2 * H.264 encoding using the x264 library
3 * Copyright (C) 2005 Mans Rullgard <mans@mansr.com>
5 * This file is part of Libav.
7 * Libav 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 * Libav 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 Libav; 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"
32 typedef struct X264Context {
65 static void X264_log(void *p, int level, const char *fmt, va_list args)
67 static const int level_map[] = {
68 [X264_LOG_ERROR] = AV_LOG_ERROR,
69 [X264_LOG_WARNING] = AV_LOG_WARNING,
70 [X264_LOG_INFO] = AV_LOG_INFO,
71 [X264_LOG_DEBUG] = AV_LOG_DEBUG
74 if (level < 0 || level > X264_LOG_DEBUG)
77 av_vlog(p, level_map[level], fmt, args);
81 static int encode_nals(AVCodecContext *ctx, uint8_t *buf, int size,
82 x264_nal_t *nals, int nnal, int skip_sei)
84 X264Context *x4 = ctx->priv_data;
88 /* Write the SEI as part of the first frame. */
89 if (x4->sei_size > 0 && nnal > 0) {
90 memcpy(p, x4->sei, x4->sei_size);
95 for (i = 0; i < nnal; i++){
96 /* Don't put the SEI in extradata. */
97 if (skip_sei && nals[i].i_type == NAL_SEI) {
98 x4->sei_size = nals[i].i_payload;
99 x4->sei = av_malloc(x4->sei_size);
100 memcpy(x4->sei, nals[i].p_payload, nals[i].i_payload);
103 memcpy(p, nals[i].p_payload, nals[i].i_payload);
104 p += nals[i].i_payload;
110 static int X264_frame(AVCodecContext *ctx, uint8_t *buf,
111 int bufsize, void *data)
113 X264Context *x4 = ctx->priv_data;
114 AVFrame *frame = data;
117 x264_picture_t pic_out;
119 x264_picture_init( &x4->pic );
120 x4->pic.img.i_csp = X264_CSP_I420;
121 x4->pic.img.i_plane = 3;
124 for (i = 0; i < 3; i++) {
125 x4->pic.img.plane[i] = frame->data[i];
126 x4->pic.img.i_stride[i] = frame->linesize[i];
129 x4->pic.i_pts = frame->pts;
131 frame->pict_type == AV_PICTURE_TYPE_I ? X264_TYPE_KEYFRAME :
132 frame->pict_type == AV_PICTURE_TYPE_P ? X264_TYPE_P :
133 frame->pict_type == AV_PICTURE_TYPE_B ? X264_TYPE_B :
135 if (x4->params.b_tff != frame->top_field_first) {
136 x4->params.b_tff = frame->top_field_first;
137 x264_encoder_reconfig(x4->enc, &x4->params);
142 if (x264_encoder_encode(x4->enc, &nal, &nnal, frame? &x4->pic: NULL, &pic_out) < 0)
145 bufsize = encode_nals(ctx, buf, bufsize, nal, nnal, 0);
148 } while (!bufsize && !frame && x264_encoder_delayed_frames(x4->enc));
150 /* FIXME: libx264 now provides DTS, but AVFrame doesn't have a field for it. */
151 x4->out_pic.pts = pic_out.i_pts;
153 switch (pic_out.i_type) {
156 x4->out_pic.pict_type = AV_PICTURE_TYPE_I;
159 x4->out_pic.pict_type = AV_PICTURE_TYPE_P;
163 x4->out_pic.pict_type = AV_PICTURE_TYPE_B;
167 x4->out_pic.key_frame = pic_out.b_keyframe;
169 x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
174 static av_cold int X264_close(AVCodecContext *avctx)
176 X264Context *x4 = avctx->priv_data;
178 av_freep(&avctx->extradata);
182 x264_encoder_close(x4->enc);
187 static av_cold int X264_init(AVCodecContext *avctx)
189 X264Context *x4 = avctx->priv_data;
191 x264_param_default(&x4->params);
193 x4->params.b_cabac = avctx->coder_type == FF_CODER_TYPE_AC;
194 x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
196 x4->params.i_keyint_min = avctx->keyint_min;
197 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
198 x4->params.i_keyint_min = x4->params.i_keyint_max;
