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 * If the encoder does not support ROI then warn the first time we
53 * encounter a frame with ROI side data.
58 static int is_keyframe(NalUnitType naltype)
61 case NAL_UNIT_CODED_SLICE_BLA_W_LP:
62 case NAL_UNIT_CODED_SLICE_BLA_W_RADL:
63 case NAL_UNIT_CODED_SLICE_BLA_N_LP:
64 case NAL_UNIT_CODED_SLICE_IDR_W_RADL:
65 case NAL_UNIT_CODED_SLICE_IDR_N_LP:
66 case NAL_UNIT_CODED_SLICE_CRA:
73 static av_cold int libx265_encode_close(AVCodecContext *avctx)
75 libx265Context *ctx = avctx->priv_data;
77 ctx->api->param_free(ctx->params);
80 ctx->api->encoder_close(ctx->encoder);
85 static av_cold int libx265_encode_init(AVCodecContext *avctx)
87 libx265Context *ctx = avctx->priv_data;
88 AVCPBProperties *cpb_props = NULL;
90 ctx->api = x265_api_get(av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth);
92 ctx->api = x265_api_get(0);
94 ctx->params = ctx->api->param_alloc();
96 av_log(avctx, AV_LOG_ERROR, "Could not allocate x265 param structure.\n");
97 return AVERROR(ENOMEM);
100 if (ctx->api->param_default_preset(ctx->params, ctx->preset, ctx->tune) < 0) {
103 av_log(avctx, AV_LOG_ERROR, "Error setting preset/tune %s/%s.\n", ctx->preset, ctx->tune);
104 av_log(avctx, AV_LOG_INFO, "Possible presets:");
105 for (i = 0; x265_preset_names[i]; i++)
106 av_log(avctx, AV_LOG_INFO, " %s", x265_preset_names[i]);
108 av_log(avctx, AV_LOG_INFO, "\n");
109 av_log(avctx, AV_LOG_INFO, "Possible tunes:");
110 for (i = 0; x265_tune_names[i]; i++)
111 av_log(avctx, AV_LOG_INFO, " %s", x265_tune_names[i]);
113 av_log(avctx, AV_LOG_INFO, "\n");
115 return AVERROR(EINVAL);
118 ctx->params->frameNumThreads = avctx->thread_count;
119 if (avctx->framerate.num > 0 && avctx->framerate.den > 0) {
120 ctx->params->fpsNum = avctx->framerate.num;
121 ctx->params->fpsDenom = avctx->framerate.den;
123 ctx->params->fpsNum = avctx->time_base.den;
124 ctx->params->fpsDenom = avctx->time_base.num * avctx->ticks_per_frame;
126 ctx->params->sourceWidth = avctx->width;
127 ctx->params->sourceHeight = avctx->height;
128 ctx->params->bEnablePsnr = !!(avctx->flags & AV_CODEC_FLAG_PSNR);
129 ctx->params->bOpenGOP = !(avctx->flags & AV_CODEC_FLAG_CLOSED_GOP);
131 /* Tune the CTU size based on input resolution. */
132 if (ctx->params->sourceWidth < 64 || ctx->params->sourceHeight < 64)
133 ctx->params->maxCUSize = 32;
134 if (ctx->params->sourceWidth < 32 || ctx->params->sourceHeight < 32)
135 ctx->params->maxCUSize = 16;
136 if (ctx->params->sourceWidth < 16 || ctx->params->sourceHeight < 16) {
137 av_log(avctx, AV_LOG_ERROR, "Image size is too small (%dx%d).\n",
138 ctx->params->sourceWidth, ctx->params->sourceHeight);
139 return AVERROR(EINVAL);
143 ctx->params->vui.bEnableVideoSignalTypePresentFlag = 1;
145 ctx->params->vui.bEnableVideoFullRangeFlag = avctx->pix_fmt == AV_PIX_FMT_YUVJ420P ||
146 avctx->pix_fmt == AV_PIX_FMT_YUVJ422P ||
147 avctx->pix_fmt == AV_PIX_FMT_YUVJ444P ||
148 avctx->color_range == AVCOL_RANGE_JPEG;
150 if ((avctx->color_primaries <= AVCOL_PRI_SMPTE432 &&
151 avctx->color_primaries != AVCOL_PRI_UNSPECIFIED) ||
152 (avctx->color_trc <= AVCOL_TRC_ARIB_STD_B67 &&
153 avctx->color_trc != AVCOL_TRC_UNSPECIFIED) ||
154 (avctx->colorspace <= AVCOL_SPC_ICTCP &&
155 avctx->colorspace != AVCOL_SPC_UNSPECIFIED)) {
157 ctx->params->vui.bEnableColorDescriptionPresentFlag = 1;
159 // x265 validates the parameters internally
160 ctx->params->vui.colorPrimaries = avctx->color_primaries;
