diff --git a/changelogs/fragments/template_output_encoding.yml b/changelogs/fragments/template_output_encoding.yml
new file mode 100644
index 0000000000..3185a9aa75
--- /dev/null
+++ b/changelogs/fragments/template_output_encoding.yml
@@ -0,0 +1,5 @@
+---
+minor_changes:
+- Explicit encoding for the output of the template module, to be able
+  to generate non-utf8 files from a utf-8 template.
+  (https://github.com/ansible/proposals/issues/121)
\ No newline at end of file
diff --git a/lib/ansible/modules/files/template.py b/lib/ansible/modules/files/template.py
index a7c879f430..3b34440961 100644
--- a/lib/ansible/modules/files/template.py
+++ b/lib/ansible/modules/files/template.py
@@ -106,6 +106,13 @@ options:
         may be specified as a symbolic mode (for example, C(u+rwx) or C(u=rw,g=r,o=r)).  As of
         version 2.6, the mode may also be the special string C(preserve).  C(preserve) means that
         the file will be given the same permissions as the source file."
+  output_encoding:
+    description:
+      - Overrides the encoding used to write the template file defined by C(dest).
+      - It defaults to C('utf-8'), but any encoding supported by python can be used.
+      - The source template file must always be encoded using C('utf-8'), for homogeneity.
+    default: 'utf-8'
+    version_added: "2.7"
 notes:
   - For Windows you can use M(win_template) which uses '\\r\\n' as C(newline_sequence).
   - Including a string that uses a date in the template will result in the template being marked 'changed' each time
diff --git a/lib/ansible/plugins/action/template.py b/lib/ansible/plugins/action/template.py
index 5cdb2f5271..49b65536ba 100644
--- a/lib/ansible/plugins/action/template.py
+++ b/lib/ansible/plugins/action/template.py
@@ -58,6 +58,7 @@ class ActionModule(ActionBase):
         block_end_string = self._task.args.get('block_end_string', None)
         trim_blocks = boolean(self._task.args.get('trim_blocks', True), strict=False)
         lstrip_blocks = boolean(self._task.args.get('lstrip_blocks', False), strict=False)
+        output_encoding = self._task.args.get('output_encoding', 'utf-8') or 'utf-8'
 
         # Option `lstrip_blocks' was added in Jinja2 version 2.7.
         if lstrip_blocks:
@@ -176,13 +177,14 @@ class ActionModule(ActionBase):
             new_task.args.pop('variable_end_string', None)
             new_task.args.pop('trim_blocks', None)
             new_task.args.pop('lstrip_blocks', None)
+            new_task.args.pop('output_encoding', None)
 
             local_tempdir = tempfile.mkdtemp(dir=C.DEFAULT_LOCAL_TMP)
 
             try:
                 result_file = os.path.join(local_tempdir, os.path.basename(source))
                 with open(to_bytes(result_file, errors='surrogate_or_strict'), 'wb') as f:
-                    f.write(to_bytes(resultant, errors='surrogate_or_strict'))
+                    f.write(to_bytes(resultant, encoding=output_encoding, errors='surrogate_or_strict'))
 
