/**
* @file
- * Ogg Vorbis codec support via libvorbisenc.
+ * Vorbis encoding support via libvorbisenc.
* @author Mark Hills <mark@pogo.org.uk>
*/
#include <vorbis/vorbisenc.h>
+#include "libavutil/fifo.h"
#include "libavutil/opt.h"
#include "avcodec.h"
+#include "audio_frame_queue.h"
#include "bytestream.h"
#include "internal.h"
#include "vorbis.h"
+#include "vorbis_parser.h"
#undef NDEBUG
#include <assert.h>
+/* Number of samples the user should send in each call.
+ * This value is used because it is the LCD of all possible frame sizes, so
+ * an output packet will always start at the same point as one of the input
+ * packets.
+ */
#define OGGVORBIS_FRAME_SIZE 64
#define BUFFER_SIZE (1024 * 64)
typedef struct OggVorbisContext {
- AVClass *av_class;
- vorbis_info vi;
- vorbis_dsp_state vd;
- vorbis_block vb;
- uint8_t buffer[BUFFER_SIZE];
- int buffer_index;
- int eof;
-
- /* decoder */
- vorbis_comment vc;
- ogg_packet op;
-
- double iblock;
+ AVClass *av_class; /**< class for AVOptions */
+ vorbis_info vi; /**< vorbis_info used during init */
+ vorbis_dsp_state vd; /**< DSP state used for analysis */
+ vorbis_block vb; /**< vorbis_block used for analysis */
+ AVFifoBuffer *pkt_fifo; /**< output packet buffer */
+ int eof; /**< end-of-file flag */
+ int dsp_initialized; /**< vd has been initialized */
+ vorbis_comment vc; /**< VorbisComment info */
+ ogg_packet op; /**< ogg packet */
+ double iblock; /**< impulse block bias option */
+ VorbisParseContext vp; /**< parse context to get durations */
+ AudioFrameQueue afq; /**< frame queue for timestamps */
} OggVorbisContext;
static const AVOption options[] = {
{ "iblock", "Sets the impulse block bias", offsetof(OggVorbisContext, iblock), AV_OPT_TYPE_DOUBLE, { .dbl = 0 }, -15, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
{ NULL }
};
+
+static const AVCodecDefault defaults[] = {
+ { "b", "0" },
+ { NULL },
+};
+
static const AVClass class = { "libvorbis", av_default_item_name, options, LIBAVUTIL_VERSION_INT };
-static const char * error(int oggerr, int *averr)
+
+static int vorbis_error_to_averror(int ov_err)
{
- switch (oggerr) {
- case OV_EFAULT: *averr = AVERROR(EFAULT); return "internal error";
- case OV_EIMPL: *averr = AVERROR(EINVAL); return "not supported";
- case OV_EINVAL: *averr = AVERROR(EINVAL); return "invalid request";
- default: *averr = AVERROR(EINVAL); return "unknown error";
+ switch (ov_err) {
+ case OV_EFAULT: return AVERROR_BUG;
+ case OV_EINVAL: return AVERROR(EINVAL);
+ case OV_EIMPL: return AVERROR(EINVAL);
+ default: return AVERROR_UNKNOWN;
}
}
-static av_cold int oggvorbis_init_encoder(vorbis_info *vi, AVCodecContext *avccontext)
+static av_cold int oggvorbis_init_encoder(vorbis_info *vi,
+ AVCodecContext *avctx)
{
- OggVorbisContext *context = avccontext->priv_data;
+ OggVorbisContext *s = avctx->priv_data;
double cfreq;
- int r;
-
- if (avccontext->flags & CODEC_FLAG_QSCALE) {
- /* variable bitrate */
- float quality = avccontext->global_quality / (float)FF_QP2LAMBDA;
- r = vorbis_encode_setup_vbr(vi, avccontext->channels,
- avccontext->sample_rate,
- quality / 10.0);
- if (r) {
- av_log(avccontext, AV_LOG_ERROR,
- "Unable to set quality to %g: %s\n", quality, error(r, &r));
- return r;
- }
+ int ret;
+
+ if (avctx->flags & CODEC_FLAG_QSCALE || !avctx->bit_rate) {
+ /* variable bitrate
+ * NOTE: we use the oggenc range of -1 to 10 for global_quality for
+ * user convenience, but libvorbis uses -0.1 to 1.0.
