#include <windows.h>
#endif
#include "compat.h"
-#include "wav_reader.h"
-#include "caf_reader.h"
+#include "pcm_reader.h"
#include "aacenc.h"
#include "m4af.h"
#include "progress.h"
{
int i, sigs[] = { SIGINT, SIGHUP, SIGTERM };
for (i = 0; i < sizeof(sigs)/sizeof(sigs[0]); ++i) {
- struct sigaction sa = { 0 };
+ struct sigaction sa;
+ memset(&sa, 0, sizeof sa);
sa.sa_handler = signal_handler;
sa.sa_flags |= SA_RESTART;
sigaction(sigs[i], &sa, 0);
" -a, --afterburner <n> Afterburner\n"
" 0: Off\n"
" 1: On(default)\n"
-" -L, --lowdelay-sbr Enable ELD-SBR (AAC ELD only)\n"
+" -L, --lowdelay-sbr <-1|0|1> Configure SBR activity on AAC ELD\n"
+" -1: Use ELD SBR auto configurator\n"
+" 0: Disable SBR on ELD (default)\n"
+" 1: Enable SBR on ELD\n"
+" -s, --sbr-ratio <0|1|2> Controls activation of downsampled SBR\n"
+" 0: Use lib default (default)\n"
+" 1: downsampled SBR (default for ELD+SBR)\n"
+" 2: dual-rate SBR (default for HE-AAC)\n"
" -f, --transport-format <n> Transport format\n"
" 0: RAW (default, muxed into M4A)\n"
" 1: ADIF\n"
" 0: iTunSMPB (default)\n"
" 1: ISO standard (edts + sgpd)\n"
" 2: Both\n"
+" --include-sbr-delay Count SBR decoder delay in encoder delay\n"
+" This is not iTunes compatible, but is default\n"
+" behavior of FDK library.\n"
" -I, --ignorelength Ignore length of WAV header\n"
" -S, --silent Don't print progress messages\n"
" --moov-before-mdat Place moov box before mdat box on m4a output\n"
char *output_filename;
FILE *output_fp;
unsigned gapless_mode;
+ unsigned include_sbr_delay;
unsigned ignore_length;
int silent;
int moov_before_mdat;
int parse_options(int argc, char **argv, aacenc_param_ex_t *params)
{
int ch;
- unsigned n;
+ int n;
+#define OPT_INCLUDE_SBR_DELAY M4AF_FOURCC('s','d','l','y')
#define OPT_MOOV_BEFORE_MDAT M4AF_FOURCC('m','o','o','v')
#define OPT_RAW_CHANNELS M4AF_FOURCC('r','c','h','n')
#define OPT_RAW_RATE M4AF_FOURCC('r','r','a','t')
{ "bitrate-mode", required_argument, 0, 'm' },
{ "bandwidth", required_argument, 0, 'w' },
{ "afterburner", required_argument, 0, 'a' },
- { "lowdelay-sbr", no_argument, 0, 'L' },
+ { "lowdelay-sbr", required_argument, 0, 'L' },
+ { "sbr-ratio", required_argument, 0, 's' },
{ "transport-format", required_argument, 0, 'f' },
{ "adts-crc-check", no_argument, 0, 'C' },
{ "header-period", required_argument, 0, 'P' },
{ "gapless-mode", required_argument, 0, 'G' },
+ { "include-sbr-delay", no_argument, 0, OPT_INCLUDE_SBR_DELAY },
{ "ignorelength", no_argument, 0, 'I' },
{ "silent", no_argument, 0, 'S' },
{ "moov-before-mdat", no_argument, 0, OPT_MOOV_BEFORE_MDAT },
params->afterburner = n;
break;
case 'L':
- params->lowdelay_sbr = 1;
+ if (sscanf(optarg, "%d", &n) != 1 || n < -1 || n > 1) {
+ fprintf(stderr, "invalid arg for lowdelay-sbr\n");
+ return -1;
+ }
+ params->lowdelay_sbr = n;
+ break;
+ case 's':
+ if (sscanf(optarg, "%u", &n) != 1 || n > 2) {
+ fprintf(stderr, "invalid arg for sbr-ratio\n");
+ return -1;
+ }
+ params->sbr_ratio = n;
break;
case 'f':
if (sscanf(optarg, "%u", &n) != 1) {
}
params->gapless_mode = n;
break;
+ case OPT_INCLUDE_SBR_DELAY:
+ params->include_sbr_delay = 1;
+ break;
case 'I':
params->ignore_length = 1;
break;
};
static
-int write_sample(FILE *ofp, m4af_ctx_t *m4af,
- const void *data, uint32_t size, uint32_t duration)
+int write_sample(FILE *ofp, m4af_ctx_t *m4af, aacenc_frame_t *frame)
{
if (!m4af) {
- fwrite(data, 1, size, ofp);
+ fwrite(frame->data, 1, frame->size, ofp);
if (ferror(ofp)) {
fprintf(stderr, "ERROR: fwrite(): %s\n", strerror(errno));
return -1;
}
- } else if (m4af_write_sample(m4af, 0, data, size, duration) < 0) {
+ } else if (m4af_write_sample(m4af, 0, frame->data, frame->size, 0) < 0) {
fprintf(stderr, "ERROR: failed to write m4a sample\n");
return -1;
}
}
static
-int encode(pcm_reader_t *reader, HANDLE_AACENCODER encoder,
- uint32_t frame_length, FILE *ofp, m4af_ctx_t *m4af,
- int show_progress)
+int encode(aacenc_param_ex_t *params, pcm_reader_t *reader,
+ HANDLE_AACENCODER encoder, uint32_t frame_length,
+ m4af_ctx_t *m4af)
{
- int16_t *ibuf = 0;
- uint8_t *obuf = 0;
- uint32_t olen;
- uint32_t osize = 0;
+ int16_t *ibuf = 0, *ip;
+ aacenc_frame_t obuf[2] = {{ 0 }}, *obp;
+ unsigned flip = 0;
int nread = 1;
- int consumed;
int rc = -1;
- int frames_written = 0;
+ int remaining, consumed;
+ int frames_written = 0, encoded = 0;
aacenc_progress_t progress = { 0 };
const pcm_sample_description_t *fmt = pcm_get_format(reader);
ibuf = malloc(frame_length * fmt->bytes_per_frame);
aacenc_progress_init(&progress, pcm_get_length(reader), fmt->sample_rate);
- do {
+
+ for (;;) {
+ /*
+ * Since we delay the write, we cannot just exit loop when interrupted.
+ * Instead, we regard it as EOF.
+ */
if (g_interrupted)
nread = 0;
- else if (nread) {
+ if (nread > 0) {
if ((nread = pcm_read_frames(reader, ibuf, frame_length)) < 0) {
fprintf(stderr, "ERROR: read failed\n");
goto END;
}
- if (show_progress)
+ if (!params->silent)
aacenc_progress_update(&progress, pcm_get_position(reader),
fmt->sample_rate * 2);
}
- if ((consumed = aac_encode_frame(encoder, fmt, ibuf, nread,
- &obuf, &olen, &osize)) < 0)
- goto END;
- if (olen > 0) {
- if (write_sample(ofp, m4af, obuf, olen, frame_length) < 0)
+ ip = ibuf;
+ remaining = nread;
+ do {
+ obp = &obuf[flip];
+ consumed = aac_encode_frame(encoder, fmt, ip, remaining, obp);
+ if (consumed < 0) goto END;
+ if (consumed == 0 && obp->size == 0) goto DONE;
+ if (obp->size == 0) break;
+
+ remaining -= consumed;
+ ip += consumed * fmt->channels_per_frame;
+ flip ^= 1;
+ /*
+ * As we pad 1 frame at beginning and ending by our extrapolator,
+ * we want to drop them.
+ * We delay output by 1 frame by double buffering, and discard
+ * second frame and final frame from the encoder.
+ * Since sbr_header is included in the first frame (in case of
+ * SBR), we cannot discard first frame. So we pick second instead.
+ */
+ ++encoded;
+ if (encoded == 1 || encoded == 3)
+ continue;
+
+ if (write_sample(params->output_fp, m4af, &obuf[flip]) < 0)
+ goto END;
+ ++frames_written;
+ } while (remaining > 0);
+ /*
+ * When interrupted, we haven't pulled out last extrapolated frames
+ * from the reader. Therefore, we have to write the final outcome.
