2 * H.264 encoding using the x264 library
3 * Copyright (C) 2005 Mans Rullgard <mans@mansr.com>
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 #include "libavutil/opt.h"
30 typedef struct X264Context {
47 static void X264_log(void *p, int level, const char *fmt, va_list args)
49 static const int level_map[] = {
50 [X264_LOG_ERROR] = AV_LOG_ERROR,
51 [X264_LOG_WARNING] = AV_LOG_WARNING,
52 [X264_LOG_INFO] = AV_LOG_INFO,
53 [X264_LOG_DEBUG] = AV_LOG_DEBUG
56 if (level < 0 || level > X264_LOG_DEBUG)
59 av_vlog(p, level_map[level], fmt, args);
63 static int encode_nals(AVCodecContext *ctx, uint8_t *buf, int size,
64 x264_nal_t *nals, int nnal, int skip_sei)
66 X264Context *x4 = ctx->priv_data;
70 /* Write the SEI as part of the first frame. */
71 if (x4->sei_size > 0 && nnal > 0) {
72 memcpy(p, x4->sei, x4->sei_size);
77 for (i = 0; i < nnal; i++){
78 /* Don't put the SEI in extradata. */
79 if (skip_sei && nals[i].i_type == NAL_SEI) {
80 x4->sei_size = nals[i].i_payload;
81 x4->sei = av_malloc(x4->sei_size);
82 memcpy(x4->sei, nals[i].p_payload, nals[i].i_payload);
85 memcpy(p, nals[i].p_payload, nals[i].i_payload);
86 p += nals[i].i_payload;
92 static int X264_frame(AVCodecContext *ctx, uint8_t *buf,
93 int bufsize, void *data)
95 X264Context *x4 = ctx->priv_data;
96 AVFrame *frame = data;
99 x264_picture_t pic_out;
101 x264_picture_init( &x4->pic );
102 x4->pic.img.i_csp = X264_CSP_I420;
103 x4->pic.img.i_plane = 3;
106 for (i = 0; i < 3; i++) {
107 x4->pic.img.plane[i] = frame->data[i];
108 x4->pic.img.i_stride[i] = frame->linesize[i];
111 x4->pic.i_pts = frame->pts;
113 frame->pict_type == AV_PICTURE_TYPE_I ? X264_TYPE_KEYFRAME :
114 frame->pict_type == AV_PICTURE_TYPE_P ? X264_TYPE_P :
115 frame->pict_type == AV_PICTURE_TYPE_B ? X264_TYPE_B :
117 if (x4->params.b_tff != frame->top_field_first) {
118 x4->params.b_tff = frame->top_field_first;
119 x264_encoder_reconfig(x4->enc, &x4->params);
124 if (x264_encoder_encode(x4->enc, &nal, &nnal, frame? &x4->pic: NULL, &pic_out) < 0)
127 bufsize = encode_nals(ctx, buf, bufsize, nal, nnal, 0);
130 } while (!bufsize && !frame && x264_encoder_delayed_frames(x4->enc));
132 /* FIXME: libx264 now provides DTS, but AVFrame doesn't have a field for it. */
133 x4->out_pic.pts = pic_out.i_pts;
135 switch (pic_out.i_type) {
138 x4->out_pic.pict_type = AV_PICTURE_TYPE_I;
141 x4->out_pic.pict_type = AV_PICTURE_TYPE_P;
145 x4->out_pic.pict_type = AV_PICTURE_TYPE_B;
149 x4->out_pic.key_frame = pic_out.b_keyframe;
151 x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
156 static av_cold int X264_close(AVCodecContext *avctx)
158 X264Context *x4 = avctx->priv_data;
160 av_freep(&avctx->extradata);
164 x264_encoder_close(x4->enc);
168 av_free(x4->profile);
171 av_free(x4->weightp);
177 * Detect default settings and use default profile to avoid libx264 failure.
