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"
23 #include "libavutil/pixdesc.h"
33 typedef struct X264Context {
71 static void X264_log(void *p, int level, const char *fmt, va_list args)
73 static const int level_map[] = {
74 [X264_LOG_ERROR] = AV_LOG_ERROR,
75 [X264_LOG_WARNING] = AV_LOG_WARNING,
76 [X264_LOG_INFO] = AV_LOG_INFO,
77 [X264_LOG_DEBUG] = AV_LOG_DEBUG
80 if (level < 0 || level > X264_LOG_DEBUG)
83 av_vlog(p, level_map[level], fmt, args);
87 static int encode_nals(AVCodecContext *ctx, uint8_t *buf, int size,
88 x264_nal_t *nals, int nnal, int skip_sei)
90 X264Context *x4 = ctx->priv_data;
94 /* Write the SEI as part of the first frame. */
95 if (x4->sei_size > 0 && nnal > 0) {
96 memcpy(p, x4->sei, x4->sei_size);
101 for (i = 0; i < nnal; i++){
102 /* Don't put the SEI in extradata. */
103 if (skip_sei && nals[i].i_type == NAL_SEI) {
104 x4->sei_size = nals[i].i_payload;
105 x4->sei = av_malloc(x4->sei_size);
106 memcpy(x4->sei, nals[i].p_payload, nals[i].i_payload);
109 memcpy(p, nals[i].p_payload, nals[i].i_payload);
110 p += nals[i].i_payload;
116 static int X264_frame(AVCodecContext *ctx, uint8_t *buf,
117 int bufsize, void *data)
119 X264Context *x4 = ctx->priv_data;
120 AVFrame *frame = data;
123 x264_picture_t pic_out;
125 x264_picture_init( &x4->pic );
126 x4->pic.img.i_csp = x4->params.i_csp;
127 if (x264_bit_depth > 8)
128 x4->pic.img.i_csp |= X264_CSP_HIGH_DEPTH;
129 x4->pic.img.i_plane = 3;
132 for (i = 0; i < 3; i++) {
133 x4->pic.img.plane[i] = frame->data[i];
134 x4->pic.img.i_stride[i] = frame->linesize[i];
137 x4->pic.i_pts = frame->pts;
139 frame->pict_type == AV_PICTURE_TYPE_I ? X264_TYPE_KEYFRAME :
140 frame->pict_type == AV_PICTURE_TYPE_P ? X264_TYPE_P :
141 frame->pict_type == AV_PICTURE_TYPE_B ? X264_TYPE_B :
143 if (x4->params.b_tff != frame->top_field_first) {
144 x4->params.b_tff = frame->top_field_first;
145 x264_encoder_reconfig(x4->enc, &x4->params);
150 if (x264_encoder_encode(x4->enc, &nal, &nnal, frame? &x4->pic: NULL, &pic_out) < 0)
153 bufsize = encode_nals(ctx, buf, bufsize, nal, nnal, 0);
156 } while (!bufsize && !frame && x264_encoder_delayed_frames(x4->enc));
158 /* FIXME: libx264 now provides DTS, but AVFrame doesn't have a field for it. */
159 x4->out_pic.pts = pic_out.i_pts;
161 switch (pic_out.i_type) {
164 x4->out_pic.pict_type = AV_PICTURE_TYPE_I;
167 x4->out_pic.pict_type = AV_PICTURE_TYPE_P;
171 x4->out_pic.pict_type = AV_PICTURE_TYPE_B;
175 x4->out_pic.key_frame = pic_out.b_keyframe;
177 x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
182 static av_cold int X264_close(AVCodecContext *avctx)
184 X264Context *x4 = avctx->priv_data;
186 av_freep(&avctx->extradata);
190 x264_encoder_close(x4->enc);
195 static int convert_pix_fmt(enum PixelFormat pix_fmt)
198 case PIX_FMT_YUV420P:
199 case PIX_FMT_YUVJ420P:
200 case PIX_FMT_YUV420P9:
201 case PIX_FMT_YUV420P10: return X264_CSP_I420;
202 case PIX_FMT_YUV422P:
203 case PIX_FMT_YUV422P10: return X264_CSP_I422;
204 case PIX_FMT_YUV444P:
205 case PIX_FMT_YUV444P9:
206 case PIX_FMT_YUV444P10: return X264_CSP_I444;
211 #define PARSE_X264_OPT(name, var)\
212 if (x4->var && x264_param_parse(&x4->params, name, x4->var) < 0) {\
213 av_log(avctx, AV_LOG_ERROR, "Error parsing option '%s' with value '%s'.\n", name, x4->var);\
214 return AVERROR(EINVAL);\
217 static av_cold int X264_init(AVCodecContext *avctx)
219 X264Context *x4 = avctx->priv_data;
221 x264_param_default(&x4->params);
223 x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER;
225 if (x4->preset || x4->tune)
226 if (x264_param_default_preset(&x4->params, x4->preset, x4->tune) < 0) {
227 av_log(avctx, AV_LOG_ERROR, "Error setting preset/tune %s/%s.\n", x4->preset, x4->tune);
228 return AVERROR(EINVAL);
231 if (avctx->level > 0)
232 x4->params.i_level_idc = avctx->level;
234 x4->params.pf_log = X264_log;
235 x4->params.p_log_private = avctx;
236 x4->params.i_log_level = X264_LOG_DEBUG;
237 x4->params.i_csp = convert_pix_fmt(avctx->pix_fmt);
239 if (avctx->bit_rate) {
240 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
