Enable spatial audio metadata injection for mp4 files.

spatialmedia/gui.py

@@ -60,13 +60,20 @@ def action_open(self):
         self.set_message("Opened file: %s\n" % ntpath.basename(self.in_file))
 
         console = Console()
-        metadata = spherical.parse_metadata(self.in_file, console.append)
+        parsedMetadata = spherical.parse_metadata(self.in_file, console.append)
+
+        metadata = None
+        audio_metadata = None
+        if parsedMetadata:
+            metadata = parsedMetadata.video
+            audio_metadata = parsedMetadata.audio
 
         for line in console.log:
             if "Error" in line:
                 self.set_error("Failed to load file %s"
                                % ntpath.basename(self.in_file))
                 self.var_spherical.set(0)
+                self.var_spatial_audio.set(0)
                 self.disable_state()
                 self.button_open.configure(state="normal")
                 self.button_quit.configure(state="normal")
@@ -75,10 +82,18 @@ def action_open(self):
         self.enable_state()
         self.checkbox_spherical.configure(state="normal")
 
+        infile = os.path.abspath(self.in_file)
+        file_extension = os.path.splitext(infile)[1].lower()
+        self.enable_spatial_audio =\
+            True if (file_extension == ".mp4") else False
+
         if not metadata:
             self.var_spherical.set(0)
             self.var_3d.set(0)
 
+        if not audio_metadata:
+            self.var_spatial_audio.set(0)
+
         if metadata:
             metadata = metadata.itervalues().next()
             self.var_spherical.set(1)
@@ -93,6 +108,12 @@ def action_open(self):
             else:
                 self.var_3d.set(0)
 
+        if audio_metadata:
+            self.var_spatial_audio.set(1)
+            self.options_ambisonics["text"] =\
+                audio_metadata.get_metadata_string()
+
+
         self.update_state()
 
     def action_inject_delay(self):
@@ -103,8 +124,19 @@ def action_inject_delay(self):
         xml = spherical.generate_spherical_xml(stereo=stereo)
 
         console = Console()
+        c = console.append
+
+        metadata = spherical.Metadata()
+        metadata.video = xml
+
+        if self.var_spatial_audio.get():
+            # Default ambisonics audio metadata
+            audio_metadata = {'ambisonic_order': 1,
+                              'ambisonic_type': 'periphonic'}
+            metadata.audio = audio_metadata
+
         spherical.inject_metadata(
-            self.in_file, self.save_file, xml, console.append)
+            self.in_file, self.save_file, metadata, console.append)
         self.set_message("Successfully saved file to %s\n"
                          % ntpath.basename(self.save_file))
         self.button_open.configure(state="normal")
@@ -131,6 +163,9 @@ def action_inject(self):
     def action_set_spherical(self):
         self.update_state()
 
+    def action_set_spatial_audio(self):
+        self.update_state()
+
     def action_set_3d(self):
         self.update_state()
 
@@ -140,8 +175,10 @@ def enable_state(self):
 
     def disable_state(self):
         self.checkbox_spherical.configure(state="disabled")
+        self.checkbox_spatial_audio.configure(state="disabled")
         self.checkbox_3D.configure(state="disabled")
         self.options_projection.configure(state="disabled")
+        self.options_ambisonics.configure(state="disabled")
         self.button_inject.configure(state="disabled")
         self.button_open.configure(state="disabled")
         self.button_quit.configure(state="disabled")
@@ -152,10 +189,22 @@ def update_state(self):
             self.checkbox_3D.configure(state="normal")
             self.options_projection.configure(state="normal")
             self.button_inject.configure(state="normal")
+            if self.enable_spatial_audio:
+                self.checkbox_spatial_audio.configure(state="normal")
+                if self.var_spatial_audio.get():
+                    self.options_ambisonics.configure(state="normal")
+                else:
+                    self.options_ambisonics.configure(state="disable")
         else:
             self.checkbox_3D.configure(state="disabled")
             self.options_projection.configure(state="disabled")
             self.button_inject.configure(state="disabled")
+            self.checkbox_spatial_audio.configure(state="disabled")
+            if self.var_spatial_audio.get():
+                self.options_ambisonics.configure(state="normal")
+            else:
+                self.options_ambisonics.configure(state="disable")
+            self.options_ambisonics.configure(state="disabled")
 
     def set_error(self, text):
         self.label_message["text"] = text
@@ -185,6 +234,20 @@ def create_widgets(self):
         self.checkbox_spherical["command"] = self.action_set_spherical
         self.checkbox_spherical.grid(row=row, column=column, padx=14, pady=2)
 
+        # Spatial Audio Checkbox
+        row += 1
+        column = 0
+        self.label_spatial_audio = Label(self)
+        self.label_spatial_audio["text"] = "Spatial Audio"
+        self.label_spatial_audio.grid(row=row, column=column)
+
+        column += 1
+        self.var_spatial_audio = IntVar()
+        self.checkbox_spatial_audio = \
+           Checkbutton(self, variable=self.var_spatial_audio)
+        self.checkbox_spatial_audio["command"] = self.action_set_spatial_audio
+        self.checkbox_spatial_audio.grid(row=row, column=column, padx=0, pady=0)
+
         # 3D
         column = 0
         row = row + 1
@@ -210,6 +273,18 @@ def create_widgets(self):
         self.options_projection["text"] = "Equirectangular"
         self.options_projection.grid(row=row, column=column, padx=14, pady=2)
 