+ */
+ float q = avctx->global_quality / (float)FF_QP2LAMBDA;
+ /* default to 3 if the user did not set quality or bitrate */
+ if (!(avctx->flags & CODEC_FLAG_QSCALE))
+ q = 3.0;
+ if ((ret = vorbis_encode_setup_vbr(vi, avctx->channels,
+ avctx->sample_rate,
+ q / 10.0)))
+ goto error;
} else {
- int minrate = avccontext->rc_min_rate > 0 ? avccontext->rc_min_rate : -1;
- int maxrate = avccontext->rc_min_rate > 0 ? avccontext->rc_max_rate : -1;
-
- /* constant bitrate */
- r = vorbis_encode_setup_managed(vi, avccontext->channels,
- avccontext->sample_rate, minrate,
- avccontext->bit_rate, maxrate);
- if (r) {
- av_log(avccontext, AV_LOG_ERROR,
- "Unable to set CBR to %d: %s\n", avccontext->bit_rate,
- error(r, &r));
- return r;
- }
+ int minrate = avctx->rc_min_rate > 0 ? avctx->rc_min_rate : -1;
+ int maxrate = avctx->rc_max_rate > 0 ? avctx->rc_max_rate : -1;
+
+ /* average bitrate */
+ if ((ret = vorbis_encode_setup_managed(vi, avctx->channels,
+ avctx->sample_rate, maxrate,
+ avctx->bit_rate, minrate)))
+ goto error;
/* variable bitrate by estimate, disable slow rate management */
if (minrate == -1 && maxrate == -1)
- if (vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE2_SET, NULL))
- return AVERROR(EINVAL); /* should not happen */
+ if ((ret = vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE2_SET, NULL)))
+ goto error; /* should not happen */
}
/* cutoff frequency */
- if (avccontext->cutoff > 0) {
- cfreq = avccontext->cutoff / 1000.0;
- if (vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq))
- return AVERROR(EINVAL); /* should not happen */
+ if (avctx->cutoff > 0) {
+ cfreq = avctx->cutoff / 1000.0;
+ if ((ret = vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq)))
+ goto error; /* should not happen */
}
- if (context->iblock) {
- vorbis_encode_ctl(vi, OV_ECTL_IBLOCK_SET, &context->iblock);
+ /* impulse block bias */
+ if (s->iblock) {
+ if ((ret = vorbis_encode_ctl(vi, OV_ECTL_IBLOCK_SET, &s->iblock)))
+ goto error;
}
- if (avccontext->channels == 3 &&
- avccontext->channel_layout != (AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER) ||
- avccontext->channels == 4 &&
- avccontext->channel_layout != AV_CH_LAYOUT_2_2 &&
- avccontext->channel_layout != AV_CH_LAYOUT_QUAD ||
- avccontext->channels == 5 &&
- avccontext->channel_layout != AV_CH_LAYOUT_5POINT0 &&
- avccontext->channel_layout != AV_CH_LAYOUT_5POINT0_BACK ||
- avccontext->channels == 6 &&
- avccontext->channel_layout != AV_CH_LAYOUT_5POINT1 &&
- avccontext->channel_layout != AV_CH_LAYOUT_5POINT1_BACK ||
- avccontext->channels == 7 &&
- avccontext->channel_layout != (AV_CH_LAYOUT_5POINT1|AV_CH_BACK_CENTER) ||
- avccontext->channels == 8 &&
- avccontext->channel_layout != AV_CH_LAYOUT_7POINT1) {
- if (avccontext->channel_layout) {
+ if (avctx->channels == 3 &&
+ avctx->channel_layout != (AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER) ||
+ avctx->channels == 4 &&
+ avctx->channel_layout != AV_CH_LAYOUT_2_2 &&
+ avctx->channel_layout != AV_CH_LAYOUT_QUAD ||
+ avctx->channels == 5 &&
+ avctx->channel_layout != AV_CH_LAYOUT_5POINT0 &&
+ avctx->channel_layout != AV_CH_LAYOUT_5POINT0_BACK ||
+ avctx->channels == 6 &&
+ avctx->channel_layout != AV_CH_LAYOUT_5POINT1 &&
+ avctx->channel_layout != AV_CH_LAYOUT_5POINT1_BACK ||
+ avctx->channels == 7 &&
+ avctx->channel_layout != (AV_CH_LAYOUT_5POINT1|AV_CH_BACK_CENTER) ||
+ avctx->channels == 8 &&
+ avctx->channel_layout != AV_CH_LAYOUT_7POINT1) {
+ if (avctx->channel_layout) {
char name[32];