179 static void check_default_settings(AVCodecContext *avctx)
181 X264Context *x4 = avctx->priv_data;
184 score += x4->params.analyse.i_me_range == 0;
185 score += x4->params.rc.i_qp_step == 3;
186 score += x4->params.i_keyint_max == 12;
187 score += x4->params.rc.i_qp_min == 2;
188 score += x4->params.rc.i_qp_max == 31;
189 score += x4->params.rc.f_qcompress == 0.5;
190 score += fabs(x4->params.rc.f_ip_factor - 1.25) < 0.01;
191 score += fabs(x4->params.rc.f_pb_factor - 1.25) < 0.01;
192 score += x4->params.analyse.inter == 0 && x4->params.analyse.i_subpel_refine == 8;
194 av_log(avctx, AV_LOG_ERROR, "Default settings detected, using medium profile\n");
195 x4->preset = av_strdup("medium");
196 if (avctx->bit_rate == 200*100)
201 #define OPT_STR(opt, param) \
203 if (param && x264_param_parse(&x4->params, opt, param) < 0) { \
204 av_log(avctx, AV_LOG_ERROR, \
205 "bad value for '%s': '%s'\n", opt, param); \
210 static av_cold int X264_init(AVCodecContext *avctx)
212 X264Context *x4 = avctx->priv_data;
215 x264_param_default(&x4->params);
217 x4->params.i_keyint_max = avctx->gop_size;
219 x4->params.i_bframe = avctx->max_b_frames;
220 x4->params.b_cabac = avctx->coder_type == FF_CODER_TYPE_AC;
221 x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
222 x4->params.i_bframe_bias = avctx->bframebias;
223 x4->params.i_bframe_pyramid = avctx->flags2 & CODEC_FLAG2_BPYRAMID ? X264_B_PYRAMID_NORMAL : X264_B_PYRAMID_NONE;
224 avctx->has_b_frames = avctx->flags2 & CODEC_FLAG2_BPYRAMID ? 2 : !!avctx->max_b_frames;
226 x4->params.i_keyint_min = avctx->keyint_min;
227 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
228 x4->params.i_keyint_min = x4->params.i_keyint_max;
230 x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
232 x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER;
233 x4->params.i_deblocking_filter_alphac0 = avctx->deblockalpha;
234 x4->params.i_deblocking_filter_beta = avctx->deblockbeta;
236 x4->params.rc.i_qp_min = avctx->qmin;
237 x4->params.rc.i_qp_max = avctx->qmax;
238 x4->params.rc.i_qp_step = avctx->max_qdiff;
240 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
241 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
242 x4->params.rc.f_complexity_blur = avctx->complexityblur;
244 x4->params.i_frame_reference = avctx->refs;
246 x4->params.analyse.inter = 0;
247 if (avctx->partitions) {
248 if (avctx->partitions & X264_PART_I4X4)
249 x4->params.analyse.inter |= X264_ANALYSE_I4x4;
250 if (avctx->partitions & X264_PART_I8X8)
251 x4->params.analyse.inter |= X264_ANALYSE_I8x8;
252 if (avctx->partitions & X264_PART_P8X8)
253 x4->params.analyse.inter |= X264_ANALYSE_PSUB16x16;
254 if (avctx->partitions & X264_PART_P4X4)
255 x4->params.analyse.inter |= X264_ANALYSE_PSUB8x8;
256 if (avctx->partitions & X264_PART_B8X8)
257 x4->params.analyse.inter |= X264_ANALYSE_BSUB16x16;
260 x4->params.analyse.i_direct_mv_pred = avctx->directpred;
262 x4->params.analyse.b_weighted_bipred = avctx->flags2 & CODEC_FLAG2_WPRED;
264 if (avctx->me_method == ME_EPZS)
265 x4->params.analyse.i_me_method = X264_ME_DIA;
266 else if (avctx->me_method == ME_HEX)
267 x4->params.analyse.i_me_method = X264_ME_HEX;
268 else if (avctx->me_method == ME_UMH)
269 x4->params.analyse.i_me_method = X264_ME_UMH;
270 else if (avctx->me_method == ME_FULL)
271 x4->params.analyse.i_me_method = X264_ME_ESA;
272 else if (avctx->me_method == ME_TESA)
273 x4->params.analyse.i_me_method = X264_ME_TESA;
274 else x4->params.analyse.i_me_method = X264_ME_HEX;
276 x4->params.rc.i_aq_mode = avctx->aq_mode;
277 x4->params.rc.f_aq_strength = avctx->aq_strength;
278 x4->params.rc.i_lookahead = avctx->rc_lookahead;
280 x4->params.analyse.b_psy = avctx->flags2 & CODEC_FLAG2_PSY;
281 x4->params.analyse.f_psy_rd = avctx->psy_rd;
282 x4->params.analyse.f_psy_trellis = avctx->psy_trellis;
284 x4->params.analyse.i_me_range = avctx->me_range;
285 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
287 x4->params.analyse.b_mixed_references = avctx->flags2 & CODEC_FLAG2_MIXED_REFS;
288 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
289 x4->params.analyse.b_transform_8x8 = avctx->flags2 & CODEC_FLAG2_8X8DCT;
290 x4->params.analyse.b_fast_pskip = avctx->flags2 & CODEC_FLAG2_FASTPSKIP;
292 x4->params.analyse.i_trellis = avctx->trellis;
293 x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
295 x4->params.rc.b_mb_tree = !!(avctx->flags2 & CODEC_FLAG2_MBTREE);
296 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
297 x4->params.rc.f_pb_factor = avctx->b_quant_factor;
