150d2772 |
/*
* Copyright (c) 2006 Paul Richards <paul.richards@gmail.com>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software |
e5a389a1 |
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
150d2772 |
*/
|
4311ff77 |
/** |
ba87f080 |
* @file |
4311ff77 |
* @brief Theora encoder using libtheora.
* @author Paul Richards <paul.richards@gmail.com> |
150d2772 |
*
* A lot of this is copy / paste from other output codecs in
* libavcodec or pure guesswork (or both).
*
* I have used t_ prefixes on variables which are libtheora types
* and o_ prefixes on variables which are libogg types.
*/
|
2d2b5a14 |
/* FFmpeg includes */ |
1d9c2dc8 |
#include "libavutil/common.h" |
d9513de6 |
#include "libavutil/intreadwrite.h" |
83f9ed42 |
#include "libavutil/pixdesc.h" |
245976da |
#include "libavutil/log.h" |
b0a6d9ca |
#include "libavutil/base64.h" |
150d2772 |
#include "avcodec.h" |
f7fa73ac |
#include "internal.h" |
150d2772 |
/* libtheora includes */ |
bdc8c488 |
#include <theora/theoraenc.h> |
150d2772 |
|
8767fb4c |
typedef struct TheoraContext { |
bdc8c488 |
th_enc_ctx *t_state; |
b0a6d9ca |
uint8_t *stats;
int stats_size;
int stats_offset; |
af89b178 |
int uv_hshift;
int uv_vshift; |
c39b94be |
int keyframe_mask; |
150d2772 |
} TheoraContext;
|
49bd8e4b |
/** Concatenate an ogg_packet into the extradata. */ |
8767fb4c |
static int concatenate_packet(unsigned int* offset,
AVCodecContext* avc_context,
const ogg_packet* packet) |
150d2772 |
{ |
3e962f31 |
const char* message = NULL; |
150d2772 |
int newsize = avc_context->extradata_size + 2 + packet->bytes; |
9b8d11a7 |
int err = AVERROR_INVALIDDATA; |
150d2772 |
if (packet->bytes < 0) {
message = "ogg_packet has negative size";
} else if (packet->bytes > 0xffff) {
message = "ogg_packet is larger than 65535 bytes";
} else if (newsize < avc_context->extradata_size) {
message = "extradata_size would overflow";
} else { |
9b8d11a7 |
if ((err = av_reallocp(&avc_context->extradata, newsize)) < 0) {
avc_context->extradata_size = 0; |
150d2772 |
message = "av_realloc failed"; |
9b8d11a7 |
} |
150d2772 |
} |
96ff02f1 |
if (message) { |
150d2772 |
av_log(avc_context, AV_LOG_ERROR, "concatenate_packet failed: %s\n", message); |
9b8d11a7 |
return err; |
150d2772 |
}
avc_context->extradata_size = newsize; |
2c124cb6 |
AV_WB16(avc_context->extradata + (*offset), packet->bytes);
*offset += 2; |
8767fb4c |
memcpy(avc_context->extradata + (*offset), packet->packet, packet->bytes); |
150d2772 |
(*offset) += packet->bytes;
return 0;
}
|
b0a6d9ca |
static int get_stats(AVCodecContext *avctx, int eos)
{ |
800841fd |
#ifdef TH_ENCCTL_2PASS_OUT |
b0a6d9ca |
TheoraContext *h = avctx->priv_data;
uint8_t *buf;
int bytes;
bytes = th_encode_ctl(h->t_state, TH_ENCCTL_2PASS_OUT, &buf, sizeof(buf));
if (bytes < 0) {
av_log(avctx, AV_LOG_ERROR, "Error getting first pass stats\n"); |
8cb76ef2 |
return AVERROR_EXTERNAL; |
b0a6d9ca |
}
if (!eos) { |
67d4d5f5 |
void *tmp = av_fast_realloc(h->stats, &h->stats_size, |
b0a6d9ca |
h->stats_offset + bytes); |
67d4d5f5 |
if (!tmp)
return AVERROR(ENOMEM);
h->stats = tmp; |
b0a6d9ca |
memcpy(h->stats + h->stats_offset, buf, bytes);
h->stats_offset += bytes;
} else { |
784824a6 |
int b64_size = AV_BASE64_SIZE(h->stats_offset); |
b0a6d9ca |
// libtheora generates a summary header at the end
memcpy(h->stats, buf, bytes);
avctx->stats_out = av_malloc(b64_size);
av_base64_encode(avctx->stats_out, b64_size, h->stats, h->stats_offset);
}
return 0; |
800841fd |
#else
av_log(avctx, AV_LOG_ERROR, "libtheora too old to support 2pass\n"); |
8cb76ef2 |
return AVERROR(ENOSUP); |
800841fd |
#endif |
b0a6d9ca |
}
// libtheora won't read the entire buffer we give it at once, so we have to
// repeatedly submit it...
