4 * Copyright (c) 2013-2014 Derek Buitenhuis
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 #define X265_API_IMPORTS 1
30 #include "libavutil/internal.h"
31 #include "libavutil/common.h"
32 #include "libavutil/opt.h"
33 #include "libavutil/pixdesc.h"
37 typedef struct libx265Context {
40 x265_encoder *encoder;
52 static int is_keyframe(NalUnitType naltype)
55 case NAL_UNIT_CODED_SLICE_BLA_W_LP:
56 case NAL_UNIT_CODED_SLICE_BLA_W_RADL:
57 case NAL_UNIT_CODED_SLICE_BLA_N_LP:
58 case NAL_UNIT_CODED_SLICE_IDR_W_RADL:
59 case NAL_UNIT_CODED_SLICE_IDR_N_LP:
60 case NAL_UNIT_CODED_SLICE_CRA:
67 static av_cold int libx265_encode_close(AVCodecContext *avctx)
69 libx265Context *ctx = avctx->priv_data;
71 ctx->api->param_free(ctx->params);
74 ctx->api->encoder_close(ctx->encoder);
79 static av_cold int libx265_encode_init(AVCodecContext *avctx)
81 libx265Context *ctx = avctx->priv_data;
82 AVCPBProperties *cpb_props = NULL;
84 ctx->api = x265_api_get(av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth);
86 ctx->api = x265_api_get(0);
88 ctx->params = ctx->api->param_alloc();
90 av_log(avctx, AV_LOG_ERROR, "Could not allocate x265 param structure.\n");
91 return AVERROR(ENOMEM);
94 if (ctx->api->param_default_preset(ctx->params, ctx->preset, ctx->tune) < 0) {
97 av_log(avctx, AV_LOG_ERROR, "Error setting preset/tune %s/%s.\n", ctx->preset, ctx->tune);
98 av_log(avctx, AV_LOG_INFO, "Possible presets:");
99 for (i = 0; x265_preset_names[i]; i++)
100 av_log(avctx, AV_LOG_INFO, " %s", x265_preset_names[i]);
102 av_log(avctx, AV_LOG_INFO, "\n");
103 av_log(avctx, AV_LOG_INFO, "Possible tunes:");
104 for (i = 0; x265_tune_names[i]; i++)
105 av_log(avctx, AV_LOG_INFO, " %s", x265_tune_names[i]);
107 av_log(avctx, AV_LOG_INFO, "\n");
109 return AVERROR(EINVAL);
112 ctx->params->frameNumThreads = avctx->thread_count;
113 ctx->params->fpsNum = avctx->time_base.den;
114 ctx->params->fpsDenom = avctx->time_base.num * avctx->ticks_per_frame;
115 ctx->params->sourceWidth = avctx->width;
116 ctx->params->sourceHeight = avctx->height;
117 ctx->params->bEnablePsnr = !!(avctx->flags & AV_CODEC_FLAG_PSNR);
118 ctx->params->bOpenGOP = !(avctx->flags & AV_CODEC_FLAG_CLOSED_GOP);
120 /* Tune the CTU size based on input resolution. */
121 if (ctx->params->sourceWidth < 64 || ctx->params->sourceHeight < 64)
122 ctx->params->maxCUSize = 32;
123 if (ctx->params->sourceWidth < 32 || ctx->params->sourceHeight < 32)
124 ctx->params->maxCUSize = 16;
125 if (ctx->params->sourceWidth < 16 || ctx->params->sourceHeight < 16) {
126 av_log(avctx, AV_LOG_ERROR, "Image size is too small (%dx%d).\n",
127 ctx->params->sourceWidth, ctx->params->sourceHeight);
128 return AVERROR(EINVAL);
131 if ((avctx->color_primaries <= AVCOL_PRI_SMPTE432 &&
132 avctx->color_primaries != AVCOL_PRI_UNSPECIFIED) ||
133 (avctx->color_trc <= AVCOL_TRC_ARIB_STD_B67 &&
134 avctx->color_trc != AVCOL_TRC_UNSPECIFIED) ||
135 (avctx->colorspace <= AVCOL_SPC_ICTCP &&
136 avctx->colorspace != AVCOL_SPC_UNSPECIFIED)) {
138 ctx->params->vui.bEnableVideoSignalTypePresentFlag = 1;
139 ctx->params->vui.bEnableColorDescriptionPresentFlag = 1;
141 // x265 validates the parameters internally
142 ctx->params->vui.colorPrimaries = avctx->color_primaries;
143 ctx->params->vui.transferCharacteristics = avctx->color_trc;
144 ctx->params->vui.matrixCoeffs = avctx->colorspace;
