libx264: add preset, tune, profile and fastfirstpass private options
[libav.git] / libavcodec / libx264.c
1 /*
2 * H.264 encoding using the x264 library
3 * Copyright (C) 2005 Mans Rullgard <mans@mansr.com>
4 *
5 * This file is part of Libav.
6 *
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.
11 *
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.
16 *
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
20 */
21
22 #include "libavutil/opt.h"
23 #include "avcodec.h"
24 #include <x264.h>
25 #include <math.h>
26 #include <stdio.h>
27 #include <stdlib.h>
28 #include <string.h>
29
30 typedef struct X264Context {
31 AVClass *class;
32 x264_param_t params;
33 x264_t *enc;
34 x264_picture_t pic;
35 uint8_t *sei;
36 int sei_size;
37 AVFrame out_pic;
38 char *preset;
39 char *tune;
40 char *profile;
41 int fastfirstpass;
42 } X264Context;
43
44 static void X264_log(void *p, int level, const char *fmt, va_list args)
45 {
46 static const int level_map[] = {
47 [X264_LOG_ERROR] = AV_LOG_ERROR,
48 [X264_LOG_WARNING] = AV_LOG_WARNING,
49 [X264_LOG_INFO] = AV_LOG_INFO,
50 [X264_LOG_DEBUG] = AV_LOG_DEBUG
51 };
52
53 if (level < 0 || level > X264_LOG_DEBUG)
54 return;
55
56 av_vlog(p, level_map[level], fmt, args);
57 }
58
59
60 static int encode_nals(AVCodecContext *ctx, uint8_t *buf, int size,
61 x264_nal_t *nals, int nnal, int skip_sei)
62 {
63 X264Context *x4 = ctx->priv_data;
64 uint8_t *p = buf;
65 int i;
66
67 /* Write the SEI as part of the first frame. */
68 if (x4->sei_size > 0 && nnal > 0) {
69 memcpy(p, x4->sei, x4->sei_size);
70 p += x4->sei_size;
71 x4->sei_size = 0;
72 }
73
74 for (i = 0; i < nnal; i++){
75 /* Don't put the SEI in extradata. */
76 if (skip_sei && nals[i].i_type == NAL_SEI) {
77 x4->sei_size = nals[i].i_payload;
78 x4->sei = av_malloc(x4->sei_size);
79 memcpy(x4->sei, nals[i].p_payload, nals[i].i_payload);
80 continue;
81 }
82 memcpy(p, nals[i].p_payload, nals[i].i_payload);
83 p += nals[i].i_payload;
84 }
85
86 return p - buf;
87 }
88
89 static int X264_frame(AVCodecContext *ctx, uint8_t *buf,
90 int bufsize, void *data)
91 {
92 X264Context *x4 = ctx->priv_data;
93 AVFrame *frame = data;
94 x264_nal_t *nal;
95 int nnal, i;
96 x264_picture_t pic_out;
97
98 x264_picture_init( &x4->pic );
99 x4->pic.img.i_csp = X264_CSP_I420;
100 x4->pic.img.i_plane = 3;
101
102 if (frame) {
103 for (i = 0; i < 3; i++) {
104 x4->pic.img.plane[i] = frame->data[i];
105 x4->pic.img.i_stride[i] = frame->linesize[i];
106 }
107
108 x4->pic.i_pts = frame->pts;
109 x4->pic.i_type =
110 frame->pict_type == AV_PICTURE_TYPE_I ? X264_TYPE_KEYFRAME :
111 frame->pict_type == AV_PICTURE_TYPE_P ? X264_TYPE_P :
112 frame->pict_type == AV_PICTURE_TYPE_B ? X264_TYPE_B :
113 X264_TYPE_AUTO;
114 if (x4->params.b_tff != frame->top_field_first) {
115 x4->params.b_tff = frame->top_field_first;
116 x264_encoder_reconfig(x4->enc, &x4->params);
117 }
118 }
119
120 do {
121 if (x264_encoder_encode(x4->enc, &nal, &nnal, frame? &x4->pic: NULL, &pic_out) < 0)
122 return -1;
123
124 bufsize = encode_nals(ctx, buf, bufsize, nal, nnal, 0);
125 if (bufsize < 0)
126 return -1;
127 } while (!bufsize && !frame && x264_encoder_delayed_frames(x4->enc));
