[xcode] Add OPUS encoding profile for RTP Chromecasting

src/transcode.c

@@ -40,6 +40,7 @@
 #include "conffile.h"
 #include "db.h"
 #include "avio_evbuffer.h"
+#include "misc.h"
 #include "transcode.h"
 
 // Interval between ICY metadata checks for streams, in seconds
@@ -221,6 +222,17 @@ init_settings(struct settings_ctx *settings, enum transcode_profile profile)
 	settings->byte_depth = 2; // Bytes per sample = 16/8
 	break;
 
+      case XCODE_OPUS:
+	settings->encode_audio = 1;
+	settings->format = "data"; // Means we get the raw packet from the encoder, no muxing
+	settings->audio_codec = AV_CODEC_ID_OPUS;
+	settings->sample_rate = 48000;
+	settings->channel_layout = AV_CH_LAYOUT_STEREO;
+	settings->channels = 2;
+	settings->sample_format = AV_SAMPLE_FMT_S16; // Only libopus support
+	settings->byte_depth = 2; // Bytes per sample = 16/8
+	break;
+
       case XCODE_JPEG:
 	settings->encode_video = 1;
 	settings->silent = 1;
@@ -1383,7 +1395,7 @@ transcode_cleanup(struct transcode_ctx **ctx)
 /*                       Encoding, decoding and transcoding                  */
 
 int
-transcode_decode(void **frame, struct decode_ctx *dec_ctx)
+transcode_decode(transcode_frame **frame, struct decode_ctx *dec_ctx)
 {
   struct transcode_ctx ctx;
   int ret;
@@ -1414,7 +1426,7 @@ transcode_decode(void **frame, struct decode_ctx *dec_ctx)
 
 // Filters and encodes
 int
-transcode_encode(struct evbuffer *evbuf, struct encode_ctx *ctx, void *frame, int eof)
+transcode_encode(struct evbuffer *evbuf, struct encode_ctx *ctx, transcode_frame *frame, int eof)
 {
   AVFrame *f = frame;
   struct stream_ctx *s;
@@ -1489,8 +1501,8 @@ transcode(struct evbuffer *evbuf, int *icy_timer, struct transcode_ctx *ctx, int
   return processed;
 }
 
-void *
+transcode_frame *
-transcode_frame_new(enum transcode_profile profile, uint8_t *data, size_t size)
+transcode_frame_new(enum transcode_profile profile, void *data, size_t size)
 {
   AVFrame *f;
   int ret;
@@ -1502,7 +1514,7 @@ transcode_frame_new(enum transcode_profile profile, uint8_t *data, size_t size)
       return NULL;
     }
 
-  f->nb_samples     = size / 4;
+  f->nb_samples     = BTOS(size);
   f->format         = AV_SAMPLE_FMT_S16;
   f->channel_layout = AV_CH_LAYOUT_STEREO;
 #ifdef HAVE_FFMPEG
@@ -1511,7 +1523,9 @@ transcode_frame_new(enum transcode_profile profile, uint8_t *data, size_t size)
   f->pts            = AV_NOPTS_VALUE;
   f->sample_rate    = 44100;
 
-  ret = avcodec_fill_audio_frame(f, 2, f->format, data, size, 0);
+  // We don't align because the frame won't be given directly to the encoder
+  // anyway, it will first go through the filter (which might align it...?)
+  ret = avcodec_fill_audio_frame(f, 2, f->format, data, size, 1);
   if (ret < 0)
     {
       DPRINTF(E_LOG, L_XCODE, "Error filling frame with rawbuf: %s\n", err2str(ret));
@@ -1523,7 +1537,7 @@ transcode_frame_new(enum transcode_profile profile, uint8_t *data, size_t size)
 }
 
 void
-transcode_frame_free(void *frame)
+transcode_frame_free(transcode_frame *frame)
 {
   AVFrame *f = frame;
 

src/transcode.h

@@ -14,6 +14,8 @@ enum transcode_profile
   XCODE_PCM16_HEADER,
   // Transcodes the best audio stream into MP3
   XCODE_MP3,
+  // Transcodes the best audio stream into OPUS
+  XCODE_OPUS,
   // Transcodes the best video stream into JPEG/PNG
   XCODE_JPEG,
   XCODE_PNG,
@@ -23,6 +25,8 @@ struct decode_ctx;
 struct encode_ctx;
 struct transcode_ctx;
 
+typedef void transcode_frame;
+
 // Setting up
 struct decode_ctx *
 transcode_decode_setup(enum transcode_profile profile, enum data_kind data_kind, const char *path, struct evbuffer *evbuf, uint32_t song_length);
@@ -60,7 +64,7 @@ transcode_cleanup(struct transcode_ctx **ctx);
  * @return         Positive if OK, negative if error, 0 if EOF
  */
 int
-transcode_decode(void **frame, struct decode_ctx *ctx);
+transcode_decode(transcode_frame **frame, struct decode_ctx *ctx);
 
 /* Encodes and remuxes a frame. Also resamples if needed.
  *
@@ -71,7 +75,7 @@ transcode_decode(void **frame, struct decode_ctx *ctx);
  * @return         Bytes added if OK, negative if error
  */
 int
-transcode_encode(struct evbuffer *evbuf, struct encode_ctx *ctx, void *frame, int eof);
+transcode_encode(struct evbuffer *evbuf, struct encode_ctx *ctx, transcode_frame *frame, int eof);
 