static int submit_stats(AVCodecContext *avctx)
{ |
800841fd |
#ifdef TH_ENCCTL_2PASS_IN |
b0a6d9ca |
TheoraContext *h = avctx->priv_data;
int bytes;
if (!h->stats) {
if (!avctx->stats_in) {
av_log(avctx, AV_LOG_ERROR, "No statsfile for second pass\n"); |
8cb76ef2 |
return AVERROR(EINVAL); |
b0a6d9ca |
}
h->stats_size = strlen(avctx->stats_in) * 3/4;
h->stats = av_malloc(h->stats_size); |
27216bf3 |
if (!h->stats) {
h->stats_size = 0;
return AVERROR(ENOMEM);
} |
b0a6d9ca |
h->stats_size = av_base64_decode(h->stats, avctx->stats_in, h->stats_size);
}
while (h->stats_size - h->stats_offset > 0) {
bytes = th_encode_ctl(h->t_state, TH_ENCCTL_2PASS_IN,
h->stats + h->stats_offset,
h->stats_size - h->stats_offset);
if (bytes < 0) {
av_log(avctx, AV_LOG_ERROR, "Error submitting stats\n"); |
8cb76ef2 |
return AVERROR_EXTERNAL; |
b0a6d9ca |
}
if (!bytes)
return 0;
h->stats_offset += bytes;
}
return 0; |
800841fd |
#else
av_log(avctx, AV_LOG_ERROR, "libtheora too old to support 2pass\n"); |
8cb76ef2 |
return AVERROR(ENOSUP); |
800841fd |
#endif |
b0a6d9ca |
}
|
5ef251e5 |
static av_cold int encode_init(AVCodecContext* avc_context) |
150d2772 |
{ |
bdc8c488 |
th_info t_info;
th_comment t_comment; |
150d2772 |
ogg_packet o_packet;
unsigned int offset;
TheoraContext *h = avc_context->priv_data; |
bdc8c488 |
uint32_t gop_size = avc_context->gop_size; |
8cb76ef2 |
int ret; |
150d2772 |
/* Set up the theora_info struct */ |
bdc8c488 |
th_info_init(&t_info);
t_info.frame_width = FFALIGN(avc_context->width, 16);
t_info.frame_height = FFALIGN(avc_context->height, 16);
t_info.pic_width = avc_context->width;
t_info.pic_height = avc_context->height;
t_info.pic_x = 0;
t_info.pic_y = 0; |
013591c5 |
/* Swap numerator and denominator as time_base in AVCodecContext gives the
* time period between frames, but theora_info needs the framerate. */ |
8767fb4c |
t_info.fps_numerator = avc_context->time_base.den; |
150d2772 |
t_info.fps_denominator = avc_context->time_base.num; |
96ff02f1 |
if (avc_context->sample_aspect_ratio.num) { |
8767fb4c |
t_info.aspect_numerator = avc_context->sample_aspect_ratio.num; |
150d2772 |
t_info.aspect_denominator = avc_context->sample_aspect_ratio.den;
} else { |
8767fb4c |
t_info.aspect_numerator = 1; |
150d2772 |
t_info.aspect_denominator = 1;
} |
6c2289b5 |
if (avc_context->color_primaries == AVCOL_PRI_BT470M)
t_info.colorspace = TH_CS_ITU_REC_470M;
else if (avc_context->color_primaries == AVCOL_PRI_BT470BG)
t_info.colorspace = TH_CS_ITU_REC_470BG;
else
t_info.colorspace = TH_CS_UNSPECIFIED; |
af89b178 |
|
716d413c |
if (avc_context->pix_fmt == AV_PIX_FMT_YUV420P) |
af89b178 |
t_info.pixel_fmt = TH_PF_420; |
716d413c |
else if (avc_context->pix_fmt == AV_PIX_FMT_YUV422P) |
af89b178 |
t_info.pixel_fmt = TH_PF_422; |
716d413c |
else if (avc_context->pix_fmt == AV_PIX_FMT_YUV444P) |
af89b178 |
t_info.pixel_fmt = TH_PF_444;
else {
av_log(avc_context, AV_LOG_ERROR, "Unsupported pix_fmt\n"); |
8cb76ef2 |
return AVERROR(EINVAL); |
af89b178 |
}
avcodec_get_chroma_sub_sample(avc_context->pix_fmt, &h->uv_hshift, &h->uv_vshift); |
150d2772 |
|
6945c2bc |
if (avc_context->flags & CODEC_FLAG_QSCALE) { |
37fce84e |
/* Clip global_quality in QP units to the [0 - 10] range
to be consistent with the libvorbis implementation.