                 new_task.args.update(
                     dict(
diff --git a/test/integration/targets/template/files/encoding_1252_utf-8.expected b/test/integration/targets/template/files/encoding_1252_utf-8.expected
new file mode 100644
index 0000000000..0d3cc35251
--- /dev/null
+++ b/test/integration/targets/template/files/encoding_1252_utf-8.expected
@@ -0,0 +1 @@
+windows-1252 Special Characters: €‚ƒ„…†‡ˆ‰Š‹ŒŽ‘’“”•–—˜™š›œžŸ ¡¢£¤¥¦§¨©ª«¬­®¯°±²³´µ¶·¸¹º»¼½¾¿ÀÁÂÃÄÅÆÇÈÉÊËÌÍÎÏÐÑÒÓÔÕÖ×ØÙÚÛÜÝÞßàáâãäåæçèéêëìíîïðñòóôõö÷øùúûüýþÿ
diff --git a/test/integration/targets/template/files/encoding_1252_windows-1252.expected b/test/integration/targets/template/files/encoding_1252_windows-1252.expected
new file mode 100644
index 0000000000..7fb94a7b06
--- /dev/null
+++ b/test/integration/targets/template/files/encoding_1252_windows-1252.expected
@@ -0,0 +1 @@
+windows-1252 Special Characters: ���������������������������������������������������������������������������������������������������������������������������
diff --git a/test/integration/targets/template/tasks/main.yml b/test/integration/targets/template/tasks/main.yml
index 7e34852afa..87e2e7d16f 100644
--- a/test/integration/targets/template/tasks/main.yml
+++ b/test/integration/targets/template/tasks/main.yml
@@ -619,5 +619,28 @@
       - 'template_results.mode == "0547"'
       - 'stat_results.stat["mode"] == "0547"'
 
+# Test output_encoding
+- name: Prepare the list of encodings we want to check, including empty string for defaults
+  set_fact:
+    template_encoding_1252_encodings: ['', 'utf-8', 'windows-1252']
+
+- name: Copy known good encoding_1252_*.expected into place
+  copy:
+    src: 'encoding_1252_{{ item | default("utf-8", true) }}.expected'
+    dest: '{{ output_dir }}/encoding_1252_{{ item }}.expected'
+  loop: '{{ template_encoding_1252_encodings }}'
+
+- name: Generate the encoding_1252_* files from templates using various encoding combinations
+  template:
+    src: 'encoding_1252.j2'
+    dest: '{{ output_dir }}/encoding_1252_{{ item }}.txt'
+    output_encoding: '{{ item }}'
+  loop: '{{ template_encoding_1252_encodings }}'
+
+- name: Compare the encoding_1252_* templated files to known good
+  command: diff -u {{ output_dir }}/encoding_1252_{{ item }}.expected {{ output_dir }}/encoding_1252_{{ item }}.txt
+  register: encoding_1252_diff_result
+  loop: '{{ template_encoding_1252_encodings }}'
+
 # aliases file requires root for template tests so this should be safe
 - include: backup_test.yml
diff --git a/test/integration/targets/template/templates/encoding_1252.j2 b/test/integration/targets/template/templates/encoding_1252.j2
new file mode 100644
index 0000000000..0d3cc35251
--- /dev/null
+++ b/test/integration/targets/template/templates/encoding_1252.j2
@@ -0,0 +1 @@
+windows-1252 Special Characters: €‚ƒ„…†‡ˆ‰Š‹ŒŽ‘’“”•–—˜™š›œžŸ ¡¢£¤¥¦§¨©ª«¬­®¯°±²³´µ¶·¸¹º»¼½¾¿ÀÁÂÃÄÅÆÇÈÉÊËÌÍÎÏÐÑÒÓÔÕÖ×ØÙÚÛÜÝÞßàáâãäåæçèéêëìíîïðñòóôõö÷øùúûüýþÿ
diff --git a/test/sanity/code-smell/no-smart-quotes.py b/test/sanity/code-smell/no-smart-quotes.py
index 9c0e0a5091..672fb35f96 100755
--- a/test/sanity/code-smell/no-smart-quotes.py
+++ b/test/sanity/code-smell/no-smart-quotes.py
@@ -12,6 +12,9 @@ def main():
         'docs/docsite/rst/dev_guide/testing/sanity/no-smart-quotes.rst',
         'test/integration/targets/unicode/unicode.yml',
         'test/integration/targets/lookup_properties/lookup-8859-15.ini',
+        'test/integration/targets/template/files/encoding_1252_utf-8.expected',
+        'test/integration/targets/template/files/encoding_1252_windows-1252.expected',
+        'test/integration/targets/template/templates/encoding_1252.j2',
     ])
 
     for path in sys.argv[1:] or sys.stdin.read().splitlines():