+
+    metadata = {}
+    if text.startswith("---"):
+        fence_splits = re.split(_meta_data_fence_pattern, text,
+                                maxsplit=2)
+        metadata_content = fence_splits[1]
+        match = re.findall(_meta_data_pattern, metadata_content)
+        if not match:
+            return text
+    else:
+        metadata_split = re.split(_meta_data_newline, text, maxsplit=1)
+        metadata_content = metadata_split[0]
+        match = re.findall(_meta_data_pattern, metadata_content)
+        if not match:
+            return text
+
+    kv = re.findall(_key_val_pat, metadata_content)
+    kvm = re.findall(_key_val_block_pat, metadata_content)
+    kvm = [item.replace(": >\n", ":", 1) for item in kvm]
+
+    for item in kv + kvm:
+        k, v = item.split(":", 1)
+        metadata[k.strip()] = v.strip()
+
+    return metadata
+
+
+def parse_frontmatter(text):
+    try:
+        metadata = _extract_metadata(text)
+        return metadata
+    except (ValueError, IndexError) as e:
+        raise ValueError(f'Error parsing frontmatter: {str(e)}.')