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;
83 ctx->api = x265_api_get(av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth);
85 ctx->api = x265_api_get(0);
87 ctx->params = ctx->api->param_alloc();
89 av_log(avctx, AV_LOG_ERROR, "Could not allocate x265 param structure.\n");
90 return AVERROR(ENOMEM);
93 if (ctx->api->param_default_preset(ctx->params, ctx->preset, ctx->tune) < 0) {
96 av_log(avctx, AV_LOG_ERROR, "Error setting preset/tune %s/%s.\n", ctx->preset, ctx->tune);
97 av_log(avctx, AV_LOG_INFO, "Possible presets:");
98 for (i = 0; x265_preset_names[i]; i++)
99 av_log(avctx, AV_LOG_INFO, " %s", x265_preset_names[i]);
101 av_log(avctx, AV_LOG_INFO, "\n");
102 av_log(avctx, AV_LOG_INFO, "Possible tunes:");
103 for (i = 0; x265_tune_names[i]; i++)
104 av_log(avctx, AV_LOG_INFO, " %s", x265_tune_names[i]);
106 av_log(avctx, AV_LOG_INFO, "\n");
108 return AVERROR(EINVAL);
111 ctx->params->frameNumThreads = avctx->thread_count;
112 ctx->params->fpsNum = avctx->time_base.den;
113 ctx->params->fpsDenom = avctx->time_base.num * avctx->ticks_per_frame;
114 ctx->params->sourceWidth = avctx->width;
115 ctx->params->sourceHeight = avctx->height;
116 ctx->params->bEnablePsnr = !!(avctx->flags & AV_CODEC_FLAG_PSNR);
117 ctx->params->bOpenGOP = !(avctx->flags & AV_CODEC_FLAG_CLOSED_GOP);
119 /* Tune the CTU size based on input resolution. */
120 if (ctx->params->sourceWidth < 64 || ctx->params->sourceHeight < 64)
121 ctx->params->maxCUSize = 32;
122 if (ctx->params->sourceWidth < 32 || ctx->params->sourceHeight < 32)
123 ctx->params->maxCUSize = 16;
124 if (ctx->params->sourceWidth < 16 || ctx->params->sourceHeight < 16) {
125 av_log(avctx, AV_LOG_ERROR, "Image size is too small (%dx%d).\n",
126 ctx->params->sourceWidth, ctx->params->sourceHeight);
127 return AVERROR(EINVAL);
130 if ((avctx->color_primaries <= AVCOL_PRI_SMPTE432 &&
131 avctx->color_primaries != AVCOL_PRI_UNSPECIFIED) ||
132 (avctx->color_trc <= AVCOL_TRC_ARIB_STD_B67 &&
133 avctx->color_trc != AVCOL_TRC_UNSPECIFIED) ||
134 (avctx->colorspace <= AVCOL_SPC_ICTCP &&
135 avctx->colorspace != AVCOL_SPC_UNSPECIFIED)) {
137 ctx->params->vui.bEnableVideoSignalTypePresentFlag = 1;
138 ctx->params->vui.bEnableColorDescriptionPresentFlag = 1;
140 // x265 validates the parameters internally
141 ctx->params->vui.colorPrimaries = avctx->color_primaries;
142 ctx->params->vui.transferCharacteristics = avctx->color_trc;
143 ctx->params->vui.matrixCoeffs = avctx->colorspace;
146 if (avctx->sample_aspect_ratio.num > 0 && avctx->sample_aspect_ratio.den > 0) {
148 int sar_num, sar_den;
150 av_reduce(&sar_num, &sar_den,
151 avctx->sample_aspect_ratio.num,
152 avctx->sample_aspect_ratio.den, 65535);
153 snprintf(sar, sizeof(sar), "%d:%d", sar_num, sar_den);
154 if (ctx->api->param_parse(ctx->params, "sar", sar) == X265_PARAM_BAD_VALUE) {
155 av_log(avctx, AV_LOG_ERROR, "Invalid SAR: %d:%d.\n", sar_num, sar_den);
156 return AVERROR_INVALIDDATA;
160 switch (avctx->pix_fmt) {
161 case AV_PIX_FMT_YUV420P:
162 case AV_PIX_FMT_YUV420P10:
163 case AV_PIX_FMT_YUV420P12:
164 ctx->params->internalCsp = X265_CSP_I420;
166 case AV_PIX_FMT_YUV422P:
167 case AV_PIX_FMT_YUV422P10:
168 case AV_PIX_FMT_YUV422P12:
169 ctx->params->internalCsp = X265_CSP_I422;
171 case AV_PIX_FMT_GBRP:
172 case AV_PIX_FMT_GBRP10:
173 case AV_PIX_FMT_GBRP12:
174 ctx->params->vui.matrixCoeffs = AVCOL_SPC_RGB;
175 ctx->params->vui.bEnableVideoSignalTypePresentFlag = 1;
176 ctx->params->vui.bEnableColorDescriptionPresentFlag = 1;
177 case AV_PIX_FMT_YUV444P:
178 case AV_PIX_FMT_YUV444P10:
179 case AV_PIX_FMT_YUV444P12:
180 ctx->params->internalCsp = X265_CSP_I444;
182 case AV_PIX_FMT_GRAY8:
183 case AV_PIX_FMT_GRAY10:
184 case AV_PIX_FMT_GRAY12:
185 if (ctx->api->api_build_number < 85) {
186 av_log(avctx, AV_LOG_ERROR,
187 "libx265 version is %d, must be at least 85 for gray encoding.\n",
188 ctx->api->api_build_number);
189 return AVERROR_INVALIDDATA;
191 ctx->params->internalCsp = X265_CSP_I400;
198 snprintf(crf, sizeof(crf), "%2.2f", ctx->crf);
199 if (ctx->api->param_parse(ctx->params, "crf", crf) == X265_PARAM_BAD_VALUE) {
200 av_log(avctx, AV_LOG_ERROR, "Invalid crf: %2.2f.\n", ctx->crf);
201 return AVERROR(EINVAL);
203 } else if (avctx->bit_rate > 0) {
204 ctx->params->rc.bitrate = avctx->bit_rate / 1000;
205 ctx->params->rc.rateControlMode = X265_RC_ABR;
208 ctx->params->rc.vbvBufferSize = avctx->rc_buffer_size / 1000;
209 ctx->params->rc.vbvMaxBitrate = avctx->rc_max_rate / 1000;
211 if (!(avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER))
212 ctx->params->bRepeatHeaders = 1;
214 if (ctx->x265_opts) {
215 AVDictionary *dict = NULL;
216 AVDictionaryEntry *en = NULL;
218 if (!av_dict_parse_string(&dict, ctx->x265_opts, "=", ":", 0)) {
219 while ((en = av_dict_get(dict, "", en, AV_DICT_IGNORE_SUFFIX))) {
220 int parse_ret = ctx->api->param_parse(ctx->params, en->key, en->value);
223 case X265_PARAM_BAD_NAME:
224 av_log(avctx, AV_LOG_WARNING,
225 "Unknown option: %s.\n", en->key);
227 case X265_PARAM_BAD_VALUE:
228 av_log(avctx, AV_LOG_WARNING,
229 "Invalid value for %s: %s.\n", en->key, en->value);
239 if (ctx->params->rc.vbvBufferSize && avctx->rc_initial_buffer_occupancy > 1000 &&
240 ctx->params->rc.vbvBufferInit == 0.9) {
241 ctx->params->rc.vbvBufferInit = (float)avctx->rc_initial_buffer_occupancy / 1000;
245 if (ctx->api->param_apply_profile(ctx->params, ctx->profile) < 0) {
247 av_log(avctx, AV_LOG_ERROR, "Invalid or incompatible profile set: %s.\n", ctx->profile);
248 av_log(avctx, AV_LOG_INFO, "Possible profiles:");
249 for (i = 0; x265_profile_names[i]; i++)
250 av_log(avctx, AV_LOG_INFO, " %s", x265_profile_names[i]);
251 av_log(avctx, AV_LOG_INFO, "\n");
252 return AVERROR(EINVAL);
256 ctx->encoder = ctx->api->encoder_open(ctx->params);
258 av_log(avctx, AV_LOG_ERROR, "Cannot open libx265 encoder.\n");
259 libx265_encode_close(avctx);
260 return AVERROR_INVALIDDATA;
263 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
267 avctx->extradata_size = ctx->api->encoder_headers(ctx->encoder, &nal, &nnal);