 /* Demuxes, decodes, encodes and remuxes from the input.
  *
@@ -87,17 +91,18 @@ int
 transcode(struct evbuffer *evbuf, int *icy_timer, struct transcode_ctx *ctx, int want_bytes);
 
 /* Converts a buffer with raw data to a frame that can be passed directly to the
- * transcode_encode() function
+ * transcode_encode() function. It does not copy, so if you free the data the
+ * frame will become invalid.
  *
  * @in  profile    Tells the function what kind of frame to create
  * @in  data       Buffer with raw data
  * @in  size       Size of buffer
  * @return         Opaque pointer to frame if OK, otherwise NULL
  */
-void *
+transcode_frame *
-transcode_frame_new(enum transcode_profile profile, uint8_t *data, size_t size);
+transcode_frame_new(enum transcode_profile profile, void *data, size_t size);
 void
-transcode_frame_free(void *frame);
+transcode_frame_free(transcode_frame *frame);
 
 /* Seek to the specified position - next transcode() will return this packet
  *

src/transcode_legacy.c

@@ -39,6 +39,7 @@
 #include "conffile.h"
 #include "db.h"
 #include "avio_evbuffer.h"
+#include "misc.h"
 #include "transcode.h"
 
 // Interval between ICY metadata checks for streams, in seconds
@@ -162,6 +163,16 @@ init_profile(struct encode_ctx *ctx, enum transcode_profile profile)
 	ctx->byte_depth = 2; // Bytes per sample = 16/8
 	return 0;
 
+      case XCODE_OPUS:
+	ctx->format = "data"; // Means we get the raw packet from the encoder, no muxing
+	ctx->audio_codec = AV_CODEC_ID_OPUS;
+	ctx->sample_rate = 48000;
+	ctx->channel_layout = AV_CH_LAYOUT_STEREO;
+	ctx->channels = 2;
+	ctx->sample_format = AV_SAMPLE_FMT_S16; // Only libopus support
+	ctx->byte_depth = 2; // Bytes per sample = 16/8
+	return 0;
+
       case XCODE_MP3:
 	ctx->format = "mp3";
 	ctx->audio_codec = AV_CODEC_ID_MP3;
@@ -1337,7 +1348,7 @@ transcode_cleanup(struct transcode_ctx **ctx)
 }
 
 void
-transcode_frame_free(void *frame)
+transcode_frame_free(transcode_frame *frame)
 {
   struct decoded_frame *decoded = frame;
 
@@ -1350,7 +1361,7 @@ transcode_frame_free(void *frame)
 
 
 int
-transcode_decode(void **frame, struct decode_ctx *ctx)
+transcode_decode(transcode_frame **frame, struct decode_ctx *ctx)
 {
   struct decoded_frame *decoded;
   AVPacket packet;
@@ -1450,7 +1461,7 @@ transcode_decode(void **frame, struct decode_ctx *ctx)
 
 // Filters and encodes
 int
-transcode_encode(struct evbuffer *evbuf, struct encode_ctx *ctx, void *frame, int eof)
+transcode_encode(struct evbuffer *evbuf, struct encode_ctx *ctx, transcode_frame *frame, int eof)
 {
   struct decoded_frame *decoded = frame;
   int stream_index;
@@ -1486,7 +1497,7 @@ transcode_encode(struct evbuffer *evbuf, struct encode_ctx *ctx, void *frame, in
 int
 transcode(struct evbuffer *evbuf, int *icy_timer, struct transcode_ctx *ctx, int want_bytes)
 {
-  void *frame;
+  transcode_frame *frame;
   int processed;
   int ret;
 
@@ -1515,8 +1526,8 @@ transcode(struct evbuffer *evbuf, int *icy_timer, struct transcode_ctx *ctx, int
   return processed;
 }
 
-void *
+transcode_frame *
-transcode_frame_new(enum transcode_profile profile, uint8_t *data, size_t size)
+transcode_frame_new(enum transcode_profile profile, void *data, size_t size)
 {
   struct decoded_frame *decoded;
   AVFrame *f;
@@ -1540,7 +1551,7 @@ transcode_frame_new(enum transcode_profile profile, uint8_t *data, size_t size)
   decoded->stream_index = 0;
   decoded->frame = f;
 
-  f->nb_samples     = size / 4;
+  f->nb_samples     = BTOS(size);
   f->format         = AV_SAMPLE_FMT_S16;
   f->channel_layout = AV_CH_LAYOUT_STEREO;
 #ifdef HAVE_FFMPEG
@@ -1549,7 +1560,7 @@ transcode_frame_new(enum transcode_profile profile, uint8_t *data, size_t size)
   f->pts            = AV_NOPTS_VALUE;
   f->sample_rate    = 44100;
 
-  ret = avcodec_fill_audio_frame(f, 2, f->format, data, size, 0);
+  ret = avcodec_fill_audio_frame(f, 2, f->format, data, size, 1);
   if (ret < 0)
     {
       DPRINTF(E_LOG, L_XCODE, "Error filling frame with rawbuf: %s\n", err2str(ret));