161 ctx->params->vui.transferCharacteristics = avctx->color_trc;
162 ctx->params->vui.matrixCoeffs = avctx->colorspace;
165 if (avctx->sample_aspect_ratio.num > 0 && avctx->sample_aspect_ratio.den > 0) {
167 int sar_num, sar_den;
169 av_reduce(&sar_num, &sar_den,
170 avctx->sample_aspect_ratio.num,
171 avctx->sample_aspect_ratio.den, 65535);
172 snprintf(sar, sizeof(sar), "%d:%d", sar_num, sar_den);
173 if (ctx->api->param_parse(ctx->params, "sar", sar) == X265_PARAM_BAD_VALUE) {
174 av_log(avctx, AV_LOG_ERROR, "Invalid SAR: %d:%d.\n", sar_num, sar_den);
175 return AVERROR_INVALIDDATA;
179 switch (avctx->pix_fmt) {
180 case AV_PIX_FMT_YUV420P:
181 case AV_PIX_FMT_YUV420P10:
182 case AV_PIX_FMT_YUV420P12:
183 ctx->params->internalCsp = X265_CSP_I420;
185 case AV_PIX_FMT_YUV422P:
186 case AV_PIX_FMT_YUV422P10:
187 case AV_PIX_FMT_YUV422P12:
188 ctx->params->internalCsp = X265_CSP_I422;
190 case AV_PIX_FMT_GBRP:
191 case AV_PIX_FMT_GBRP10:
192 case AV_PIX_FMT_GBRP12:
193 ctx->params->vui.matrixCoeffs = AVCOL_SPC_RGB;
194 ctx->params->vui.bEnableVideoSignalTypePresentFlag = 1;
195 ctx->params->vui.bEnableColorDescriptionPresentFlag = 1;
196 case AV_PIX_FMT_YUV444P:
197 case AV_PIX_FMT_YUV444P10:
198 case AV_PIX_FMT_YUV444P12:
199 ctx->params->internalCsp = X265_CSP_I444;
201 case AV_PIX_FMT_GRAY8:
202 case AV_PIX_FMT_GRAY10:
203 case AV_PIX_FMT_GRAY12:
204 if (ctx->api->api_build_number < 85) {
205 av_log(avctx, AV_LOG_ERROR,
206 "libx265 version is %d, must be at least 85 for gray encoding.\n",
207 ctx->api->api_build_number);
208 return AVERROR_INVALIDDATA;
210 ctx->params->internalCsp = X265_CSP_I400;
217 snprintf(crf, sizeof(crf), "%2.2f", ctx->crf);
218 if (ctx->api->param_parse(ctx->params, "crf", crf) == X265_PARAM_BAD_VALUE) {
219 av_log(avctx, AV_LOG_ERROR, "Invalid crf: %2.2f.\n", ctx->crf);
220 return AVERROR(EINVAL);
222 } else if (avctx->bit_rate > 0) {
223 ctx->params->rc.bitrate = avctx->bit_rate / 1000;
224 ctx->params->rc.rateControlMode = X265_RC_ABR;
227 ctx->params->rc.vbvBufferSize = avctx->rc_buffer_size / 1000;
228 ctx->params->rc.vbvMaxBitrate = avctx->rc_max_rate / 1000;
230 cpb_props = ff_add_cpb_side_data(avctx);
232 return AVERROR(ENOMEM);
233 cpb_props->buffer_size = ctx->params->rc.vbvBufferSize * 1000;
234 cpb_props->max_bitrate = ctx->params->rc.vbvMaxBitrate * 1000;
235 cpb_props->avg_bitrate = ctx->params->rc.bitrate * 1000;
237 if (!(avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER))
238 ctx->params->bRepeatHeaders = 1;
240 if (ctx->x265_opts) {
241 AVDictionary *dict = NULL;
242 AVDictionaryEntry *en = NULL;
244 if (!av_dict_parse_string(&dict, ctx->x265_opts, "=", ":", 0)) {
245 while ((en = av_dict_get(dict, "", en, AV_DICT_IGNORE_SUFFIX))) {
246 int parse_ret = ctx->api->param_parse(ctx->params, en->key, en->value);
249 case X265_PARAM_BAD_NAME:
250 av_log(avctx, AV_LOG_WARNING,
251 "Unknown option: %s.\n", en->key);
253 case X265_PARAM_BAD_VALUE:
254 av_log(avctx, AV_LOG_WARNING,
255 "Invalid value for %s: %s.\n", en->key, en->value);
265 if (ctx->params->rc.vbvBufferSize && avctx->rc_initial_buffer_occupancy > 1000 &&
266 ctx->params->rc.vbvBufferInit == 0.9) {
267 ctx->params->rc.vbvBufferInit = (float)avctx->rc_initial_buffer_occupancy / 1000;
271 if (ctx->api->param_apply_profile(ctx->params, ctx->profile) < 0) {
273 av_log(avctx, AV_LOG_ERROR, "Invalid or incompatible profile set: %s.\n", ctx->profile);
274 av_log(avctx, AV_LOG_INFO, "Possible profiles:");
275 for (i = 0; x265_profile_names[i]; i++)