268 if (avctx->extradata_size <= 0) {
269 av_log(avctx, AV_LOG_ERROR, "Cannot encode headers.\n");
270 libx265_encode_close(avctx);
271 return AVERROR_INVALIDDATA;
274 avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
275 if (!avctx->extradata) {
276 av_log(avctx, AV_LOG_ERROR,
277 "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size);
278 libx265_encode_close(avctx);
279 return AVERROR(ENOMEM);
282 memcpy(avctx->extradata, nal[0].payload, avctx->extradata_size);
288 static av_cold int libx265_encode_set_roi(libx265Context *ctx, const AVFrame *frame, x265_picture* pic)
290 AVFrameSideData *sd = av_frame_get_side_data(frame, AV_FRAME_DATA_REGIONS_OF_INTEREST);
292 if (ctx->params->rc.aqMode == X265_AQ_NONE) {
293 av_log(ctx, AV_LOG_WARNING, "Adaptive quantization must be enabled to use ROI encoding, skipping ROI.\n");
295 /* 8x8 block when qg-size is 8, 16*16 block otherwise. */
296 int mb_size = (ctx->params->rc.qgSize == 8) ? 8 : 16;
297 int mbx = (frame->width + mb_size - 1) / mb_size;
298 int mby = (frame->height + mb_size - 1) / mb_size;
300 AVRegionOfInterest *roi;
301 float *qoffsets; /* will be freed after encode is called. */
302 qoffsets = av_mallocz_array(mbx * mby, sizeof(*qoffsets));
304 return AVERROR(ENOMEM);
306 nb_rois = sd->size / sizeof(AVRegionOfInterest);
307 roi = (AVRegionOfInterest*)sd->data;
308 for (int count = 0; count < nb_rois; count++) {
309 int starty = FFMIN(mby, roi->top / mb_size);
310 int endy = FFMIN(mby, (roi->bottom + mb_size - 1)/ mb_size);
311 int startx = FFMIN(mbx, roi->left / mb_size);
312 int endx = FFMIN(mbx, (roi->right + mb_size - 1)/ mb_size);
315 if (roi->self_size == 0) {
317 av_log(ctx, AV_LOG_ERROR, "AVRegionOfInterest.self_size must be set to sizeof(AVRegionOfInterest).\n");
318 return AVERROR(EINVAL);
321 if (roi->qoffset.den == 0) {
323 av_log(ctx, AV_LOG_ERROR, "AVRegionOfInterest.qoffset.den must not be zero.\n");
324 return AVERROR(EINVAL);
326 qoffset = roi->qoffset.num * 1.0f / roi->qoffset.den;
327 qoffset = av_clipf(qoffset, -1.0f, 1.0f);
329 /* qp range of x265 is from 0 to 51, just choose 25 as the scale value,
330 * so the range of final qoffset is [-25.0, 25.0].
332 qoffset = qoffset * 25;
334 for (int y = starty; y < endy; y++)
335 for (int x = startx; x < endx; x++)
336 qoffsets[x + y*mbx] = qoffset;
338 roi = (AVRegionOfInterest*)((char*)roi + roi->self_size);
341 pic->quantOffsets = qoffsets;
347 static int libx265_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
348 const AVFrame *pic, int *got_packet)
350 libx265Context *ctx = avctx->priv_data;
351 x265_picture x265pic;
352 x265_picture x265pic_out = { 0 };
360 ctx->api->picture_init(ctx->params, &x265pic);
363 for (i = 0; i < 3; i++) {
364 x265pic.planes[i] = pic->data[i];
365 x265pic.stride[i] = pic->linesize[i];
368 x265pic.pts = pic->pts;
369 x265pic.bitDepth = av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth;
371 x265pic.sliceType = pic->pict_type == AV_PICTURE_TYPE_I ?
