diff --git a/src/calibre/ebooks/metadata/epub.py b/src/calibre/ebooks/metadata/epub.py
index 1f3141b113..2a7e0a3db6 100644
--- a/src/calibre/ebooks/metadata/epub.py
+++ b/src/calibre/ebooks/metadata/epub.py
@@ -17,11 +17,11 @@ from calibre.ebooks.metadata.opf import get_metadata as get_metadata_from_opf
 from calibre.ebooks.metadata.opf import set_metadata as set_metadata_opf
 from calibre.ebooks.metadata.opf2 import OPF
 from calibre.ptempfile import TemporaryDirectory
+from calibre.utils.imghdr import what as what_image_type
 from calibre.utils.localunzip import LocalZipFile
 from calibre.utils.xml_parse import safe_xml_fromstring
 from calibre.utils.zipfile import BadZipfile, ZipFile, safe_replace
 
-
 class EPubException(Exception):
     pass
 
@@ -36,7 +36,7 @@ class ContainerException(OCFException):
 
 class Container(dict):
 
-    def __init__(self, stream=None):
+    def __init__(self, stream=None, file_exists=None):
         if not stream:
             return
         container = safe_xml_fromstring(stream.read())
@@ -49,6 +49,12 @@ class Container(dict):
             mt, fp = rootfile.get('media-type'), rootfile.get('full-path')
             if not mt or not fp:
                 raise EPubException("<rootfile/> element malformed")
+
+            if file_exists and not file_exists(fp):
+                # Some Kobo epubs have multiple rootfile entries, but only one
+                # exists.  Ignore the ones that don't exist.
+                continue
+
             self[mt] = fp
 
 
@@ -95,7 +101,7 @@ class OCFReader(OCF):
 
         try:
             with closing(self.open(OCF.CONTAINER_PATH)) as f:
-                self.container = Container(f)
+                self.container = Container(f, self.exists)
         except KeyError:
             raise EPubException("missing OCF container.xml file")
         self.opf_path = self.container[OPF.MIMETYPE]
@@ -125,6 +131,14 @@ class OCFReader(OCF):
     def read_bytes(self, name):
         return self.open(name).read()
 
+    def exists(self, path):
+        try:
+            self.open(path)
+            return True
+        except OSError:
+            return False
+
+
 
 class OCFZipReader(OCFReader):
 
@@ -153,6 +167,13 @@ class OCFZipReader(OCFReader):
     def read_bytes(self, name):
         return self.archive.read(name)
 
+    def exists(self, path):
+        try:
+            self.archive.getinfo(path)
+            return True
+        except KeyError:
+            return False
+
 
 def get_zip_reader(stream, root=None):
     try:
@@ -192,6 +213,18 @@ def render_cover(cpage, zf, reader=None):
             cpage = os.path.join(tdir, cpage)
             if not os.path.exists(cpage):
                 return
+
+            # In the case of manga, the first spine item may be an image
+            # already, so treat it as a raster cover.
+            file_format = what_image_type(cpage)
+            if file_format == "jpeg":
+                # Only JPEG is allowed since elsewhere we assume raster covers
+                # are JPEG.  In principle we could convert other image formats
+                # but this is already an out-of-spec case that happens to
+                # arise in books from some stores.
+                with open(cpage, "rb") as source:
+                    return source.read()
+
             return render_html_svg_workaround(cpage, default_log, root=tdir)