241 x4->params.rc.i_rc_method = X264_RC_ABR;
243 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
244 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
245 x4->params.rc.b_stat_write = avctx->flags & CODEC_FLAG_PASS1;
246 if (avctx->flags & CODEC_FLAG_PASS2) {
247 x4->params.rc.b_stat_read = 1;
250 x4->params.rc.i_rc_method = X264_RC_CRF;
251 x4->params.rc.f_rf_constant = x4->crf;
252 } else if (x4->cqp >= 0) {
253 x4->params.rc.i_rc_method = X264_RC_CQP;
254 x4->params.rc.i_qp_constant = x4->cqp;
257 if (x4->crf_max >= 0)
258 x4->params.rc.f_rf_constant_max = x4->crf_max;
261 if (avctx->rc_buffer_size && avctx->rc_initial_buffer_occupancy &&
262 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
263 x4->params.rc.f_vbv_buffer_init =
264 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
267 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
268 x4->params.rc.f_pb_factor = avctx->b_quant_factor;
269 x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
271 if (avctx->me_method == ME_EPZS)
272 x4->params.analyse.i_me_method = X264_ME_DIA;
273 else if (avctx->me_method == ME_HEX)
274 x4->params.analyse.i_me_method = X264_ME_HEX;
275 else if (avctx->me_method == ME_UMH)
276 x4->params.analyse.i_me_method = X264_ME_UMH;
277 else if (avctx->me_method == ME_FULL)
278 x4->params.analyse.i_me_method = X264_ME_ESA;
279 else if (avctx->me_method == ME_TESA)
280 x4->params.analyse.i_me_method = X264_ME_TESA;
282 if (avctx->gop_size >= 0)
283 x4->params.i_keyint_max = avctx->gop_size;
284 if (avctx->max_b_frames >= 0)
285 x4->params.i_bframe = avctx->max_b_frames;
286 if (avctx->scenechange_threshold >= 0)
287 x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
288 if (avctx->qmin >= 0)
289 x4->params.rc.i_qp_min = avctx->qmin;
290 if (avctx->qmax >= 0)
291 x4->params.rc.i_qp_max = avctx->qmax;
292 if (avctx->max_qdiff >= 0)
293 x4->params.rc.i_qp_step = avctx->max_qdiff;
294 if (avctx->qblur >= 0)
295 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
296 if (avctx->qcompress >= 0)
297 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
298 if (avctx->refs >= 0)
299 x4->params.i_frame_reference = avctx->refs;
300 if (avctx->trellis >= 0)
301 x4->params.analyse.i_trellis = avctx->trellis;
302 if (avctx->me_range >= 0)
303 x4->params.analyse.i_me_range = avctx->me_range;
304 if (avctx->noise_reduction >= 0)
305 x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
306 if (avctx->me_subpel_quality >= 0)
307 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
308 if (avctx->b_frame_strategy >= 0)
309 x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
310 if (avctx->keyint_min >= 0)
311 x4->params.i_keyint_min = avctx->keyint_min;
312 if (avctx->coder_type >= 0)
313 x4->params.b_cabac = avctx->coder_type == FF_CODER_TYPE_AC;
314 if (avctx->me_cmp >= 0)
315 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
317 if (x4->aq_mode >= 0)
318 x4->params.rc.i_aq_mode = x4->aq_mode;
319 if (x4->aq_strength >= 0)
320 x4->params.rc.f_aq_strength = x4->aq_strength;
321 PARSE_X264_OPT("psy-rd", psy_rd);
322 PARSE_X264_OPT("deblock", deblock);
323 PARSE_X264_OPT("partitions", partitions);
325 x4->params.analyse.b_psy = x4->psy;
326 if (x4->rc_lookahead >= 0)
327 x4->params.rc.i_lookahead = x4->rc_lookahead;
328 if (x4->weightp >= 0)
329 x4->params.analyse.i_weighted_pred = x4->weightp;
330 if (x4->weightb >= 0)
331 x4->params.analyse.b_weighted_bipred = x4->weightb;
332 if (x4->cplxblur >= 0)
333 x4->params.rc.f_complexity_blur = x4->cplxblur;
336 x4->params.analyse.b_ssim = x4->ssim;
337 if (x4->intra_refresh >= 0)
338 x4->params.b_intra_refresh = x4->intra_refresh;
339 if (x4->b_bias != INT_MIN)
340 x4->params.i_bframe_bias = x4->b_bias;
341 if (x4->b_pyramid >= 0)
342 x4->params.i_bframe_pyramid = x4->b_pyramid;
343 if (x4->mixed_refs >= 0)
344 x4->params.analyse.b_mixed_references = x4->mixed_refs;
346 x4->params.analyse.b_transform_8x8 = x4->dct8x8;
347 if (x4->fast_pskip >= 0)
348 x4->params.analyse.b_fast_pskip = x4->fast_pskip;
350 x4->params.b_aud = x4->aud;