200 x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER;
201 x4->params.i_deblocking_filter_alphac0 = avctx->deblockalpha;
202 x4->params.i_deblocking_filter_beta = avctx->deblockbeta;
204 x4->params.rc.f_complexity_blur = avctx->complexityblur;
206 x4->params.analyse.inter = 0;
207 if (avctx->partitions) {
208 if (avctx->partitions & X264_PART_I4X4)
209 x4->params.analyse.inter |= X264_ANALYSE_I4x4;
210 if (avctx->partitions & X264_PART_I8X8)
211 x4->params.analyse.inter |= X264_ANALYSE_I8x8;
212 if (avctx->partitions & X264_PART_P8X8)
213 x4->params.analyse.inter |= X264_ANALYSE_PSUB16x16;
214 if (avctx->partitions & X264_PART_P4X4)
215 x4->params.analyse.inter |= X264_ANALYSE_PSUB8x8;
216 if (avctx->partitions & X264_PART_B8X8)
217 x4->params.analyse.inter |= X264_ANALYSE_BSUB16x16;
220 x4->params.analyse.i_direct_mv_pred = avctx->directpred;
222 if (avctx->me_method == ME_EPZS)
223 x4->params.analyse.i_me_method = X264_ME_DIA;
224 else if (avctx->me_method == ME_HEX)
225 x4->params.analyse.i_me_method = X264_ME_HEX;
226 else if (avctx->me_method == ME_UMH)
227 x4->params.analyse.i_me_method = X264_ME_UMH;
228 else if (avctx->me_method == ME_FULL)
229 x4->params.analyse.i_me_method = X264_ME_ESA;
230 else if (avctx->me_method == ME_TESA)
231 x4->params.analyse.i_me_method = X264_ME_TESA;
232 else x4->params.analyse.i_me_method = X264_ME_HEX;
234 x4->params.analyse.i_me_range = avctx->me_range;
235 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
237 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
239 x4->params.analyse.i_trellis = avctx->trellis;
240 x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
242 if (avctx->level > 0)
243 x4->params.i_level_idc = avctx->level;
245 if (x4->preset || x4->tune)
246 if (x264_param_default_preset(&x4->params, x4->preset, x4->tune) < 0) {
247 av_log(avctx, AV_LOG_ERROR, "Error setting preset/tune %s/%s.\n", x4->preset, x4->tune);
248 return AVERROR(EINVAL);
251 x4->params.pf_log = X264_log;
252 x4->params.p_log_private = avctx;
253 x4->params.i_log_level = X264_LOG_DEBUG;
255 if (avctx->bit_rate) {
256 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
257 x4->params.rc.i_rc_method = X264_RC_ABR;
259 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
260 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
261 x4->params.rc.b_stat_write = avctx->flags & CODEC_FLAG_PASS1;
262 if (avctx->flags & CODEC_FLAG_PASS2) {
263 x4->params.rc.b_stat_read = 1;
265 #if FF_API_X264_GLOBAL_OPTS
267 x4->params.rc.i_rc_method = X264_RC_CRF;
268 x4->params.rc.f_rf_constant = avctx->crf;
269 x4->params.rc.f_rf_constant_max = avctx->crf_max;
270 } else if (avctx->cqp > -1) {
271 x4->params.rc.i_rc_method = X264_RC_CQP;
272 x4->params.rc.i_qp_constant = avctx->cqp;
277 x4->params.rc.i_rc_method = X264_RC_CRF;
278 x4->params.rc.f_rf_constant = x4->crf;
279 } else if (x4->cqp >= 0) {
280 x4->params.rc.i_rc_method = X264_RC_CQP;
281 x4->params.rc.i_qp_constant = x4->cqp;
284 if (x4->crf_max >= 0)
285 x4->params.rc.f_rf_constant_max = x4->crf_max;
288 if (avctx->rc_buffer_size && avctx->rc_initial_buffer_occupancy &&
289 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
290 x4->params.rc.f_vbv_buffer_init =
291 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
294 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