+ */
+ if (g_interrupted) {
+ if (write_sample(params->output_fp, m4af, &obp[flip^1]) < 0)
goto END;
++frames_written;
}
- } while (nread > 0 || olen > 0);
-
- if (show_progress)
+ }
+DONE:
+ if (!params->silent)
aacenc_progress_finish(&progress, pcm_get_position(reader));
rc = frames_written;
END:
if (ibuf) free(ibuf);
- if (obuf) free(obuf);
+ if (obuf[0].data) free(obuf[0].data);
+ if (obuf[1].data) free(obuf[1].data);
return rc;
}
{
char tool_info[256];
char *p = tool_info;
- LIB_INFO *lib_info = 0;
+ LIB_INFO lib_info;
p += sprintf(p, PROGNAME " %s, ", fdkaac_version);
-
- lib_info = calloc(FDK_MODULE_LAST, sizeof(LIB_INFO));
- if (aacEncGetLibInfo(lib_info) == AACENC_OK) {
- int i;
- for (i = 0; i < FDK_MODULE_LAST; ++i)
- if (lib_info[i].module_id == FDK_AACENC)
- break;
- p += sprintf(p, "libfdk-aac %s, ", lib_info[i].versionStr);
- }
- free(lib_info);
+ aacenc_get_lib_info(&lib_info);
+ p += sprintf(p, "libfdk-aac %s, ", lib_info.versionStr);
if (params->bitrate_mode)
sprintf(p, "VBR mode %d", params->bitrate_mode);
else
const char *ext_org = strrchr(base, '.');
if (ext_org) ilen = ext_org - base;
p = malloc(ilen + ext_len + 1);
- sprintf(p, "%.*s%s", ilen, base, ext);
+ sprintf(p, "%.*s%s", (int)ilen, base, ext);
}
return p;
}
static pcm_io_vtbl_t pcm_io_vtbl = {
read_callback, seek_callback, tell_callback
};
-static pcm_io_vtbl_t pcm_io_vtbl_noseek = { read_callback, 0, 0 };
+static pcm_io_vtbl_t pcm_io_vtbl_noseek = { read_callback, 0, tell_callback };
static
pcm_reader_t *open_input(aacenc_param_ex_t *params)
if ((params->input_fp = aacenc_fopen(params->input_filename, "rb")) == 0) {
aacenc_fprintf(stderr, "ERROR: %s: %s\n", params->input_filename,
strerror(errno));
- goto END;
+ goto FAIL;
}
io.cookie = params->input_fp;
if (fstat(fileno(params->input_fp), &stb) == 0
pcm_sample_description_t desc = { 0 };
if (parse_raw_spec(params->raw_format, &desc) < 0) {
fprintf(stderr, "ERROR: invalid raw-format spec\n");
- goto END;
+ goto FAIL;
}
desc.sample_rate = params->raw_rate;
desc.channels_per_frame = params->raw_channels;
desc.bytes_per_frame = params->raw_channels * bytes_per_channel;
if ((reader = raw_open(&io, &desc)) == 0) {
fprintf(stderr, "ERROR: failed to open raw input\n");
- goto END;
+ goto FAIL;
}
} else {
int c;
case 'R':
if ((reader = wav_open(&io, params->ignore_length)) == 0) {
fprintf(stderr, "ERROR: broken / unsupported input file\n");
- goto END;
+ goto FAIL;
}
break;
case 'c':
aacenc_translate_generic_text_tag,
¶ms->source_tag_ctx)) == 0) {
fprintf(stderr, "ERROR: broken / unsupported input file\n");
- goto END;
+ goto FAIL;
}
break;
default:
- goto END;
+ fprintf(stderr, "ERROR: unsupported input file\n");
+ goto FAIL;
}
}
- return pcm_open_sint16_converter(reader);
-END:
+ reader = pcm_open_native_converter(reader);
+ if (reader && PCM_IS_FLOAT(pcm_get_format(reader)))
+ reader = limiter_open(reader);
+ if (reader && (reader = pcm_open_sint16_converter(reader)) != 0)
+ reader = extrapolater_open(reader);
+ return reader;
+FAIL:
return 0;
}
AACENC_InfoStruct aacinfo = { 0 };
m4af_ctx_t *m4af = 0;
const pcm_sample_description_t *sample_format;
- int downsampled_timescale = 0;
int frame_count = 0;
int sbr_mode = 0;
+ unsigned scale_shift = 0;
setlocale(LC_CTYPE, "");
setbuf(stderr, 0);
sample_format = pcm_get_format(reader);
- /*
- * We use explicit/hierarchical signaling for LOAS.