298 x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
301 check_default_settings(avctx);
303 if (x4->preset || x4->tune) {
304 if (x264_param_default_preset(&x4->params, x4->preset, x4->tune) < 0)
308 x4->params.pf_log = X264_log;
309 x4->params.p_log_private = avctx;
310 x4->params.i_log_level = X264_LOG_DEBUG;
312 OPT_STR("weightp", x4->weightp);
314 x4->params.b_intra_refresh = avctx->flags2 & CODEC_FLAG2_INTRA_REFRESH;
315 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
316 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
317 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
318 x4->params.rc.b_stat_write = avctx->flags & CODEC_FLAG_PASS1;
319 if (avctx->flags & CODEC_FLAG_PASS2) {
320 x4->params.rc.b_stat_read = 1;
323 x4->params.rc.i_rc_method = X264_RC_CRF;
324 x4->params.rc.f_rf_constant = avctx->crf;
325 x4->params.rc.f_rf_constant_max = avctx->crf_max;
326 } else if (avctx->cqp > -1) {
327 x4->params.rc.i_rc_method = X264_RC_CQP;
328 x4->params.rc.i_qp_constant = avctx->cqp;
332 OPT_STR("stats", x4->stats);
334 // if neither crf nor cqp modes are selected we have to enable the RC
335 // we do it this way because we cannot check if the bitrate has been set
336 if (!(avctx->crf || (avctx->cqp > -1)))
337 x4->params.rc.i_rc_method = X264_RC_ABR;
339 if (avctx->rc_buffer_size && avctx->rc_initial_buffer_occupancy &&
340 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
341 x4->params.rc.f_vbv_buffer_init =
342 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
345 OPT_STR("level", x4->level);
347 if (x4->fastfirstpass)
348 x264_param_apply_fastfirstpass(&x4->params);
351 if (x264_param_apply_profile(&x4->params, x4->profile) < 0)
354 x4->params.i_width = avctx->width;
355 x4->params.i_height = avctx->height;
356 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
357 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
358 x4->params.i_fps_num = x4->params.i_timebase_den = avctx->time_base.den;
359 x4->params.i_fps_den = x4->params.i_timebase_num = avctx->time_base.num;
361 x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR;
362 x4->params.analyse.b_ssim = avctx->flags2 & CODEC_FLAG2_SSIM;
364 x4->params.b_aud = avctx->flags2 & CODEC_FLAG2_AUD;
366 x4->params.i_threads = avctx->thread_count;
368 x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT;
370 x4->params.i_slice_count = avctx->slices;
372 x4->params.vui.b_fullrange = avctx->pix_fmt == PIX_FMT_YUVJ420P;
374 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
375 x4->params.b_repeat_headers = 0;
377 // update AVCodecContext with x264 parameters
378 avctx->has_b_frames = x4->params.i_bframe ?
379 x4->params.i_bframe_pyramid ? 2 : 1 : 0;
380 avctx->bit_rate = x4->params.rc.i_bitrate*1000;
381 avctx->crf = x4->params.rc.f_rf_constant;
383 x4->enc = x264_encoder_open(&x4->params);
387 avctx->coded_frame = &x4->out_pic;
389 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
393 s = x264_encoder_headers(x4->enc, &nal, &nnal);
395 for (i = 0; i < nnal; i++)
396 if (nal[i].i_type == NAL_SEI)
397 av_log(avctx, AV_LOG_INFO, "%s\n", nal[i].p_payload+25);
399 avctx->extradata = av_malloc(s);
400 avctx->extradata_size = encode_nals(avctx, avctx->extradata, s, nal, nnal, 1);
406 #define OFFSET(x) offsetof(X264Context,x)
407 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
409 static const AVOption options[] = {
410 {"preset", "Set the encoding preset", OFFSET(preset), FF_OPT_TYPE_STRING, 0, 0, 0, VE},
411 {"tune", "Tune the encoding params", OFFSET(tune), FF_OPT_TYPE_STRING, 0, 0, 0, VE},
412 {"fastfirstpass", "Use fast settings when encoding first pass", OFFSET(fastfirstpass), FF_OPT_TYPE_INT, 1, 0, 1, VE},
413 {"profile", "Set profile restrictions", OFFSET(profile), FF_OPT_TYPE_STRING, 0, 0, 0, VE},
414 {"level", "Specify level (as defined by Annex A)", OFFSET(level), FF_OPT_TYPE_STRING, 0, 0, 0, VE},
415 {"passlogfile", "Filename for 2 pass stats", OFFSET(stats), FF_OPT_TYPE_STRING, 0, 0, 0, VE},
416 {"wpredp", "Weighted prediction for P-frames", OFFSET(weightp), FF_OPT_TYPE_STRING, 0, 0, 0, VE},
420 static const AVClass class = { "libx264", av_default_item_name, options, LIBAVUTIL_VERSION_INT };
422 AVCodec ff_libx264_encoder = {
424 .type = AVMEDIA_TYPE_VIDEO,
426 .priv_data_size = sizeof(X264Context),
428 .encode = X264_frame,
430 .capabilities = CODEC_CAP_DELAY,
431 .pix_fmts = (const enum PixelFormat[]) { PIX_FMT_YUV420P, PIX_FMT_YUVJ420P, PIX_FMT_NONE },
432 .long_name = NULL_IF_CONFIG_SMALL("libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
433 .priv_class = &class,