128
129 /* FIXME: libx264 now provides DTS, but AVFrame doesn't have a field for it. */
130 x4->out_pic.pts = pic_out.i_pts;
131
132 switch (pic_out.i_type) {
133 case X264_TYPE_IDR:
134 case X264_TYPE_I:
135 x4->out_pic.pict_type = AV_PICTURE_TYPE_I;
136 break;
137 case X264_TYPE_P:
138 x4->out_pic.pict_type = AV_PICTURE_TYPE_P;
139 break;
140 case X264_TYPE_B:
141 case X264_TYPE_BREF:
142 x4->out_pic.pict_type = AV_PICTURE_TYPE_B;
143 break;
144 }
145
146 x4->out_pic.key_frame = pic_out.b_keyframe;
147 x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
148
149 return bufsize;
150 }
151
152 static av_cold int X264_close(AVCodecContext *avctx)
153 {
154 X264Context *x4 = avctx->priv_data;
155
156 av_freep(&avctx->extradata);
157 av_free(x4->sei);
158
159 if (x4->enc)
160 x264_encoder_close(x4->enc);
161
162 return 0;
163 }
164
165 static av_cold int X264_init(AVCodecContext *avctx)
166 {
167 X264Context *x4 = avctx->priv_data;
168
169 x4->sei_size = 0;
170 x264_param_default(&x4->params);
171
172 x4->params.i_keyint_max = avctx->gop_size;
173
174 x4->params.i_bframe = avctx->max_b_frames;
175 x4->params.b_cabac = avctx->coder_type == FF_CODER_TYPE_AC;
176 x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
177 x4->params.i_bframe_bias = avctx->bframebias;
178 x4->params.i_bframe_pyramid = avctx->flags2 & CODEC_FLAG2_BPYRAMID ? X264_B_PYRAMID_NORMAL : X264_B_PYRAMID_NONE;
179 avctx->has_b_frames = avctx->flags2 & CODEC_FLAG2_BPYRAMID ? 2 : !!avctx->max_b_frames;
180
181 x4->params.i_keyint_min = avctx->keyint_min;
182 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
183 x4->params.i_keyint_min = x4->params.i_keyint_max;
184
185 x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
186
187 x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER;
188 x4->params.i_deblocking_filter_alphac0 = avctx->deblockalpha;
189 x4->params.i_deblocking_filter_beta = avctx->deblockbeta;
190
191 x4->params.rc.i_qp_min = avctx->qmin;
192 x4->params.rc.i_qp_max = avctx->qmax;
193 x4->params.rc.i_qp_step = avctx->max_qdiff;
194
195 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
196 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
197 x4->params.rc.f_complexity_blur = avctx->complexityblur;
198
199 x4->params.i_frame_reference = avctx->refs;
200
201 x4->params.analyse.inter = 0;
202 if (avctx->partitions) {
203 if (avctx->partitions & X264_PART_I4X4)
204 x4->params.analyse.inter |= X264_ANALYSE_I4x4;
205 if (avctx->partitions & X264_PART_I8X8)
206 x4->params.analyse.inter |= X264_ANALYSE_I8x8;
207 if (avctx->partitions & X264_PART_P8X8)
208 x4->params.analyse.inter |= X264_ANALYSE_PSUB16x16;
209 if (avctx->partitions & X264_PART_P4X4)
210 x4->params.analyse.inter |= X264_ANALYSE_PSUB8x8;
211 if (avctx->partitions & X264_PART_B8X8)
212 x4->params.analyse.inter |= X264_ANALYSE_BSUB16x16;
213 }
214
215 x4->params.analyse.i_direct_mv_pred = avctx->directpred;
216
217 x4->params.analyse.b_weighted_bipred = avctx->flags2 & CODEC_FLAG2_WPRED;
218 x4->params.analyse.i_weighted_pred = avctx->weighted_p_pred;
219
220 if (avctx->me_method == ME_EPZS)
221 x4->params.analyse.i_me_method = X264_ME_DIA;
222 else if (avctx->me_method == ME_HEX)
