2 * Copyright (c) 2006 Paul Richards <paul.richards@gmail.com>
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 * \file libtheoraenc.c
23 * \brief Theora encoder using libtheora.
24 * \author Paul Richards <paul.richards@gmail.com>
26 * A lot of this is copy / paste from other output codecs in
27 * libavcodec or pure guesswork (or both).
29 * I have used t_ prefixes on variables which are libtheora types
30 * and o_ prefixes on variables which are libogg types.
34 #include "libavutil/intreadwrite.h"
35 #include "libavutil/log.h"
36 #include "libavutil/base64.h"
39 /* libtheora includes */
40 #include <theora/theoraenc.h>
42 typedef struct TheoraContext {
50 Concatenates an ogg_packet into the extradata.
52 static int concatenate_packet(unsigned int* offset,
53 AVCodecContext* avc_context,
54 const ogg_packet* packet)
56 const char* message = NULL;
57 uint8_t* newdata = NULL;
58 int newsize = avc_context->extradata_size + 2 + packet->bytes;
60 if (packet->bytes < 0) {
61 message = "ogg_packet has negative size";
62 } else if (packet->bytes > 0xffff) {
63 message = "ogg_packet is larger than 65535 bytes";
64 } else if (newsize < avc_context->extradata_size) {
65 message = "extradata_size would overflow";
67 newdata = av_realloc(avc_context->extradata, newsize);
69 message = "av_realloc failed";
72 av_log(avc_context, AV_LOG_ERROR, "concatenate_packet failed: %s\n", message);
76 avc_context->extradata = newdata;
77 avc_context->extradata_size = newsize;
78 AV_WB16(avc_context->extradata + (*offset), packet->bytes);
80 memcpy(avc_context->extradata + (*offset), packet->packet, packet->bytes);
81 (*offset) += packet->bytes;
85 static int get_stats(AVCodecContext *avctx, int eos)
87 TheoraContext *h = avctx->priv_data;
91 bytes = th_encode_ctl(h->t_state, TH_ENCCTL_2PASS_OUT, &buf, sizeof(buf));
93 av_log(avctx, AV_LOG_ERROR, "Error getting first pass stats\n");
97 h->stats = av_fast_realloc(h->stats, &h->stats_size,
98 h->stats_offset + bytes);
99 memcpy(h->stats + h->stats_offset, buf, bytes);
100 h->stats_offset += bytes;
102 int b64_size = ((h->stats_offset + 2) / 3) * 4 + 1;
103 // libtheora generates a summary header at the end
104 memcpy(h->stats, buf, bytes);
105 avctx->stats_out = av_malloc(b64_size);
106 av_base64_encode(avctx->stats_out, b64_size, h->stats, h->stats_offset);
111 // libtheora won't read the entire buffer we give it at once, so we have to
112 // repeatedly submit it...
113 static int submit_stats(AVCodecContext *avctx)
115 TheoraContext *h = avctx->priv_data;
118 if (!avctx->stats_in) {
119 av_log(avctx, AV_LOG_ERROR, "No statsfile for second pass\n");
122 h->stats_size = strlen(avctx->stats_in) * 3/4;
123 h->stats = av_malloc(h->stats_size);
124 h->stats_size = av_base64_decode(h->stats, avctx->stats_in, h->stats_size);
126 while (h->stats_size - h->stats_offset > 0) {
127 bytes = th_encode_ctl(h->t_state, TH_ENCCTL_2PASS_IN,
128 h->stats + h->stats_offset,
129 h->stats_size - h->stats_offset);
131 av_log(avctx, AV_LOG_ERROR, "Error submitting stats\n");
136 h->stats_offset += bytes;
141 static av_cold int encode_init(AVCodecContext* avc_context)
144 th_comment t_comment;
147 TheoraContext *h = avc_context->priv_data;
148 uint32_t gop_size = avc_context->gop_size;
150 /* Set up the theora_info struct */
151 th_info_init(&t_info);
152 t_info.frame_width = FFALIGN(avc_context->width, 16);
153 t_info.frame_height = FFALIGN(avc_context->height, 16);
154 t_info.pic_width = avc_context->width;
155 t_info.pic_height = avc_context->height;
158 /* Swap numerator and denominator as time_base in AVCodecContext gives the
159 * time period between frames, but theora_info needs the framerate. */
160 t_info.fps_numerator = avc_context->time_base.den;
161 t_info.fps_denominator = avc_context->time_base.num;
162 if (avc_context->sample_aspect_ratio.num) {
163 t_info.aspect_numerator = avc_context->sample_aspect_ratio.num;
164 t_info.aspect_denominator = avc_context->sample_aspect_ratio.den;
166 t_info.aspect_numerator = 1;
167 t_info.aspect_denominator = 1;
169 t_info.colorspace = TH_CS_UNSPECIFIED;
170 t_info.pixel_fmt = TH_PF_420;
172 if (avc_context->flags & CODEC_FLAG_QSCALE) {
173 /* to be constant with the libvorbis implementation, clip global_quality to 0 - 10
174 Theora accepts a quality parameter p, which is:
178 t_info.quality = av_clip(avc_context->global_quality / (float)FF_QP2LAMBDA, 0, 10) * 6.3;
179 t_info.target_bitrate = 0;
181 t_info.target_bitrate = avc_context->bit_rate;
185 /* Now initialise libtheora */
186 h->t_state = th_encode_alloc(&t_info);
188 av_log(avc_context, AV_LOG_ERROR, "theora_encode_init failed\n");
192 /* Clear up theora_info struct */
193 th_info_clear(&t_info);
195 if (th_encode_ctl(h->t_state, TH_ENCCTL_SET_KEYFRAME_FREQUENCY_FORCE,
196 &gop_size, sizeof(gop_size))) {
197 av_log(avc_context, AV_LOG_ERROR, "Error setting GOP size\n");
201 // need to enable 2 pass (via TH_ENCCTL_2PASS_) before encoding headers
202 if (avc_context->flags & CODEC_FLAG_PASS1) {
203 if (get_stats(avc_context, 0))
205 } else if (avc_context->flags & CODEC_FLAG_PASS2) {
206 if (submit_stats(avc_context))
211 Output first header packet consisting of theora
212 header, comment, and tables.
