Add avfilter to transcode_aac to respect frame size.

avutil/avutil.ml

@@ -681,8 +681,17 @@ let add_audio_opts ?channels ?channel_layout ~sample_rate ~sample_format
   Hashtbl.add opts "ar" (`Int sample_rate);
   on_opt channels (fun channels -> Hashtbl.add opts "ac" (`Int channels));
   on_opt channel_layout (fun channel_layout ->
-      Hashtbl.add opts "channel_layout"
-        (`String (Channel_layout.get_description channel_layout)));
+      let param =
+        match Channel_layout.get_native_id channel_layout with
+          | Some id -> `Int64 id
+          | None ->
+              let channel_layout =
+                Channel_layout.get_default
+                  (Channel_layout.get_nb_channels channel_layout)
+              in
+              `Int64 (Option.get (Channel_layout.get_native_id channel_layout))
+      in
+      Hashtbl.add opts "channel_layout" param);
   Hashtbl.add opts "sample_fmt" (`Int (Sample_format.get_id sample_format));
   Hashtbl.add opts "time_base" (`String (string_of_rational time_base))
 

examples/dune

@@ -96,7 +96,7 @@
 (executable
  (name transcode_aac)
  (modules transcode_aac)
- (libraries ffmpeg-av))
+ (libraries ffmpeg-av ffmpeg-avfilter))
 
 (executable
  (name transcoding)

examples/transcode_aac.ml

@@ -5,13 +5,29 @@ let () =
 
   if Array.length Sys.argv = 4 && Sys.argv.(3) = "macosx" then exit 0;
 
-  Avutil.Log.set_level `Debug;
-  Avutil.Log.set_callback print_string;
+  Avutil.Log.set_level `Verbose;
 
   let src = Av.open_input Sys.argv.(1) in
 
   let idx, istream, params = Av.find_best_audio_stream src in
 
+  let () =
+    match Avcodec.descriptor params with
+      | Some { name = d; _ } ->
+          Printf.printf
+            "Input format: %s (channel layout: %s, sample format: %s, sample \
+             rate: %d)\n\
+             %!"
+            d
+            (Avutil.Channel_layout.get_description
+               (Avcodec.Audio.get_channel_layout params))
+            (Option.value ~default:"(unknown)"
+               (Avutil.Sample_format.get_name
+                  (Avcodec.Audio.get_sample_format params)))
+            (Avcodec.Audio.get_sample_rate params)
+      | _ -> Printf.printf "Unknown input format..\n!"
+  in
+
   let codec = Avcodec.Audio.find_encoder_by_name "aac" in
 
   let channel_layout = Avcodec.Audio.get_channel_layout params in
@@ -25,10 +41,69 @@ let () =
          ~time_base ~codec
   in
 
+  let () =
+    match Avcodec.descriptor (Av.get_codec_params ostream) with
+      | Some { name = d; _ } ->
+          Printf.printf
+            "Output format: %s (channel layout: %s, sample format: %s, sample \
+             rate: %d)\n\
+             %!"
+            d
+            (Avutil.Channel_layout.get_description
+               (Avcodec.Audio.get_channel_layout params))
+            (Option.value ~default:"(unknown)"
+               (Avutil.Sample_format.get_name
+                  (Avcodec.Audio.get_sample_format params)))
+            (Avcodec.Audio.get_sample_rate params)
+      | _ -> Printf.printf "Unknown output format..\n!"
+  in
+
+  let filter = ref None in
+  let get_filter frame =
+    match !filter with
+      | Some f -> f
+      | None ->
+          let in_params =
+            {
+              Avfilter.Utils.sample_rate =
+                Avutil.Audio.frame_get_sample_rate frame;
+              channel_layout = Avutil.Audio.frame_get_channel_layout frame;
+              sample_format = Avutil.Audio.frame_get_sample_format frame;
+            }
+          in
+          let in_time_base = { Avutil.num = 1; den = sample_rate } in
+          let out_frame_size =
+            if List.mem `Variable_frame_size (Avcodec.capabilities codec) then
+              512
+            else Av.get_frame_size ostream
+          in
+          let out_params =
+            { Avfilter.Utils.sample_rate; sample_format; channel_layout }
+          in
+          let f =
+            Avfilter.Utils.init_audio_converter ~in_params ~in_time_base
+              ~out_params ~out_frame_size ()
+          in
+          filter := Some f;
+          f
+  in
+
+  let pts = ref 0L in
+  let on_frame frame =
+    Avutil.Frame.set_pts frame (Some !pts);
+    pts := Int64.add !pts (Int64.of_int (Avutil.Audio.frame_nb_samples frame));
+    Av.write_frame ostream frame
+  in
+
+  let write_frame frame =
+    let filter = get_filter frame in
+    Avfilter.Utils.convert_audio filter on_frame (`Frame frame)
+  in
+
   let rec f () =
     match Av.read_input ~audio_frame:[istream] src with
       | `Audio_frame (i, frame) when i = idx ->
-          Av.write_frame ostream frame;
+          write_frame frame;
           f ()
       | exception Avutil.Error `Eof -> ()
       | _ -> f ()