295 x4->params.rc.f_pb_factor = avctx->b_quant_factor;
296 x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
298 #if FF_API_X264_GLOBAL_OPTS
299 if (avctx->aq_mode >= 0)
300 x4->params.rc.i_aq_mode = avctx->aq_mode;
301 if (avctx->aq_strength >= 0)
302 x4->params.rc.f_aq_strength = avctx->aq_strength;
303 if (avctx->psy_rd >= 0)
304 x4->params.analyse.f_psy_rd = avctx->psy_rd;
305 if (avctx->psy_trellis >= 0)
306 x4->params.analyse.f_psy_trellis = avctx->psy_trellis;
307 if (avctx->rc_lookahead >= 0)
308 x4->params.rc.i_lookahead = avctx->rc_lookahead;
309 if (avctx->weighted_p_pred >= 0)
310 x4->params.analyse.i_weighted_pred = avctx->weighted_p_pred;
311 if (avctx->bframebias)
312 x4->params.i_bframe_bias = avctx->bframebias;
313 x4->params.analyse.b_ssim = avctx->flags2 & CODEC_FLAG2_SSIM;
314 x4->params.b_intra_refresh = avctx->flags2 & CODEC_FLAG2_INTRA_REFRESH;
315 x4->params.i_bframe_pyramid = avctx->flags2 & CODEC_FLAG2_BPYRAMID ? X264_B_PYRAMID_NORMAL : X264_B_PYRAMID_NONE;
316 x4->params.analyse.b_weighted_bipred = avctx->flags2 & CODEC_FLAG2_WPRED;
317 x4->params.analyse.b_mixed_references = avctx->flags2 & CODEC_FLAG2_MIXED_REFS;
318 x4->params.analyse.b_transform_8x8 = avctx->flags2 & CODEC_FLAG2_8X8DCT;
319 x4->params.analyse.b_fast_pskip = avctx->flags2 & CODEC_FLAG2_FASTPSKIP;
320 x4->params.b_aud = avctx->flags2 & CODEC_FLAG2_AUD;
321 x4->params.analyse.b_psy = avctx->flags2 & CODEC_FLAG2_PSY;
322 x4->params.rc.b_mb_tree = !!(avctx->flags2 & CODEC_FLAG2_MBTREE);
325 if (avctx->gop_size >= 0)
326 x4->params.i_keyint_max = avctx->gop_size;
327 if (avctx->max_b_frames >= 0)
328 x4->params.i_bframe = avctx->max_b_frames;
329 if (avctx->scenechange_threshold >= 0)
330 x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
331 if (avctx->qmin >= 0)
332 x4->params.rc.i_qp_min = avctx->qmin;
333 if (avctx->qmax >= 0)
334 x4->params.rc.i_qp_max = avctx->qmax;
335 if (avctx->max_qdiff >= 0)
336 x4->params.rc.i_qp_step = avctx->max_qdiff;
337 if (avctx->qblur >= 0)
338 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
339 if (avctx->qcompress >= 0)
340 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
341 if (avctx->refs >= 0)
342 x4->params.i_frame_reference = avctx->refs;
344 if (x4->aq_mode >= 0)
345 x4->params.rc.i_aq_mode = x4->aq_mode;
346 if (x4->aq_strength >= 0)
347 x4->params.rc.f_aq_strength = x4->aq_strength;
348 if (x4->psy_rd && x264_param_parse(&x4->params, "psy-rd", x4->psy_rd) < 0) {
349 av_log(avctx, AV_LOG_ERROR, "Error parsing option 'psy-rd' with value '%s'.\n", x4->psy_rd);
350 return AVERROR(EINVAL);
353 x4->params.analyse.b_psy = x4->psy;
354 if (x4->rc_lookahead >= 0)
355 x4->params.rc.i_lookahead = x4->rc_lookahead;
356 if (x4->weightp >= 0)
357 x4->params.analyse.i_weighted_pred = x4->weightp;
358 if (x4->weightb >= 0)
359 x4->params.analyse.b_weighted_bipred = x4->weightb;
362 x4->params.analyse.b_ssim = x4->ssim;
363 if (x4->intra_refresh >= 0)
364 x4->params.b_intra_refresh = x4->intra_refresh;
365 if (x4->b_bias != INT_MIN)
366 x4->params.i_bframe_bias = x4->b_bias;
367 if (x4->b_pyramid >= 0)
368 x4->params.i_bframe_pyramid = x4->b_pyramid;
369 if (x4->mixed_refs >= 0)
370 x4->params.analyse.b_mixed_references = x4->mixed_refs;
372 x4->params.analyse.b_transform_8x8 = x4->dct8x8;