Theora accepts a quality parameter which is an int value in
the [0 - 63] range.
*/ |
467c0338 |
t_info.quality = av_clipf(avc_context->global_quality / (float)FF_QP2LAMBDA, 0, 10) * 6.3; |
6945c2bc |
t_info.target_bitrate = 0;
} else {
t_info.target_bitrate = avc_context->bit_rate; |
8767fb4c |
t_info.quality = 0; |
6945c2bc |
}
|
150d2772 |
/* Now initialise libtheora */ |
bdc8c488 |
h->t_state = th_encode_alloc(&t_info);
if (!h->t_state) { |
150d2772 |
av_log(avc_context, AV_LOG_ERROR, "theora_encode_init failed\n"); |
8cb76ef2 |
return AVERROR_EXTERNAL; |
150d2772 |
}
|
c39b94be |
h->keyframe_mask = (1 << t_info.keyframe_granule_shift) - 1; |
150d2772 |
/* Clear up theora_info struct */ |
bdc8c488 |
th_info_clear(&t_info);
if (th_encode_ctl(h->t_state, TH_ENCCTL_SET_KEYFRAME_FREQUENCY_FORCE,
&gop_size, sizeof(gop_size))) {
av_log(avc_context, AV_LOG_ERROR, "Error setting GOP size\n"); |
8cb76ef2 |
return AVERROR_EXTERNAL; |
bdc8c488 |
} |
150d2772 |
|
b0a6d9ca |
// need to enable 2 pass (via TH_ENCCTL_2PASS_) before encoding headers
if (avc_context->flags & CODEC_FLAG_PASS1) { |
8cb76ef2 |
if ((ret = get_stats(avc_context, 0)) < 0)
return ret; |
b0a6d9ca |
} else if (avc_context->flags & CODEC_FLAG_PASS2) { |
8cb76ef2 |
if ((ret = submit_stats(avc_context)) < 0)
return ret; |
b0a6d9ca |
}
|
150d2772 |
/*
Output first header packet consisting of theora
header, comment, and tables.
Each one is prefixed with a 16bit size, then they |
2f5df0b1 |
are concatenated together into libavcodec's extradata. |
150d2772 |
*/
offset = 0;
|
bdc8c488 |
/* Headers */
th_comment_init(&t_comment); |
150d2772 |
|
bdc8c488 |
while (th_encode_flushheader(h->t_state, &t_comment, &o_packet)) |
8cb76ef2 |
if ((ret = concatenate_packet(&offset, avc_context, &o_packet)) < 0)
return ret; |
bdc8c488 |
th_comment_clear(&t_comment); |
150d2772 |
/* Set up the output AVFrame */ |
5b9c3b45 |
avc_context->coded_frame = av_frame_alloc(); |
150d2772 |
return 0;
}
|
f7fa73ac |
static int encode_frame(AVCodecContext* avc_context, AVPacket *pkt,
const AVFrame *frame, int *got_packet) |
150d2772 |
{ |
bdc8c488 |
th_ycbcr_buffer t_yuv_buffer; |
150d2772 |
TheoraContext *h = avc_context->priv_data;
ogg_packet o_packet; |
f7fa73ac |
int result, i, ret; |
150d2772 |
|
b0a6d9ca |
// EOS, finish and get 1st pass stats if applicable
if (!frame) {
th_encode_packetout(h->t_state, 1, &o_packet);
if (avc_context->flags & CODEC_FLAG_PASS1) |
8cb76ef2 |
if ((ret = get_stats(avc_context, 1)) < 0)
return ret; |
b0a6d9ca |
return 0;
}
|
150d2772 |
/* Copy planes to the theora yuv_buffer */ |
bdc8c488 |
for (i = 0; i < 3; i++) { |
af89b178 |
t_yuv_buffer[i].width = FFALIGN(avc_context->width, 16) >> (i && h->uv_hshift);
t_yuv_buffer[i].height = FFALIGN(avc_context->height, 16) >> (i && h->uv_vshift); |
bdc8c488 |
t_yuv_buffer[i].stride = frame->linesize[i];
t_yuv_buffer[i].data = frame->data[i]; |
150d2772 |
}
|
b0a6d9ca |
if (avc_context->flags & CODEC_FLAG_PASS2) |