276 av_log(avctx, AV_LOG_INFO, " %s", x265_profile_names[i]);
277 av_log(avctx, AV_LOG_INFO, "\n");
278 return AVERROR(EINVAL);
282 ctx->encoder = ctx->api->encoder_open(ctx->params);
284 av_log(avctx, AV_LOG_ERROR, "Cannot open libx265 encoder.\n");
285 libx265_encode_close(avctx);
286 return AVERROR_INVALIDDATA;
289 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
293 avctx->extradata_size = ctx->api->encoder_headers(ctx->encoder, &nal, &nnal);
294 if (avctx->extradata_size <= 0) {
295 av_log(avctx, AV_LOG_ERROR, "Cannot encode headers.\n");
296 libx265_encode_close(avctx);
297 return AVERROR_INVALIDDATA;
300 avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
301 if (!avctx->extradata) {
302 av_log(avctx, AV_LOG_ERROR,
303 "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size);
304 libx265_encode_close(avctx);
305 return AVERROR(ENOMEM);
308 memcpy(avctx->extradata, nal[0].payload, avctx->extradata_size);
314 static av_cold int libx265_encode_set_roi(libx265Context *ctx, const AVFrame *frame, x265_picture* pic)
316 AVFrameSideData *sd = av_frame_get_side_data(frame, AV_FRAME_DATA_REGIONS_OF_INTEREST);
318 if (ctx->params->rc.aqMode == X265_AQ_NONE) {
319 if (!ctx->roi_warned) {
321 av_log(ctx, AV_LOG_WARNING, "Adaptive quantization must be enabled to use ROI encoding, skipping ROI.\n");
324 /* 8x8 block when qg-size is 8, 16*16 block otherwise. */
325 int mb_size = (ctx->params->rc.qgSize == 8) ? 8 : 16;
326 int mbx = (frame->width + mb_size - 1) / mb_size;
327 int mby = (frame->height + mb_size - 1) / mb_size;
328 int qp_range = 51 + 6 * (pic->bitDepth - 8);
330 const AVRegionOfInterest *roi;
332 float *qoffsets; /* will be freed after encode is called. */
334 roi = (const AVRegionOfInterest*)sd->data;
335 roi_size = roi->self_size;
336 if (!roi_size || sd->size % roi_size != 0) {
337 av_log(ctx, AV_LOG_ERROR, "Invalid AVRegionOfInterest.self_size.\n");
338 return AVERROR(EINVAL);
340 nb_rois = sd->size / roi_size;
342 qoffsets = av_mallocz_array(mbx * mby, sizeof(*qoffsets));
344 return AVERROR(ENOMEM);
346 // This list must be iterated in reverse because the first
347 // region in the list applies when regions overlap.
348 for (int i = nb_rois - 1; i >= 0; i--) {
349 int startx, endx, starty, endy;
352 roi = (const AVRegionOfInterest*)(sd->data + roi_size * i);
354 starty = FFMIN(mby, roi->top / mb_size);
355 endy = FFMIN(mby, (roi->bottom + mb_size - 1)/ mb_size);
356 startx = FFMIN(mbx, roi->left / mb_size);
357 endx = FFMIN(mbx, (roi->right + mb_size - 1)/ mb_size);
359 if (roi->qoffset.den == 0) {
361 av_log(ctx, AV_LOG_ERROR, "AVRegionOfInterest.qoffset.den must not be zero.\n");
362 return AVERROR(EINVAL);
364 qoffset = roi->qoffset.num * 1.0f / roi->qoffset.den;
365 qoffset = av_clipf(qoffset * qp_range, -qp_range, +qp_range);
367 for (int y = starty; y < endy; y++)
368 for (int x = startx; x < endx; x++)
369 qoffsets[x + y*mbx] = qoffset;
372 pic->quantOffsets = qoffsets;
378 static int libx265_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
379 const AVFrame *pic, int *got_packet)
381 libx265Context *ctx = avctx->priv_data;
382 x265_picture x265pic;
383 x265_picture x265pic_out = { 0 };
391 ctx->api->picture_init(ctx->params, &x265pic);
394 for (i = 0; i < 3; i++) {
395 x265pic.planes[i] = pic->data[i];
396 x265pic.stride[i] = pic->linesize[i];
399 x265pic.pts = pic->pts;
400 x265pic.bitDepth = av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth;
402 x265pic.sliceType = pic->pict_type == AV_PICTURE_TYPE_I ?
