diff --git a/compose/config/config.py b/compose/config/config.py
index 5bc534fe9..ff3ae780c 100644
--- a/compose/config/config.py
+++ b/compose/config/config.py
@@ -457,7 +457,7 @@ def parse_environment(environment):
 
 def split_env(env):
     if isinstance(env, six.binary_type):
-        env = env.decode('utf-8')
+        env = env.decode('utf-8', 'replace')
     if '=' in env:
         return env.split('=', 1)
     else:
diff --git a/compose/utils.py b/compose/utils.py
index c8fddc5f1..08f6034fc 100644
--- a/compose/utils.py
+++ b/compose/utils.py
@@ -95,7 +95,7 @@ def stream_as_text(stream):
     """
     for data in stream:
         if not isinstance(data, six.text_type):
-            data = data.decode('utf-8')
+            data = data.decode('utf-8', 'replace')
         yield data
 
 
diff --git a/tests/unit/utils_test.py b/tests/unit/utils_test.py
index b272c7349..e3d0bc00b 100644
--- a/tests/unit/utils_test.py
+++ b/tests/unit/utils_test.py
@@ -1,3 +1,6 @@
+# encoding: utf-8
+from __future__ import unicode_literals
+
 from .. import unittest
 from compose import utils
 
@@ -14,3 +17,16 @@ class JsonSplitterTestCase(unittest.TestCase):
             utils.json_splitter(data),
             ({'foo': 'bar'}, '{"next": "obj"}')
         )
+
+
+class StreamAsTextTestCase(unittest.TestCase):
+
+    def test_stream_with_non_utf_unicode_character(self):
+        stream = [b'\xed\xf3\xf3']
+        output, = utils.stream_as_text(stream)
+        assert output == '���'
+
+    def test_stream_with_utf_character(self):
+        stream = ['ěĝ'.encode('utf-8')]
+        output, = utils.stream_as_text(stream)
+        assert output == 'ěĝ'