- * Other than that, we request implicit signaling to FDK library, then
- * append explicit/backward-compatible signaling to ASC in case of MP4FF.
- *
- * Explicit/backward-compatible signaling of SBR is the most recommended
- * way in MPEG4 part3 spec, and seems the only way supported by iTunes.
- * Since FDK library does not support it, we have to do it on our side.
- */
- params.sbr_signaling = (params.transport_format == TT_MP4_LOAS) ? 2 : 0;
+ sbr_mode = aacenc_is_sbr_active((aacenc_param_t*)¶ms);
+ if (sbr_mode && !aacenc_is_sbr_ratio_available()) {
+ fprintf(stderr, "WARNING: Only dual-rate SBR is available "
+ "for this version\n");
+ params.sbr_ratio = 2;
+ }
+ scale_shift = aacenc_is_dual_rate_sbr((aacenc_param_t*)¶ms);
+ params.sbr_signaling =
+ (params.transport_format == TT_MP4_LOAS) ? 2 :
+ (params.transport_format == TT_MP4_RAW) ? 1 : 0;
+ if (sbr_mode && !scale_shift)
+ params.sbr_signaling = 2;
if (aacenc_init(&encoder, (aacenc_param_t*)¶ms, sample_format,
&aacinfo) < 0)
goto END;
}
handle_signals();
- sbr_mode = aacenc_is_sbr_active((aacenc_param_t*)¶ms);
+
if (!params.transport_format) {
uint32_t scale;
- uint8_t mp4asc[32];
- uint32_t ascsize = sizeof(mp4asc);
unsigned framelen = aacinfo.frameLength;
- if (sbr_mode)
- downsampled_timescale = 1;
- scale = sample_format->sample_rate >> downsampled_timescale;
+ scale = sample_format->sample_rate >> scale_shift;
if ((m4af = m4af_create(M4AF_CODEC_MP4A, scale, &m4af_io,
params.output_fp)) < 0)
goto END;
- aacenc_mp4asc((aacenc_param_t*)¶ms, aacinfo.confBuf,
- aacinfo.confSize, mp4asc, &ascsize);
- m4af_set_decoder_specific_info(m4af, 0, mp4asc, ascsize);
- m4af_set_fixed_frame_duration(m4af, 0,
- framelen >> downsampled_timescale);
+ m4af_set_num_channels(m4af, 0, sample_format->channels_per_frame);
+ m4af_set_decoder_specific_info(m4af, 0,
+ aacinfo.confBuf, aacinfo.confSize);
+ m4af_set_fixed_frame_duration(m4af, 0, framelen >> scale_shift);
m4af_set_vbr_mode(m4af, 0, params.bitrate_mode);
m4af_set_priming_mode(m4af, params.gapless_mode + 1);
m4af_begin_write(m4af);
}
- frame_count = encode(reader, encoder, aacinfo.frameLength,
- params.output_fp, m4af, !params.silent);
+ if (scale_shift && (aacinfo.encoderDelay & 1)) {
+ /*
+ * Since odd delay cannot be exactly expressed in downsampled scale,
+ * we push one zero frame to the encoder here, to make delay even
+ */
+ int16_t zero[8] = { 0 };
+ aacenc_frame_t frame = { 0 };
+ aac_encode_frame(encoder, sample_format, zero, 1, &frame);
+ free(frame.data);
+ }
+ frame_count = encode(¶ms, reader, encoder, aacinfo.frameLength, m4af);
if (frame_count < 0)
goto END;
if (m4af) {
uint32_t delay = aacinfo.encoderDelay;
+ uint32_t padding;
int64_t frames_read = pcm_get_position(reader);
- uint32_t padding = frame_count * aacinfo.frameLength
- - frames_read - aacinfo.encoderDelay;
- m4af_set_priming(m4af, 0, delay >> downsampled_timescale,
- padding >> downsampled_timescale);
+
+ if (sbr_mode && params.profile != AOT_ER_AAC_ELD &&
+ !params.include_sbr_delay)
+ delay -= 481 << scale_shift;
+ if (scale_shift && (delay & 1))
+ ++delay;
+ padding = frame_count * aacinfo.frameLength - frames_read - delay;
+ m4af_set_priming(m4af, 0, delay >> scale_shift, padding >> scale_shift);
if (finalize_m4a(m4af, ¶ms, encoder) < 0)
goto END;
}