403 (ctx->forced_idr ? X265_TYPE_IDR : X265_TYPE_I) :
404 pic->pict_type == AV_PICTURE_TYPE_P ? X265_TYPE_P :
405 pic->pict_type == AV_PICTURE_TYPE_B ? X265_TYPE_B :
408 ret = libx265_encode_set_roi(ctx, pic, &x265pic);
413 ret = ctx->api->encoder_encode(ctx->encoder, &nal, &nnal,
414 pic ? &x265pic : NULL, &x265pic_out);
416 av_freep(&x265pic.quantOffsets);
419 return AVERROR_EXTERNAL;
424 for (i = 0; i < nnal; i++)
425 payload += nal[i].sizeBytes;
427 ret = ff_alloc_packet2(avctx, pkt, payload, payload);
429 av_log(avctx, AV_LOG_ERROR, "Error getting output packet.\n");
434 for (i = 0; i < nnal; i++) {
435 memcpy(dst, nal[i].payload, nal[i].sizeBytes);
436 dst += nal[i].sizeBytes;
438 if (is_keyframe(nal[i].type))
439 pkt->flags |= AV_PKT_FLAG_KEY;
442 pkt->pts = x265pic_out.pts;
443 pkt->dts = x265pic_out.dts;
445 #if FF_API_CODED_FRAME
446 FF_DISABLE_DEPRECATION_WARNINGS
447 switch (x265pic_out.sliceType) {
450 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
453 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
456 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_B;
459 FF_ENABLE_DEPRECATION_WARNINGS
462 #if X265_BUILD >= 130
463 if (x265pic_out.sliceType == X265_TYPE_B)
465 if (x265pic_out.frameData.sliceType == 'b')
467 pkt->flags |= AV_PKT_FLAG_DISPOSABLE;
473 static const enum AVPixelFormat x265_csp_eight[] = {
485 static const enum AVPixelFormat x265_csp_ten[] = {
493 AV_PIX_FMT_YUV420P10,
494 AV_PIX_FMT_YUV422P10,
495 AV_PIX_FMT_YUV444P10,
502 static const enum AVPixelFormat x265_csp_twelve[] = {
510 AV_PIX_FMT_YUV420P10,
511 AV_PIX_FMT_YUV422P10,
512 AV_PIX_FMT_YUV444P10,
514 AV_PIX_FMT_YUV420P12,
515 AV_PIX_FMT_YUV422P12,
516 AV_PIX_FMT_YUV444P12,
524 static av_cold void libx265_encode_init_csp(AVCodec *codec)
526 if (x265_api_get(12))
527 codec->pix_fmts = x265_csp_twelve;
528 else if (x265_api_get(10))
529 codec->pix_fmts = x265_csp_ten;
530 else if (x265_api_get(8))
531 codec->pix_fmts = x265_csp_eight;
534 #define OFFSET(x) offsetof(libx265Context, x)
535 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
536 static const AVOption options[] = {
537 { "crf", "set the x265 crf", OFFSET(crf), AV_OPT_TYPE_FLOAT, { .dbl = -1 }, -1, FLT_MAX, VE },
538 { "forced-idr", "if forcing keyframes, force them as IDR frames", OFFSET(forced_idr),AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
539 { "preset", "set the x265 preset", OFFSET(preset), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
540 { "tune", "set the x265 tune parameter", OFFSET(tune), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
541 { "profile", "set the x265 profile", OFFSET(profile), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
542 { "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 },
546 static const AVClass class = {
547 .class_name = "libx265",
548 .item_name = av_default_item_name,
550 .version = LIBAVUTIL_VERSION_INT,
553 static const AVCodecDefault x265_defaults[] = {
558 AVCodec ff_libx265_encoder = {
560 .long_name = NULL_IF_CONFIG_SMALL("libx265 H.265 / HEVC"),
561 .type = AVMEDIA_TYPE_VIDEO,
562 .id = AV_CODEC_ID_HEVC,
563 .init = libx265_encode_init,
564 .init_static_data = libx265_encode_init_csp,
565 .encode2 = libx265_encode_frame,
566 .close = libx265_encode_close,
567 .priv_data_size = sizeof(libx265Context),
568 .priv_class = &class,
569 .defaults = x265_defaults,
570 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
571 .wrapper_name = "libx265",