- av_get_channel_layout_string(name, sizeof(name), avccontext->channels,
- avccontext->channel_layout);
- av_log(avccontext, AV_LOG_ERROR, "%s not supported by Vorbis: "
+ av_get_channel_layout_string(name, sizeof(name), avctx->channels,
+ avctx->channel_layout);
+ av_log(avctx, AV_LOG_ERROR, "%s not supported by Vorbis: "
"output stream will have incorrect "
"channel layout.\n", name);
} else {
- av_log(avccontext, AV_LOG_WARNING, "No channel layout specified. The encoder "
+ av_log(avctx, AV_LOG_WARNING, "No channel layout specified. The encoder "
"will use Vorbis channel layout for "
- "%d channels.\n", avccontext->channels);
+ "%d channels.\n", avctx->channels);
}
}
- return vorbis_encode_setup_init(vi);
+ if ((ret = vorbis_encode_setup_init(vi)))
+ goto error;
+
+ return 0;
+error:
+ return vorbis_error_to_averror(ret);
}
/* How many bytes are needed for a buffer of length 'l' */
return 1 + l / 255 + l;
}
-static av_cold int oggvorbis_encode_init(AVCodecContext *avccontext)
+static av_cold int oggvorbis_encode_close(AVCodecContext *avctx)
{
- OggVorbisContext *context = avccontext->priv_data;
+ OggVorbisContext *s = avctx->priv_data;
+
+ /* notify vorbisenc this is EOF */
+ if (s->dsp_initialized)
+ vorbis_analysis_wrote(&s->vd, 0);
+
+ vorbis_block_clear(&s->vb);
+ vorbis_dsp_clear(&s->vd);
+ vorbis_info_clear(&s->vi);
+
+ av_fifo_free(s->pkt_fifo);
+ ff_af_queue_close(&s->afq);
+#if FF_API_OLD_ENCODE_AUDIO
+ av_freep(&avctx->coded_frame);
+#endif
+ av_freep(&avctx->extradata);
+
+ return 0;
+}
+
+static av_cold int oggvorbis_encode_init(AVCodecContext *avctx)
+{
+ OggVorbisContext *s = avctx->priv_data;
ogg_packet header, header_comm, header_code;
uint8_t *p;
unsigned int offset;
- int r;
+ int ret;
- vorbis_info_init(&context->vi);
- r = oggvorbis_init_encoder(&context->vi, avccontext);
- if (r < 0) {
- av_log(avccontext, AV_LOG_ERROR, "oggvorbis_encode_init failed\n");
- return r;
+ vorbis_info_init(&s->vi);
+ if ((ret = oggvorbis_init_encoder(&s->vi, avctx))) {
+ av_log(avctx, AV_LOG_ERROR, "encoder setup failed\n");
+ goto error;
+ }
+ if ((ret = vorbis_analysis_init(&s->vd, &s->vi))) {
+ av_log(avctx, AV_LOG_ERROR, "analysis init failed\n");
+ ret = vorbis_error_to_averror(ret);
+ goto error;
+ }
+ s->dsp_initialized = 1;
+ if ((ret = vorbis_block_init(&s->vd, &s->vb))) {
+ av_log(avctx, AV_LOG_ERROR, "dsp init failed\n");
+ ret = vorbis_error_to_averror(ret);
+ goto error;
}
- vorbis_analysis_init(&context->vd, &context->vi);
- vorbis_block_init(&context->vd, &context->vb);
- vorbis_comment_init(&context->vc);
- vorbis_comment_add_tag(&context->vc, "encoder", LIBAVCODEC_IDENT);
+ vorbis_comment_init(&s->vc);
+ vorbis_comment_add_tag(&s->vc, "encoder", LIBAVCODEC_IDENT);
- vorbis_analysis_headerout(&context->vd, &context->vc, &header,
- &header_comm, &header_code);
+ if ((ret = vorbis_analysis_headerout(&s->vd, &s->vc, &header, &header_comm,
+ &header_code))) {
+ ret = vorbis_error_to_averror(ret);
+ goto error;
+ }
- avccontext->extradata_size =
- 1 + xiph_len(header.bytes) + xiph_len(header_comm.bytes) +
- header_code.bytes;
- p = avccontext->extradata =
- av_malloc(avccontext->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
+ avctx->extradata_size = 1 + xiph_len(header.bytes) +
+ xiph_len(header_comm.bytes) +
+ header_code.bytes;
+ p = avctx->extradata = av_malloc(avctx->extradata_size +
+ FF_INPUT_BUFFER_PADDING_SIZE);
+ if (!p) {
+ ret = AVERROR(ENOMEM);
+ goto error;
+ }
p[0] = 2;
offset = 1;