147 if (avctx->sample_aspect_ratio.num > 0 && avctx->sample_aspect_ratio.den > 0) {
149 int sar_num, sar_den;
151 av_reduce(&sar_num, &sar_den,
152 avctx->sample_aspect_ratio.num,
153 avctx->sample_aspect_ratio.den, 65535);
154 snprintf(sar, sizeof(sar), "%d:%d", sar_num, sar_den);
155 if (ctx->api->param_parse(ctx->params, "sar", sar) == X265_PARAM_BAD_VALUE) {
156 av_log(avctx, AV_LOG_ERROR, "Invalid SAR: %d:%d.\n", sar_num, sar_den);
157 return AVERROR_INVALIDDATA;
161 switch (avctx->pix_fmt) {
162 case AV_PIX_FMT_YUV420P:
163 case AV_PIX_FMT_YUV420P10:
164 case AV_PIX_FMT_YUV420P12:
165 ctx->params->internalCsp = X265_CSP_I420;
167 case AV_PIX_FMT_YUV422P:
168 case AV_PIX_FMT_YUV422P10:
169 case AV_PIX_FMT_YUV422P12:
170 ctx->params->internalCsp = X265_CSP_I422;
172 case AV_PIX_FMT_GBRP:
173 case AV_PIX_FMT_GBRP10:
174 case AV_PIX_FMT_GBRP12:
175 ctx->params->vui.matrixCoeffs = AVCOL_SPC_RGB;
176 ctx->params->vui.bEnableVideoSignalTypePresentFlag = 1;
177 ctx->params->vui.bEnableColorDescriptionPresentFlag = 1;
178 case AV_PIX_FMT_YUV444P:
179 case AV_PIX_FMT_YUV444P10:
180 case AV_PIX_FMT_YUV444P12:
181 ctx->params->internalCsp = X265_CSP_I444;
183 case AV_PIX_FMT_GRAY8:
184 case AV_PIX_FMT_GRAY10:
185 case AV_PIX_FMT_GRAY12:
186 if (ctx->api->api_build_number < 85) {
187 av_log(avctx, AV_LOG_ERROR,
188 "libx265 version is %d, must be at least 85 for gray encoding.\n",
189 ctx->api->api_build_number);
190 return AVERROR_INVALIDDATA;
192 ctx->params->internalCsp = X265_CSP_I400;
199 snprintf(crf, sizeof(crf), "%2.2f", ctx->crf);
200 if (ctx->api->param_parse(ctx->params, "crf", crf) == X265_PARAM_BAD_VALUE) {
201 av_log(avctx, AV_LOG_ERROR, "Invalid crf: %2.2f.\n", ctx->crf);
202 return AVERROR(EINVAL);
204 } else if (avctx->bit_rate > 0) {
205 ctx->params->rc.bitrate = avctx->bit_rate / 1000;
206 ctx->params->rc.rateControlMode = X265_RC_ABR;
209 ctx->params->rc.vbvBufferSize = avctx->rc_buffer_size / 1000;
210 ctx->params->rc.vbvMaxBitrate = avctx->rc_max_rate / 1000;
212 cpb_props = ff_add_cpb_side_data(avctx);
214 return AVERROR(ENOMEM);
215 cpb_props->buffer_size = ctx->params->rc.vbvBufferSize * 1000;
216 cpb_props->max_bitrate = ctx->params->rc.vbvMaxBitrate * 1000;
217 cpb_props->avg_bitrate = ctx->params->rc.bitrate * 1000;
219 if (!(avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER))
220 ctx->params->bRepeatHeaders = 1;
222 if (ctx->x265_opts) {
223 AVDictionary *dict = NULL;
224 AVDictionaryEntry *en = NULL;
226 if (!av_dict_parse_string(&dict, ctx->x265_opts, "=", ":", 0)) {
227 while ((en = av_dict_get(dict, "", en, AV_DICT_IGNORE_SUFFIX))) {
228 int parse_ret = ctx->api->param_parse(ctx->params, en->key, en->value);
231 case X265_PARAM_BAD_NAME:
232 av_log(avctx, AV_LOG_WARNING,
233 "Unknown option: %s.\n", en->key);
235 case X265_PARAM_BAD_VALUE:
236 av_log(avctx, AV_LOG_WARNING,
237 "Invalid value for %s: %s.\n", en->key, en->value);
247 if (ctx->params->rc.vbvBufferSize && avctx->rc_initial_buffer_occupancy > 1000 &&
248 ctx->params->rc.vbvBufferInit == 0.9) {
249 ctx->params->rc.vbvBufferInit = (float)avctx->rc_initial_buffer_occupancy / 1000;
253 if (ctx->api->param_apply_profile(ctx->params, ctx->profile) < 0) {
255 av_log(avctx, AV_LOG_ERROR, "Invalid or incompatible profile set: %s.\n", ctx->profile);
256 av_log(avctx, AV_LOG_INFO, "Possible profiles:");
257 for (i = 0; x265_profile_names[i]; i++)
258 av_log(avctx, AV_LOG_INFO, " %s", x265_profile_names[i]);