214 Each one is prefixed with a 16bit size, then they
215 are concatenated together into ffmpeg's extradata.
220 th_comment_init(&t_comment);
222 while (th_encode_flushheader(h->t_state, &t_comment, &o_packet))
223 if (concatenate_packet(&offset, avc_context, &o_packet))
226 th_comment_clear(&t_comment);
228 /* Set up the output AVFrame */
229 avc_context->coded_frame= avcodec_alloc_frame();
234 static int encode_frame(AVCodecContext* avc_context, uint8_t *outbuf,
235 int buf_size, void *data)
237 th_ycbcr_buffer t_yuv_buffer;
238 TheoraContext *h = avc_context->priv_data;
239 AVFrame *frame = data;
243 assert(avc_context->pix_fmt == PIX_FMT_YUV420P);
245 // EOS, finish and get 1st pass stats if applicable
247 th_encode_packetout(h->t_state, 1, &o_packet);
248 if (avc_context->flags & CODEC_FLAG_PASS1)
249 if (get_stats(avc_context, 1))
254 /* Copy planes to the theora yuv_buffer */
255 for (i = 0; i < 3; i++) {
256 t_yuv_buffer[i].width = FFALIGN(avc_context->width, 16) >> !!i;
257 t_yuv_buffer[i].height = FFALIGN(avc_context->height, 16) >> !!i;
258 t_yuv_buffer[i].stride = frame->linesize[i];
259 t_yuv_buffer[i].data = frame->data[i];
262 if (avc_context->flags & CODEC_FLAG_PASS2)
263 if (submit_stats(avc_context))
266 /* Now call into theora_encode_YUVin */
267 result = th_encode_ycbcr_in(h->t_state, t_yuv_buffer);
272 message = "differing frame sizes";
275 message = "encoder is not ready or is finished";
278 message = "unknown reason";
281 av_log(avc_context, AV_LOG_ERROR, "theora_encode_YUVin failed (%s) [%d]\n", message, result);
285 if (avc_context->flags & CODEC_FLAG_PASS1)
286 if (get_stats(avc_context, 0))
289 /* Pick up returned ogg_packet */
290 result = th_encode_packetout(h->t_state, 0, &o_packet);
293 /* No packet is ready */
296 /* Success, we have a packet */
299 av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed [%d]\n", result);
303 /* Copy ogg_packet content out to buffer */
304 if (buf_size < o_packet.bytes) {
305 av_log(avc_context, AV_LOG_ERROR, "encoded frame too large\n");
308 memcpy(outbuf, o_packet.packet, o_packet.bytes);
310 // HACK: does not take codec delay into account (neither does the decoder though)
311 avc_context->coded_frame->pts = frame->pts;
313 return o_packet.bytes;
316 static av_cold int encode_close(AVCodecContext* avc_context)
318 TheoraContext *h = avc_context->priv_data;
320 th_encode_free(h->t_state);
322 av_freep(&avc_context->coded_frame);
323 av_freep(&avc_context->stats_out);
324 av_freep(&avc_context->extradata);
325 avc_context->extradata_size = 0;
330 static const enum PixelFormat supported_pixel_formats[] = { PIX_FMT_YUV420P, PIX_FMT_NONE };
332 /*! AVCodec struct exposed to libavcodec */
333 AVCodec libtheora_encoder = {
335 .type = CODEC_TYPE_VIDEO,
336 .id = CODEC_ID_THEORA,
337 .priv_data_size = sizeof(TheoraContext),
339 .close = encode_close,
340 .encode = encode_frame,
341 .capabilities = CODEC_CAP_DELAY, // needed to get the statsfile summary
342 .pix_fmts = supported_pixel_formats,
343 .long_name = NULL_IF_CONFIG_SMALL("libtheora Theora"),