offset += av_xiphlacing(&p[offset], header.bytes);
offset += header_comm.bytes;
memcpy(&p[offset], header_code.packet, header_code.bytes);
offset += header_code.bytes;
- assert(offset == avccontext->extradata_size);
+ assert(offset == avctx->extradata_size);
-#if 0
- vorbis_block_clear(&context->vb);
- vorbis_dsp_clear(&context->vd);
- vorbis_info_clear(&context->vi);
-#endif
- vorbis_comment_clear(&context->vc);
+ if ((ret = avpriv_vorbis_parse_extradata(avctx, &s->vp)) < 0) {
+ av_log(avctx, AV_LOG_ERROR, "invalid extradata\n");
+ return ret;
+ }
- avccontext->frame_size = OGGVORBIS_FRAME_SIZE;
+ vorbis_comment_clear(&s->vc);
- avccontext->coded_frame = avcodec_alloc_frame();
- avccontext->coded_frame->key_frame = 1;
+ avctx->frame_size = OGGVORBIS_FRAME_SIZE;
+ ff_af_queue_init(avctx, &s->afq);
+
+ s->pkt_fifo = av_fifo_alloc(BUFFER_SIZE);
+ if (!s->pkt_fifo) {
+ ret = AVERROR(ENOMEM);
+ goto error;
+ }
+
+#if FF_API_OLD_ENCODE_AUDIO
+ avctx->coded_frame = avcodec_alloc_frame();
+ if (!avctx->coded_frame) {
+ ret = AVERROR(ENOMEM);
+ goto error;
+ }
+#endif
return 0;
+error:
+ oggvorbis_encode_close(avctx);
+ return ret;
}
-static int oggvorbis_encode_frame(AVCodecContext *avccontext,
- unsigned char *packets,
- int buf_size, void *data)
+static int oggvorbis_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
+ const AVFrame *frame, int *got_packet_ptr)
{
- OggVorbisContext *context = avccontext->priv_data;
+ OggVorbisContext *s = avctx->priv_data;
ogg_packet op;
- signed short *audio = data;
- int l;
+ int ret, duration;
- if (data) {
- const int samples = avccontext->frame_size;
+ /* send samples to libvorbis */
+ if (frame) {
+ const float *audio = (const float *)frame->data[0];
+ const int samples = frame->nb_samples;
float **buffer;
- int c, channels = context->vi.channels;
+ int c, channels = s->vi.channels;
- buffer = vorbis_analysis_buffer(&context->vd, samples);
+ buffer = vorbis_analysis_buffer(&s->vd, samples);
for (c = 0; c < channels; c++) {
+ int i;
int co = (channels > 8) ? c :
ff_vorbis_encoding_channel_layout_offsets[channels - 1][c];
- for (l = 0; l < samples; l++)
- buffer[c][l] = audio[l * channels + co] / 32768.f;
+ for (i = 0; i < samples; i++)
+ buffer[c][i] = audio[i * channels + co];
+ }
+ if ((ret = vorbis_analysis_wrote(&s->vd, samples)) < 0) {
+ av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
+ return vorbis_error_to_averror(ret);
}
- vorbis_analysis_wrote(&context->vd, samples);
+ if ((ret = ff_af_queue_add(&s->afq, frame) < 0))
+ return ret;
} else {
- if (!context->eof)
- vorbis_analysis_wrote(&context->vd, 0);
- context->eof = 1;
+ if (!s->eof)
+ if ((ret = vorbis_analysis_wrote(&s->vd, 0)) < 0) {
+ av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
+ return vorbis_error_to_averror(ret);
+ }
+ s->eof = 1;
}
- while (vorbis_analysis_blockout(&context->vd, &context->vb) == 1) {
- vorbis_analysis(&context->vb, NULL);
- vorbis_bitrate_addblock(&context->vb);
-
- while (vorbis_bitrate_flushpacket(&context->vd, &op)) {
- /* i'd love to say the following line is a hack, but sadly it's
- * not, apparently the end of stream decision is in libogg. */
- if (op.bytes == 1 && op.e_o_s)
- continue;
- if (context->buffer_index + sizeof(ogg_packet) + op.bytes > BUFFER_SIZE) {
- av_log(avccontext, AV_LOG_ERROR, "libvorbis: buffer overflow.\n");
- return AVERROR(EINVAL);
+ /* retrieve available packets from libvorbis */
+ while ((ret = vorbis_analysis_blockout(&s->vd, &s->vb)) == 1) {
+ if ((ret = vorbis_analysis(&s->vb, NULL)) < 0)