8cb76ef2 |
if ((ret = submit_stats(avc_context)) < 0)
return ret; |
b0a6d9ca |
|
150d2772 |
/* Now call into theora_encode_YUVin */ |
bdc8c488 |
result = th_encode_ycbcr_in(h->t_state, t_yuv_buffer); |
96ff02f1 |
if (result) { |
150d2772 |
const char* message;
switch (result) { |
8767fb4c |
case -1:
message = "differing frame sizes";
break; |
bdc8c488 |
case TH_EINVAL: |
8767fb4c |
message = "encoder is not ready or is finished";
break;
default:
message = "unknown reason";
break; |
150d2772 |
}
av_log(avc_context, AV_LOG_ERROR, "theora_encode_YUVin failed (%s) [%d]\n", message, result); |
8cb76ef2 |
return AVERROR_EXTERNAL; |
150d2772 |
}
|
b0a6d9ca |
if (avc_context->flags & CODEC_FLAG_PASS1) |
8cb76ef2 |
if ((ret = get_stats(avc_context, 0)) < 0)
return ret; |
b0a6d9ca |
|
150d2772 |
/* Pick up returned ogg_packet */ |
bdc8c488 |
result = th_encode_packetout(h->t_state, 0, &o_packet); |
150d2772 |
switch (result) { |
8767fb4c |
case 0:
/* No packet is ready */
return 0;
case 1:
/* Success, we have a packet */
break;
default:
av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed [%d]\n", result); |
8cb76ef2 |
return AVERROR_EXTERNAL; |
150d2772 |
}
/* Copy ogg_packet content out to buffer */ |
ae2c33b0 |
if ((ret = ff_alloc_packet2(avc_context, pkt, o_packet.bytes)) < 0) |
f7fa73ac |
return ret;
memcpy(pkt->data, o_packet.packet, o_packet.bytes); |
150d2772 |
|
a4914ac7 |
// HACK: assumes no encoder delay, this is true until libtheora becomes |
511cf612 |
// multithreaded (which will be disabled unless explicitly requested) |
f7fa73ac |
pkt->pts = pkt->dts = frame->pts; |
c39b94be |
avc_context->coded_frame->key_frame = !(o_packet.granulepos & h->keyframe_mask); |
f7fa73ac |
if (avc_context->coded_frame->key_frame)
pkt->flags |= AV_PKT_FLAG_KEY;
*got_packet = 1; |
3bb98498 |
|
f7fa73ac |
return 0; |
150d2772 |
}
|
5ef251e5 |
static av_cold int encode_close(AVCodecContext* avc_context) |
150d2772 |
{
TheoraContext *h = avc_context->priv_data;
|
bdc8c488 |
th_encode_free(h->t_state); |
b0a6d9ca |
av_freep(&h->stats); |
546d69eb |
av_frame_free(&avc_context->coded_frame); |
b0a6d9ca |
av_freep(&avc_context->stats_out); |
870ee6f7 |
av_freep(&avc_context->extradata);
avc_context->extradata_size = 0;
|
b0a6d9ca |
return 0; |
150d2772 |
}
|
4311ff77 |
/** AVCodec struct exposed to libavcodec */ |
e7e2df27 |
AVCodec ff_libtheora_encoder = { |
00c3b67b |
.name = "libtheora", |
b2bed932 |
.long_name = NULL_IF_CONFIG_SMALL("libtheora Theora"), |
00c3b67b |
.type = AVMEDIA_TYPE_VIDEO, |
36ef5369 |
.id = AV_CODEC_ID_THEORA, |
150d2772 |
.priv_data_size = sizeof(TheoraContext), |
00c3b67b |
.init = encode_init,
.close = encode_close,
.encode2 = encode_frame,
.capabilities = CODEC_CAP_DELAY, // needed to get the statsfile summary |
716d413c |
.pix_fmts = (const enum AVPixelFormat[]){
AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV444P, AV_PIX_FMT_NONE |
00c3b67b |
}, |
150d2772 |
}; |