259 av_log(avctx, AV_LOG_INFO, "\n");
260 return AVERROR(EINVAL);
264 ctx->encoder = ctx->api->encoder_open(ctx->params);
266 av_log(avctx, AV_LOG_ERROR, "Cannot open libx265 encoder.\n");
267 libx265_encode_close(avctx);
268 return AVERROR_INVALIDDATA;
271 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
275 avctx->extradata_size = ctx->api->encoder_headers(ctx->encoder, &nal, &nnal);
276 if (avctx->extradata_size <= 0) {
277 av_log(avctx, AV_LOG_ERROR, "Cannot encode headers.\n");
278 libx265_encode_close(avctx);
279 return AVERROR_INVALIDDATA;
282 avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
283 if (!avctx->extradata) {
284 av_log(avctx, AV_LOG_ERROR,
285 "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size);
286 libx265_encode_close(avctx);
287 return AVERROR(ENOMEM);
290 memcpy(avctx->extradata, nal[0].payload, avctx->extradata_size);
296 static av_cold int libx265_encode_set_roi(libx265Context *ctx, const AVFrame *frame, x265_picture* pic)
298 AVFrameSideData *sd = av_frame_get_side_data(frame, AV_FRAME_DATA_REGIONS_OF_INTEREST);
300 if (ctx->params->rc.aqMode == X265_AQ_NONE) {
301 av_log(ctx, AV_LOG_WARNING, "Adaptive quantization must be enabled to use ROI encoding, skipping ROI.\n");
303 /* 8x8 block when qg-size is 8, 16*16 block otherwise. */
304 int mb_size = (ctx->params->rc.qgSize == 8) ? 8 : 16;
305 int mbx = (frame->width + mb_size - 1) / mb_size;
306 int mby = (frame->height + mb_size - 1) / mb_size;
308 AVRegionOfInterest *roi;
309 float *qoffsets; /* will be freed after encode is called. */
310 qoffsets = av_mallocz_array(mbx * mby, sizeof(*qoffsets));
312 return AVERROR(ENOMEM);
314 nb_rois = sd->size / sizeof(AVRegionOfInterest);
315 roi = (AVRegionOfInterest*)sd->data;
316 for (int count = 0; count < nb_rois; count++) {
317 int starty = FFMIN(mby, roi->top / mb_size);
318 int endy = FFMIN(mby, (roi->bottom + mb_size - 1)/ mb_size);
319 int startx = FFMIN(mbx, roi->left / mb_size);
320 int endx = FFMIN(mbx, (roi->right + mb_size - 1)/ mb_size);
323 if (roi->self_size == 0) {
325 av_log(ctx, AV_LOG_ERROR, "AVRegionOfInterest.self_size must be set to sizeof(AVRegionOfInterest).\n");
326 return AVERROR(EINVAL);
329 if (roi->qoffset.den == 0) {
331 av_log(ctx, AV_LOG_ERROR, "AVRegionOfInterest.qoffset.den must not be zero.\n");
332 return AVERROR(EINVAL);
334 qoffset = roi->qoffset.num * 1.0f / roi->qoffset.den;
335 qoffset = av_clipf(qoffset, -1.0f, 1.0f);
337 /* qp range of x265 is from 0 to 51, just choose 25 as the scale value,
338 * so the range of final qoffset is [-25.0, 25.0].
340 qoffset = qoffset * 25;
342 for (int y = starty; y < endy; y++)
343 for (int x = startx; x < endx; x++)
344 qoffsets[x + y*mbx] = qoffset;
346 roi = (AVRegionOfInterest*)((char*)roi + roi->self_size);
349 pic->quantOffsets = qoffsets;
355 static int libx265_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
356 const AVFrame *pic, int *got_packet)
358 libx265Context *ctx = avctx->priv_data;
359 x265_picture x265pic;
360 x265_picture x265pic_out = { 0 };
368 ctx->api->picture_init(ctx->params, &x265pic);
371 for (i = 0; i < 3; i++) {
372 x265pic.planes[i] = pic->data[i];
373 x265pic.stride[i] = pic->linesize[i];
376 x265pic.pts = pic->pts;
377 x265pic.bitDepth = av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth;
379 x265pic.sliceType = pic->pict_type == AV_PICTURE_TYPE_I ?