352 x4->params.rc.b_mb_tree = x4->mbtree;
353 if (x4->direct_pred >= 0)
354 x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
356 if (x4->slice_max_size >= 0)
357 x4->params.i_slice_max_size = x4->slice_max_size;
359 if (x4->fastfirstpass)
360 x264_param_apply_fastfirstpass(&x4->params);
363 if (x264_param_apply_profile(&x4->params, x4->profile) < 0) {
364 av_log(avctx, AV_LOG_ERROR, "Error setting profile %s.\n", x4->profile);
365 return AVERROR(EINVAL);
368 x4->params.i_width = avctx->width;
369 x4->params.i_height = avctx->height;
370 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
371 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
372 x4->params.i_fps_num = x4->params.i_timebase_den = avctx->time_base.den;
373 x4->params.i_fps_den = x4->params.i_timebase_num = avctx->time_base.num;
375 x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR;
377 x4->params.i_threads = avctx->thread_count;
379 x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT;
381 x4->params.b_open_gop = !(avctx->flags & CODEC_FLAG_CLOSED_GOP);
383 x4->params.i_slice_count = avctx->slices;
385 x4->params.vui.b_fullrange = avctx->pix_fmt == PIX_FMT_YUVJ420P;
387 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
388 x4->params.b_repeat_headers = 0;
390 // update AVCodecContext with x264 parameters
391 avctx->has_b_frames = x4->params.i_bframe ?
392 x4->params.i_bframe_pyramid ? 2 : 1 : 0;
393 if (avctx->max_b_frames < 0)
394 avctx->max_b_frames = 0;
396 avctx->bit_rate = x4->params.rc.i_bitrate*1000;
398 x4->enc = x264_encoder_open(&x4->params);
402 avctx->coded_frame = &x4->out_pic;
404 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
408 s = x264_encoder_headers(x4->enc, &nal, &nnal);
410 for (i = 0; i < nnal; i++)
411 if (nal[i].i_type == NAL_SEI)
412 av_log(avctx, AV_LOG_INFO, "%s\n", nal[i].p_payload+25);
414 avctx->extradata = av_malloc(s);
415 avctx->extradata_size = encode_nals(avctx, avctx->extradata, s, nal, nnal, 1);
421 static const enum PixelFormat pix_fmts_8bit[] = {
428 static const enum PixelFormat pix_fmts_9bit[] = {
433 static const enum PixelFormat pix_fmts_10bit[] = {
440 static av_cold void X264_init_static(AVCodec *codec)
442 if (x264_bit_depth == 8)
443 codec->pix_fmts = pix_fmts_8bit;
444 else if (x264_bit_depth == 9)
445 codec->pix_fmts = pix_fmts_9bit;
446 else if (x264_bit_depth == 10)
447 codec->pix_fmts = pix_fmts_10bit;
450 #define OFFSET(x) offsetof(X264Context, x)
451 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
452 static const AVOption options[] = {
453 { "preset", "Set the encoding preset (cf. x264 --fullhelp)", OFFSET(preset), AV_OPT_TYPE_STRING, { .str = "medium" }, 0, 0, VE},
454 { "tune", "Tune the encoding params (cf. x264 --fullhelp)", OFFSET(tune), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
455 { "profile", "Set profile restrictions (cf. x264 --fullhelp) ", OFFSET(profile), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
456 { "fastfirstpass", "Use fast settings when encoding first pass", OFFSET(fastfirstpass), AV_OPT_TYPE_INT, { 1 }, 0, 1, VE},
457 { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
458 { "crf_max", "In CRF mode, prevents VBV from lowering quality beyond this point.",OFFSET(crf_max), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
459 { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
460 { "aq-mode", "AQ method", OFFSET(aq_mode), AV_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "aq_mode"},
461 { "none", NULL, 0, AV_OPT_TYPE_CONST, {X264_AQ_NONE}, INT_MIN, INT_MAX, VE, "aq_mode" },
462 { "variance", "Variance AQ (complexity mask)", 0, AV_OPT_TYPE_CONST, {X264_AQ_VARIANCE}, INT_MIN, INT_MAX, VE, "aq_mode" },
463 { "autovariance", "Auto-variance AQ (experimental)", 0, AV_OPT_TYPE_CONST, {X264_AQ_AUTOVARIANCE}, INT_MIN, INT_MAX, VE, "aq_mode" },
464 { "aq-strength", "AQ strength. Reduces blocking and blurring in flat and textured areas.", OFFSET(aq_strength), AV_OPT_TYPE_FLOAT, {-1}, -1, FLT_MAX, VE},
465 { "psy", "Use psychovisual optimizations.", OFFSET(psy), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE },
466 { "psy-rd", "Strength of psychovisual optimization, in <psy-rd>:<psy-trellis> format.", OFFSET(psy_rd), AV_OPT_TYPE_STRING, {0 }, 0, 0, VE},
467 { "rc-lookahead", "Number of frames to look ahead for frametype and ratecontrol", OFFSET(rc_lookahead), AV_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
468 { "weightb", "Weighted prediction for B-frames.", OFFSET(weightb), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE },
469 { "weightp", "Weighted prediction analysis method.", OFFSET(weightp), AV_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "weightp" },
470 { "none", NULL, 0, AV_OPT_TYPE_CONST, {X264_WEIGHTP_NONE}, INT_MIN, INT_MAX, VE, "weightp" },
471 { "simple", NULL, 0, AV_OPT_TYPE_CONST, {X264_WEIGHTP_SIMPLE}, INT_MIN, INT_MAX, VE, "weightp" },
472 { "smart", NULL, 0, AV_OPT_TYPE_CONST, {X264_WEIGHTP_SMART}, INT_MIN, INT_MAX, VE, "weightp" },
473 { "ssim", "Calculate and print SSIM stats.", OFFSET(ssim), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE },
474 { "intra-refresh", "Use Periodic Intra Refresh instead of IDR frames.",OFFSET(intra_refresh),AV_OPT_TYPE_INT, {-1 }, -1, 1, VE },
475 { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {INT_MIN}, INT_MIN, INT_MAX, VE },
476 { "b-pyramid", "Keep some B-frames as references.", OFFSET(b_pyramid), AV_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "b_pyramid" },
477 { "none", NULL, 0, AV_OPT_TYPE_CONST, {X264_B_PYRAMID_NONE}, INT_MIN, INT_MAX, VE, "b_pyramid" },
478 { "strict", "Strictly hierarchical pyramid", 0, AV_OPT_TYPE_CONST, {X264_B_PYRAMID_STRICT}, INT_MIN, INT_MAX, VE, "b_pyramid" },
479 { "normal", "Non-strict (not Blu-ray compatible)", 0, AV_OPT_TYPE_CONST, {X264_B_PYRAMID_NORMAL}, INT_MIN, INT_MAX, VE, "b_pyramid" },
480 { "mixed-refs", "One reference per partition, as opposed to one reference per macroblock", OFFSET(mixed_refs), AV_OPT_TYPE_INT, {-1}, -1, 1, VE },
481 { "8x8dct", "High profile 8x8 transform.", OFFSET(dct8x8), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE},
482 { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE},
483 { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE},
484 { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE},
485 { "deblock", "Loop filter parameters, in <alpha:beta> form.", OFFSET(deblock), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
486 { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE},
487 { "partitions", "A comma-separated list of partitions to consider. "
488 "Possible values: p8x8, p4x4, b8x8, i8x8, i4x4, none, all", OFFSET(partitions), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
489 { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "direct-pred" },
490 { "none", NULL, 0, AV_OPT_TYPE_CONST, { X264_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
491 { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { X264_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
492 { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { X264_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
493 { "auto", NULL, 0, AV_OPT_TYPE_CONST, { X264_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
494 { "slice-max-size","Constant quantization parameter rate control method",OFFSET(slice_max_size), AV_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
498 static const AVClass class = {
499 .class_name = "libx264",
500 .item_name = av_default_item_name,
502 .version = LIBAVUTIL_VERSION_INT,
505 static const AVCodecDefault x264_defaults[] = {
515 { "sc_threshold", "-1" },
518 { "me_range", "-1" },
519 { "me_method", "-1" },
521 { "b_strategy", "-1" },
522 { "keyint_min", "-1" },
525 { "threads", AV_STRINGIFY(X264_THREADS_AUTO) },
529 AVCodec ff_libx264_encoder = {
531 .type = AVMEDIA_TYPE_VIDEO,
533 .priv_data_size = sizeof(X264Context),
535 .encode = X264_frame,
537 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
538 .long_name = NULL_IF_CONFIG_SMALL("libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
539 .priv_class = &class,
540 .defaults = x264_defaults,
541 .init_static_data = X264_init_static,