+        # Ambisonics Type
+        column = 0
+        row = row + 1
+        self.label_ambisonics = Label(self)
+        self.label_ambisonics["text"] = "Ambisonics Type"
+        self.label_ambisonics.grid(row=row, column=column, padx=14, pady=2)
+        column += 1
+
+        self.options_ambisonics = Label(self)
+        self.options_ambisonics["text"] = "1st Order, ACN, SN3D, Periphonic"
+        self.options_ambisonics.grid(row=row, column=column, padx=14, pady=2)
+
         # Message Box
         row = row + 1
         column = 0
@@ -261,7 +336,7 @@ def __init__(self, master=None):
         master.attributes("-topmost", True)
         master.focus_force()
         self.after(50, lambda: master.attributes("-topmost", False))
-
+        self.enable_spatial_audio = False
 
 def main():
     root = Tk()

spatialmedia/mpeg/__init__.py

@@ -15,6 +15,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import spatialmedia.mpeg.sa3d
 import spatialmedia.mpeg.box
 import spatialmedia.mpeg.constants
 import spatialmedia.mpeg.container
@@ -23,7 +24,8 @@
 load = mpeg4_container.load
 
 Box = box.Box
+SA3DBox = sa3d.SA3DBox
 Container = container.Container
 Mpeg4Container = mpeg4_container.Mpeg4Container
 
-__all__ = ["box", "mpeg4", "container", "constants"]
+__all__ = ["box", "mpeg4", "container", "constants", "sa3d"]

spatialmedia/mpeg/constants.py

@@ -30,22 +30,29 @@
 TAG_XML = "xml "
 TAG_HDLR = "hdlr"
 TAG_FTYP = "ftyp"
+TAG_ESDS = "esds"
+TAG_SOUN = "soun"
 
 # Container types.
 TAG_MOOV = "moov"
 TAG_UDTA = "udta"
 TAG_META = "meta"
 TAG_TRAK = "trak"
 TAG_MDIA = "mdia"
+TAG_MP4A = "mp4a"
 TAG_MINF = "minf"
 TAG_STBL = "stbl"
+TAG_STSD = "stsd"
 TAG_UUID = "uuid"
+TAG_SA3D = "SA3D"
 
 CONTAINERS_LIST = [
     TAG_MDIA,
     TAG_MINF,
+    TAG_MP4A,
     TAG_MOOV,
     TAG_STBL,
+    TAG_STSD,
     TAG_TRAK,
     TAG_UDTA,
     ]

spatialmedia/mpeg/container.py

@@ -25,7 +25,7 @@
 
 from spatialmedia.mpeg import box
 from spatialmedia.mpeg import constants
-
+from spatialmedia.mpeg import sa3d
 
 def load(fh, position, end):
     if position is None:
@@ -37,7 +37,10 @@ def load(fh, position, end):
     name = fh.read(4)
 
     if name not in constants.CONTAINERS_LIST:
-        return box.load(fh, position, end)
+        if name == constants.TAG_SA3D:
+            return sa3d.load(fh, position, end)
+        else:
+            return box.load(fh, position, end)
 
     if size == 1:
         size = struct.unpack(">Q", fh.read(8))[0]
@@ -51,13 +54,30 @@ def load(fh, position, end):
         print "Error: Container box size exceeds bounds."
         return None
 
+    padding = 0
+    stsd_version = 0
+    if (name == constants.TAG_STSD):
+        padding = 8
+
+    if (name == constants.TAG_MP4A):
+        current_pos = fh.tell()
+        fh.seek(current_pos + 8)
+        sample_description_version = struct.unpack(">h", fh.read(2))[0]
+        fh.seek(current_pos)
+
+        if sample_description_version == 1:
+          padding = 28+16 # Mov
+        else:
+          padding = 28 # Mp4
+
     new_box = Container()
     new_box.name = name
     new_box.position = position
     new_box.header_size = header_size
     new_box.content_size = size - header_size
+    new_box.padding = padding
     new_box.contents = load_multiple(
-        fh, position + header_size, position + size)
+        fh, position + header_size + padding, position + size)
 
     if new_box.contents is None:
         return None
@@ -81,16 +101,17 @@ def load_multiple(fh, position=None, end=None):
 class Container(box.Box):
     """MPEG4 container box contents / behaviour."""
 
-    def __init__(self):
+    def __init__(self, padding=0):
         self.name = ""
         self.position = 0
         self.header_size = 0
         self.content_size = 0
         self.contents = list()
+        self.padding = padding
 
     def resize(self):
         """Recomputes the box size and recurses on contents."""
-        self.content_size = 0
+        self.content_size = self.padding
         for element in self.contents:
             if isinstance(element, Container):
                 element.resize()
@@ -176,5 +197,9 @@ def save(self, in_fh, out_fh, delta):
             out_fh.write(struct.pack(">I", self.size()))
             out_fh.write(self.name)
 
+        if self.padding > 0:
+            in_fh.seek(self.content_start())
+            box.tag_copy(in_fh, out_fh, self.padding)
+
         for element in self.contents:
             element.save(in_fh, out_fh, delta)

spatialmedia/mpeg/mpeg4_container.py

@@ -94,6 +94,7 @@ def __init__(self):
         self.first_mdat_box = None
         self.ftyp_box = None
         self.first_mdat_position = None
+        self.padding = 0
 
     def merge(self, element):
         """Mpeg4 containers do not support merging."""