+ break;
+ if ((ret = vorbis_bitrate_addblock(&s->vb)) < 0)
+ break;
+
+ /* add any available packets to the output packet buffer */
+ while ((ret = vorbis_bitrate_flushpacket(&s->vd, &op)) == 1) {
+ if (av_fifo_space(s->pkt_fifo) < sizeof(ogg_packet) + op.bytes) {
+ av_log(avctx, AV_LOG_ERROR, "packet buffer is too small\n");
+ return AVERROR_BUG;
}
- memcpy(context->buffer + context->buffer_index, &op, sizeof(ogg_packet));
- context->buffer_index += sizeof(ogg_packet);
- memcpy(context->buffer + context->buffer_index, op.packet, op.bytes);
- context->buffer_index += op.bytes;
-// av_log(avccontext, AV_LOG_DEBUG, "e%d / %d\n", context->buffer_index, op.bytes);
+ av_fifo_generic_write(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);
+ av_fifo_generic_write(s->pkt_fifo, op.packet, op.bytes, NULL);
}
- }
-
- l = 0;
- if (context->buffer_index) {
- ogg_packet *op2 = (ogg_packet *)context->buffer;
- op2->packet = context->buffer + sizeof(ogg_packet);
-
- l = op2->bytes;
- avccontext->coded_frame->pts = ff_samples_to_time_base(avccontext,
- op2->granulepos);
- //FIXME we should reorder the user supplied pts and not assume that they are spaced by 1/sample_rate
-
- if (l > buf_size) {
- av_log(avccontext, AV_LOG_ERROR, "libvorbis: buffer overflow.\n");
- return AVERROR(EINVAL);
+ if (ret < 0) {
+ av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
+ break;
}
-
- memcpy(packets, op2->packet, l);
- context->buffer_index -= l + sizeof(ogg_packet);
- memmove(context->buffer, context->buffer + l + sizeof(ogg_packet), context->buffer_index);
-// av_log(avccontext, AV_LOG_DEBUG, "E%d\n", l);
+ }
+ if (ret < 0) {
+ av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
+ return vorbis_error_to_averror(ret);
}
- return l;
-}
+ /* check for available packets */
+ if (av_fifo_size(s->pkt_fifo) < sizeof(ogg_packet))
+ return 0;
-static av_cold int oggvorbis_encode_close(AVCodecContext *avccontext)
-{
- OggVorbisContext *context = avccontext->priv_data;
-/* ogg_packet op ; */
+ av_fifo_generic_read(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);
- vorbis_analysis_wrote(&context->vd, 0); /* notify vorbisenc this is EOF */
+ if ((ret = ff_alloc_packet2(avctx, avpkt, op.bytes)))
+ return ret;
+ av_fifo_generic_read(s->pkt_fifo, avpkt->data, op.bytes, NULL);
- vorbis_block_clear(&context->vb);
- vorbis_dsp_clear(&context->vd);
- vorbis_info_clear(&context->vi);
+ avpkt->pts = ff_samples_to_time_base(avctx, op.granulepos);
- av_freep(&avccontext->coded_frame);
- av_freep(&avccontext->extradata);
+ duration = avpriv_vorbis_parse_frame(&s->vp, avpkt->data, avpkt->size);
+ if (duration > 0) {
+ /* we do not know encoder delay until we get the first packet from
+ * libvorbis, so we have to update the AudioFrameQueue counts */
+ if (!avctx->delay) {
+ avctx->delay = duration;
+ s->afq.remaining_delay += duration;
+ s->afq.remaining_samples += duration;
+ }
+ ff_af_queue_remove(&s->afq, duration, &avpkt->pts, &avpkt->duration);
+ }
+ *got_packet_ptr = 1;
return 0;
}
.id = CODEC_ID_VORBIS,
.priv_data_size = sizeof(OggVorbisContext),
.init = oggvorbis_encode_init,
- .encode = oggvorbis_encode_frame,
+ .encode2 = oggvorbis_encode_frame,
.close = oggvorbis_encode_close,
.capabilities = CODEC_CAP_DELAY,
- .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE },
+ .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT,
+ AV_SAMPLE_FMT_NONE },
.long_name = NULL_IF_CONFIG_SMALL("libvorbis Vorbis"),
.priv_class = &class,
+ .defaults = defaults,
};