223 x4->params.analyse.i_me_method = X264_ME_HEX;
224 else if (avctx->me_method == ME_UMH)
225 x4->params.analyse.i_me_method = X264_ME_UMH;
226 else if (avctx->me_method == ME_FULL)
227 x4->params.analyse.i_me_method = X264_ME_ESA;
228 else if (avctx->me_method == ME_TESA)
229 x4->params.analyse.i_me_method = X264_ME_TESA;
230 else x4->params.analyse.i_me_method = X264_ME_HEX;
231
232 x4->params.rc.i_aq_mode = avctx->aq_mode;
233 x4->params.rc.f_aq_strength = avctx->aq_strength;
234 x4->params.rc.i_lookahead = avctx->rc_lookahead;
235
236 x4->params.analyse.b_psy = avctx->flags2 & CODEC_FLAG2_PSY;
237 x4->params.analyse.f_psy_rd = avctx->psy_rd;
238 x4->params.analyse.f_psy_trellis = avctx->psy_trellis;
239
240 x4->params.analyse.i_me_range = avctx->me_range;
241 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
242
243 x4->params.analyse.b_mixed_references = avctx->flags2 & CODEC_FLAG2_MIXED_REFS;
244 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
245 x4->params.analyse.b_transform_8x8 = avctx->flags2 & CODEC_FLAG2_8X8DCT;
246 x4->params.analyse.b_fast_pskip = avctx->flags2 & CODEC_FLAG2_FASTPSKIP;
247
248 x4->params.analyse.i_trellis = avctx->trellis;
249 x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
250
251 if (avctx->level > 0)
252 x4->params.i_level_idc = avctx->level;
253
254 if (x4->preset || x4->tune)
255 if (x264_param_default_preset(&x4->params, x4->preset, x4->tune) < 0) {
256 av_log(avctx, AV_LOG_ERROR, "Error setting preset/tune %s/%s.\n", x4->preset, x4->tune);
257 return AVERROR(EINVAL);
258 }
259
260 x4->params.pf_log = X264_log;
261 x4->params.p_log_private = avctx;
262 x4->params.i_log_level = X264_LOG_DEBUG;
263
264 x4->params.b_intra_refresh = avctx->flags2 & CODEC_FLAG2_INTRA_REFRESH;
265 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
266 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
267 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
268 x4->params.rc.b_stat_write = avctx->flags & CODEC_FLAG_PASS1;
269 if (avctx->flags & CODEC_FLAG_PASS2) {
270 x4->params.rc.b_stat_read = 1;
271 } else {
272 if (avctx->crf) {
273 x4->params.rc.i_rc_method = X264_RC_CRF;
274 x4->params.rc.f_rf_constant = avctx->crf;
275 x4->params.rc.f_rf_constant_max = avctx->crf_max;
276 } else if (avctx->cqp > -1) {
277 x4->params.rc.i_rc_method = X264_RC_CQP;
278 x4->params.rc.i_qp_constant = avctx->cqp;
279 }
280 }
281
282 // if neither crf nor cqp modes are selected we have to enable the RC
283 // we do it this way because we cannot check if the bitrate has been set
284 if (!(avctx->crf || (avctx->cqp > -1)))
285 x4->params.rc.i_rc_method = X264_RC_ABR;
286
287 if (avctx->rc_buffer_size && avctx->rc_initial_buffer_occupancy &&
288 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
289 x4->params.rc.f_vbv_buffer_init =
290 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
291 }
292
293 x4->params.rc.b_mb_tree = !!(avctx->flags2 & CODEC_FLAG2_MBTREE);
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;
297
298 if (x4->fastfirstpass)
299 x264_param_apply_fastfirstpass(&x4->params);
300
301 if (x4->profile)
302 if (x264_param_apply_profile(&x4->params, x4->profile) < 0) {