373 if (x4->fast_pskip >= 0)
374 x4->params.analyse.b_fast_pskip = x4->fast_pskip;
376 x4->params.b_aud = x4->aud;
378 x4->params.rc.b_mb_tree = x4->mbtree;
380 if (x4->fastfirstpass)
381 x264_param_apply_fastfirstpass(&x4->params);
384 if (x264_param_apply_profile(&x4->params, x4->profile) < 0) {
385 av_log(avctx, AV_LOG_ERROR, "Error setting profile %s.\n", x4->profile);
386 return AVERROR(EINVAL);
389 x4->params.i_width = avctx->width;
390 x4->params.i_height = avctx->height;
391 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
392 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
393 x4->params.i_fps_num = x4->params.i_timebase_den = avctx->time_base.den;
394 x4->params.i_fps_den = x4->params.i_timebase_num = avctx->time_base.num;
396 x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR;
398 x4->params.i_threads = avctx->thread_count;
400 x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT;
402 x4->params.b_open_gop = !(avctx->flags & CODEC_FLAG_CLOSED_GOP);
404 x4->params.i_slice_count = avctx->slices;
406 x4->params.vui.b_fullrange = avctx->pix_fmt == PIX_FMT_YUVJ420P;
408 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
409 x4->params.b_repeat_headers = 0;
411 // update AVCodecContext with x264 parameters
412 avctx->has_b_frames = x4->params.i_bframe ?
413 x4->params.i_bframe_pyramid ? 2 : 1 : 0;
414 avctx->bit_rate = x4->params.rc.i_bitrate*1000;
415 #if FF_API_X264_GLOBAL_OPTS
416 avctx->crf = x4->params.rc.f_rf_constant;
419 x4->enc = x264_encoder_open(&x4->params);
423 avctx->coded_frame = &x4->out_pic;
425 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
429 s = x264_encoder_headers(x4->enc, &nal, &nnal);
431 for (i = 0; i < nnal; i++)
432 if (nal[i].i_type == NAL_SEI)
433 av_log(avctx, AV_LOG_INFO, "%s\n", nal[i].p_payload+25);
435 avctx->extradata = av_malloc(s);
436 avctx->extradata_size = encode_nals(avctx, avctx->extradata, s, nal, nnal, 1);
442 #define OFFSET(x) offsetof(X264Context, x)
443 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
444 static const AVOption options[] = {
445 { "preset", "Set the encoding preset (cf. x264 --fullhelp)", OFFSET(preset), FF_OPT_TYPE_STRING, { .str = "medium" }, 0, 0, VE},
446 { "tune", "Tune the encoding params (cf. x264 --fullhelp)", OFFSET(tune), FF_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
447 { "profile", "Set profile restrictions (cf. x264 --fullhelp) ", OFFSET(profile), FF_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
448 { "fastfirstpass", "Use fast settings when encoding first pass", OFFSET(fastfirstpass), FF_OPT_TYPE_INT, { 1 }, 0, 1, VE},
449 { "crf", "Select the quality for constant quality mode", OFFSET(crf), FF_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
450 { "crf_max", "In CRF mode, prevents VBV from lowering quality beyond this point.",OFFSET(crf_max), FF_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
451 { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
452 { "aq-mode", "AQ method", OFFSET(aq_mode), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "aq_mode"},
453 { "none", NULL, 0, FF_OPT_TYPE_CONST, {X264_AQ_NONE}, INT_MIN, INT_MAX, VE, "aq_mode" },
454 { "variance", "Variance AQ (complexity mask)", 0, FF_OPT_TYPE_CONST, {X264_AQ_VARIANCE}, INT_MIN, INT_MAX, VE, "aq_mode" },
