diff --git a/src/calibre/ebooks/rb/__init__.py b/src/calibre/ebooks/rb/__init__.py
index acf9c04995..f45b966dad 100644
--- a/src/calibre/ebooks/rb/__init__.py
+++ b/src/calibre/ebooks/rb/__init__.py
@@ -6,7 +6,7 @@ __docformat__ = 'restructuredtext en'
 
 import os
 
-HEADER = '\xb0\x0c\xb0\x0c\x02\x00NUVO\x00\x00\x00\x00'
+HEADER = b'\xb0\x0c\xb0\x0c\x02\x00NUVO\x00\x00\x00\x00'
 
 
 class RocketBookError(Exception):
diff --git a/src/calibre/ebooks/rb/writer.py b/src/calibre/ebooks/rb/writer.py
index dc83476b39..9f4818725f 100644
--- a/src/calibre/ebooks/rb/writer.py
+++ b/src/calibre/ebooks/rb/writer.py
@@ -73,13 +73,13 @@ class RBWriter(object):
         out_stream.write(struct.pack('<I', page_count))
         offset = out_stream.tell() + (len(toc_items) * 44)
         for item in toc_items:
-            out_stream.write(item.name)
+            out_stream.write(item.name.encode('utf-8'))
             out_stream.write(struct.pack('<I', item.size))
             out_stream.write(struct.pack('<I', offset))
             out_stream.write(struct.pack('<I', item.flags))
             offset += item.size
 
-        out_stream.write(info[0][1])
+        out_stream.write(info[0][1].encode('utf-8'))
 
         self.log.debug('Writing compressed RB HTHML...')
         # Compressed text with proper heading
@@ -92,7 +92,10 @@ class RBWriter(object):
 
         self.log.debug('Writing images...')
         for item in hidx+images:
-            out_stream.write(item[1])
+            w = item[1]
+            if not isinstance(w, bytes):
+                w = w.encode('utf-8')
+            out_stream.write(w)
 
         total_size = out_stream.tell()
         out_stream.seek(0x1c)
@@ -104,7 +107,7 @@ class RBWriter(object):
         size = len(text)
 
         pages = []
-        for i in range(0, (len(text) + TEXT_RECORD_SIZE-1) / TEXT_RECORD_SIZE):
+        for i in range(0, (len(text) + TEXT_RECORD_SIZE-1) // TEXT_RECORD_SIZE):
             zobj = zlib.compressobj(9, zlib.DEFLATED, 13, 8, 0)
             pages.append(zobj.compress(text[i * TEXT_RECORD_SIZE : (i * TEXT_RECORD_SIZE) + TEXT_RECORD_SIZE]) + zobj.flush())