303 av_log(avctx, AV_LOG_ERROR, "Error setting profile %s.\n", x4->profile);
304 return AVERROR(EINVAL);
305 }
306
307 x4->params.i_width = avctx->width;
308 x4->params.i_height = avctx->height;
309 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
310 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
311 x4->params.i_fps_num = x4->params.i_timebase_den = avctx->time_base.den;
312 x4->params.i_fps_den = x4->params.i_timebase_num = avctx->time_base.num;
313
314 x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR;
315 x4->params.analyse.b_ssim = avctx->flags2 & CODEC_FLAG2_SSIM;
316
317 x4->params.b_aud = avctx->flags2 & CODEC_FLAG2_AUD;
318
319 x4->params.i_threads = avctx->thread_count;
320
321 x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT;
322
323 x4->params.b_open_gop = !(avctx->flags & CODEC_FLAG_CLOSED_GOP);
324
325 x4->params.i_slice_count = avctx->slices;
326
327 x4->params.vui.b_fullrange = avctx->pix_fmt == PIX_FMT_YUVJ420P;
328
329 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
330 x4->params.b_repeat_headers = 0;
331
332 // update AVCodecContext with x264 parameters
333 avctx->has_b_frames = x4->params.i_bframe ?
334 x4->params.i_bframe_pyramid ? 2 : 1 : 0;
335 avctx->bit_rate = x4->params.rc.i_bitrate*1000;
336 avctx->crf = x4->params.rc.f_rf_constant;
337
338 x4->enc = x264_encoder_open(&x4->params);
339 if (!x4->enc)
340 return -1;
341
342 avctx->coded_frame = &x4->out_pic;
343
344 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
345 x264_nal_t *nal;
346 int nnal, s, i;
347
348 s = x264_encoder_headers(x4->enc, &nal, &nnal);
349
350 for (i = 0; i < nnal; i++)
351 if (nal[i].i_type == NAL_SEI)
352 av_log(avctx, AV_LOG_INFO, "%s\n", nal[i].p_payload+25);
353
354 avctx->extradata = av_malloc(s);
355 avctx->extradata_size = encode_nals(avctx, avctx->extradata, s, nal, nnal, 1);
356 }
357
358 return 0;
359 }
360
361 #define OFFSET(x) offsetof(X264Context, x)
362 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
363 static const AVOption options[] = {
364 { "preset", "Set the encoding preset", OFFSET(preset), FF_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
365 { "tune", "Tune the encoding params", OFFSET(tune), FF_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
366 { "profile", "Set profile restrictions", OFFSET(profile), FF_OPT_TYPE_STRING, { 0 }, 0, 0, VE},
367 { "fastfirstpass", "Use fast settings when encoding first pass", OFFSET(fastfirstpass), FF_OPT_TYPE_INT, { 1 }, 0, 1, VE},
368 { NULL },
369 };
370
371 static const AVClass class = {
372 .class_name = "libx264",
373 .item_name = av_default_item_name,
374 .option = options,
375 .version = LIBAVUTIL_VERSION_INT,
376 };
377
378 AVCodec ff_libx264_encoder = {
379 .name = "libx264",
380 .type = AVMEDIA_TYPE_VIDEO,
381 .id = CODEC_ID_H264,
382 .priv_data_size = sizeof(X264Context),
383 .init = X264_init,
384 .encode = X264_frame,
385 .close = X264_close,
386 .capabilities = CODEC_CAP_DELAY,
387 .pix_fmts = (const enum PixelFormat[]) { PIX_FMT_YUV420P, PIX_FMT_YUVJ420P, PIX_FMT_NONE },
388 .long_name = NULL_IF_CONFIG_SMALL("libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
389 .priv_class = &class,
390 };