455 { "autovariance", "Auto-variance AQ (experimental)", 0, FF_OPT_TYPE_CONST, {X264_AQ_AUTOVARIANCE}, INT_MIN, INT_MAX, VE, "aq_mode" },
456 { "aq-strength", "AQ strength. Reduces blocking and blurring in flat and textured areas.", OFFSET(aq_strength), FF_OPT_TYPE_FLOAT, {-1}, -1, FLT_MAX, VE},
457 { "psy", "Use psychovisual optimizations.", OFFSET(psy), FF_OPT_TYPE_INT, {-1 }, -1, 1, VE },
458 { "psy-rd", "Strength of psychovisual optimization, in <psy-rd>:<psy-trellis> format.", OFFSET(psy_rd), FF_OPT_TYPE_STRING, {0 }, 0, 0, VE},
459 { "rc-lookahead", "Number of frames to look ahead for frametype and ratecontrol", OFFSET(rc_lookahead), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
460 { "weightb", "Weighted prediction for B-frames.", OFFSET(weightb), FF_OPT_TYPE_INT, {-1 }, -1, 1, VE },
461 { "weightp", "Weighted prediction analysis method.", OFFSET(weightp), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "weightp" },
462 { "none", NULL, 0, FF_OPT_TYPE_CONST, {X264_WEIGHTP_NONE}, INT_MIN, INT_MAX, VE, "weightp" },
463 { "simple", NULL, 0, FF_OPT_TYPE_CONST, {X264_WEIGHTP_SIMPLE}, INT_MIN, INT_MAX, VE, "weightp" },
464 { "smart", NULL, 0, FF_OPT_TYPE_CONST, {X264_WEIGHTP_SMART}, INT_MIN, INT_MAX, VE, "weightp" },
465 { "ssim", "Calculate and print SSIM stats.", OFFSET(ssim), FF_OPT_TYPE_INT, {-1 }, -1, 1, VE },
466 { "intra-refresh", "Use Periodic Intra Refresh instead of IDR frames.",OFFSET(intra_refresh),FF_OPT_TYPE_INT, {-1 }, -1, 1, VE },
467 { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), FF_OPT_TYPE_INT, {INT_MIN}, INT_MIN, INT_MAX, VE },
468 { "b-pyramid", "Keep some B-frames as references.", OFFSET(b_pyramid), FF_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "b_pyramid" },
469 { "none", NULL, 0, FF_OPT_TYPE_CONST, {X264_B_PYRAMID_NONE}, INT_MIN, INT_MAX, VE, "b_pyramid" },
470 { "strict", "Strictly hierarchical pyramid", 0, FF_OPT_TYPE_CONST, {X264_B_PYRAMID_STRICT}, INT_MIN, INT_MAX, VE, "b_pyramid" },
471 { "normal", "Non-strict (not Blu-ray compatible)", 0, FF_OPT_TYPE_CONST, {X264_B_PYRAMID_NORMAL}, INT_MIN, INT_MAX, VE, "b_pyramid" },
472 { "mixed-refs", "One reference per partition, as opposed to one reference per macroblock", OFFSET(mixed_refs), FF_OPT_TYPE_INT, {-1}, -1, 1, VE },
473 { "8x8dct", "High profile 8x8 transform.", OFFSET(dct8x8), FF_OPT_TYPE_INT, {-1 }, -1, 1, VE},
474 { "fast-pskip", NULL, OFFSET(fast_pskip), FF_OPT_TYPE_INT, {-1 }, -1, 1, VE},
475 { "aud", "Use access unit delimiters.", OFFSET(aud), FF_OPT_TYPE_INT, {-1 }, -1, 1, VE},
476 { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), FF_OPT_TYPE_INT, {-1 }, -1, 1, VE},
480 static const AVClass class = {
481 .class_name = "libx264",
482 .item_name = av_default_item_name,
484 .version = LIBAVUTIL_VERSION_INT,
487 static const AVCodecDefault x264_defaults[] = {
497 { "sc_threshold", "-1" },
498 { "threads", AV_STRINGIFY(X264_THREADS_AUTO) },
502 AVCodec ff_libx264_encoder = {
504 .type = AVMEDIA_TYPE_VIDEO,
506 .priv_data_size = sizeof(X264Context),
508 .encode = X264_frame,
510 .capabilities = CODEC_CAP_DELAY,
511 .pix_fmts = (const enum PixelFormat[]) { PIX_FMT_YUV420P, PIX_FMT_YUVJ420P, PIX_FMT_NONE },
512 .long_name = NULL_IF_CONFIG_SMALL("libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
513 .priv_class = &class,
514 .defaults = x264_defaults,