380 (ctx->forced_idr ? X265_TYPE_IDR : X265_TYPE_I) :
381 pic->pict_type == AV_PICTURE_TYPE_P ? X265_TYPE_P :
382 pic->pict_type == AV_PICTURE_TYPE_B ? X265_TYPE_B :
385 ret = libx265_encode_set_roi(ctx, pic, &x265pic);
390 ret = ctx->api->encoder_encode(ctx->encoder, &nal, &nnal,
391 pic ? &x265pic : NULL, &x265pic_out);
393 av_freep(&x265pic.quantOffsets);
396 return AVERROR_EXTERNAL;
401 for (i = 0; i < nnal; i++)
402 payload += nal[i].sizeBytes;
404 ret = ff_alloc_packet2(avctx, pkt, payload, payload);
406 av_log(avctx, AV_LOG_ERROR, "Error getting output packet.\n");
411 for (i = 0; i < nnal; i++) {
412 memcpy(dst, nal[i].payload, nal[i].sizeBytes);
413 dst += nal[i].sizeBytes;
415 if (is_keyframe(nal[i].type))
416 pkt->flags |= AV_PKT_FLAG_KEY;
419 pkt->pts = x265pic_out.pts;
420 pkt->dts = x265pic_out.dts;
422 #if FF_API_CODED_FRAME
423 FF_DISABLE_DEPRECATION_WARNINGS
424 switch (x265pic_out.sliceType) {
427 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
430 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
433 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_B;
436 FF_ENABLE_DEPRECATION_WARNINGS
439 #if X265_BUILD >= 130
440 if (x265pic_out.sliceType == X265_TYPE_B)
442 if (x265pic_out.frameData.sliceType == 'b')
444 pkt->flags |= AV_PKT_FLAG_DISPOSABLE;
450 static const enum AVPixelFormat x265_csp_eight[] = {
459 static const enum AVPixelFormat x265_csp_ten[] = {
464 AV_PIX_FMT_YUV420P10,
465 AV_PIX_FMT_YUV422P10,
466 AV_PIX_FMT_YUV444P10,
473 static const enum AVPixelFormat x265_csp_twelve[] = {
478 AV_PIX_FMT_YUV420P10,
479 AV_PIX_FMT_YUV422P10,
480 AV_PIX_FMT_YUV444P10,
482 AV_PIX_FMT_YUV420P12,
483 AV_PIX_FMT_YUV422P12,
484 AV_PIX_FMT_YUV444P12,
492 static av_cold void libx265_encode_init_csp(AVCodec *codec)
494 if (x265_api_get(12))
495 codec->pix_fmts = x265_csp_twelve;
496 else if (x265_api_get(10))
497 codec->pix_fmts = x265_csp_ten;
498 else if (x265_api_get(8))
499 codec->pix_fmts = x265_csp_eight;
502 #define OFFSET(x) offsetof(libx265Context, x)
503 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
504 static const AVOption options[] = {
505 { "crf", "set the x265 crf", OFFSET(crf), AV_OPT_TYPE_FLOAT, { .dbl = -1 }, -1, FLT_MAX, VE },
506 { "forced-idr", "if forcing keyframes, force them as IDR frames", OFFSET(forced_idr),AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
507 { "preset", "set the x265 preset", OFFSET(preset), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
508 { "tune", "set the x265 tune parameter", OFFSET(tune), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
509 { "profile", "set the x265 profile", OFFSET(profile), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
510 { "x265-params", "set the x265 configuration using a :-separated list of key=value parameters", OFFSET(x265_opts), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
514 static const AVClass class = {
515 .class_name = "libx265",
516 .item_name = av_default_item_name,
518 .version = LIBAVUTIL_VERSION_INT,
521 static const AVCodecDefault x265_defaults[] = {
526 AVCodec ff_libx265_encoder = {
528 .long_name = NULL_IF_CONFIG_SMALL("libx265 H.265 / HEVC"),
529 .type = AVMEDIA_TYPE_VIDEO,
530 .id = AV_CODEC_ID_HEVC,
531 .init = libx265_encode_init,
532 .init_static_data = libx265_encode_init_csp,
533 .encode2 = libx265_encode_frame,
534 .close = libx265_encode_close,
535 .priv_data_size = sizeof(libx265Context),
536 .priv_class = &class,
537 .defaults = x265_defaults,
538 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
539 .wrapper_name = "libx265",