372 (ctx->forced_idr ? X265_TYPE_IDR : X265_TYPE_I) :
373 pic->pict_type == AV_PICTURE_TYPE_P ? X265_TYPE_P :
374 pic->pict_type == AV_PICTURE_TYPE_B ? X265_TYPE_B :
377 ret = libx265_encode_set_roi(ctx, pic, &x265pic);
382 ret = ctx->api->encoder_encode(ctx->encoder, &nal, &nnal,
383 pic ? &x265pic : NULL, &x265pic_out);
385 av_freep(&x265pic.quantOffsets);
388 return AVERROR_EXTERNAL;
393 for (i = 0; i < nnal; i++)
394 payload += nal[i].sizeBytes;
396 ret = ff_alloc_packet2(avctx, pkt, payload, payload);
398 av_log(avctx, AV_LOG_ERROR, "Error getting output packet.\n");
403 for (i = 0; i < nnal; i++) {
404 memcpy(dst, nal[i].payload, nal[i].sizeBytes);
405 dst += nal[i].sizeBytes;
407 if (is_keyframe(nal[i].type))
408 pkt->flags |= AV_PKT_FLAG_KEY;
411 pkt->pts = x265pic_out.pts;
412 pkt->dts = x265pic_out.dts;
414 #if FF_API_CODED_FRAME
415 FF_DISABLE_DEPRECATION_WARNINGS
416 switch (x265pic_out.sliceType) {
419 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
422 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
425 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_B;
428 FF_ENABLE_DEPRECATION_WARNINGS
431 #if X265_BUILD >= 130
432 if (x265pic_out.sliceType == X265_TYPE_B)
434 if (x265pic_out.frameData.sliceType == 'b')
436 pkt->flags |= AV_PKT_FLAG_DISPOSABLE;
442 static const enum AVPixelFormat x265_csp_eight[] = {
451 static const enum AVPixelFormat x265_csp_ten[] = {
456 AV_PIX_FMT_YUV420P10,
457 AV_PIX_FMT_YUV422P10,
458 AV_PIX_FMT_YUV444P10,
465 static const enum AVPixelFormat x265_csp_twelve[] = {
470 AV_PIX_FMT_YUV420P10,
471 AV_PIX_FMT_YUV422P10,
472 AV_PIX_FMT_YUV444P10,
474 AV_PIX_FMT_YUV420P12,
475 AV_PIX_FMT_YUV422P12,
476 AV_PIX_FMT_YUV444P12,
484 static av_cold void libx265_encode_init_csp(AVCodec *codec)
486 if (x265_api_get(12))
487 codec->pix_fmts = x265_csp_twelve;
488 else if (x265_api_get(10))
489 codec->pix_fmts = x265_csp_ten;
490 else if (x265_api_get(8))
491 codec->pix_fmts = x265_csp_eight;
494 #define OFFSET(x) offsetof(libx265Context, x)
495 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
496 static const AVOption options[] = {
497 { "crf", "set the x265 crf", OFFSET(crf), AV_OPT_TYPE_FLOAT, { .dbl = -1 }, -1, FLT_MAX, VE },
498 { "forced-idr", "if forcing keyframes, force them as IDR frames", OFFSET(forced_idr),AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
499 { "preset", "set the x265 preset", OFFSET(preset), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
500 { "tune", "set the x265 tune parameter", OFFSET(tune), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
501 { "profile", "set the x265 profile", OFFSET(profile), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
502 { "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 },
506 static const AVClass class = {
507 .class_name = "libx265",
508 .item_name = av_default_item_name,
510 .version = LIBAVUTIL_VERSION_INT,
513 static const AVCodecDefault x265_defaults[] = {
518 AVCodec ff_libx265_encoder = {
520 .long_name = NULL_IF_CONFIG_SMALL("libx265 H.265 / HEVC"),
521 .type = AVMEDIA_TYPE_VIDEO,
522 .id = AV_CODEC_ID_HEVC,
523 .init = libx265_encode_init,
524 .init_static_data = libx265_encode_init_csp,
525 .encode2 = libx265_encode_frame,
526 .close = libx265_encode_close,
527 .priv_data_size = sizeof(libx265Context),
528 .priv_class = &class,
529 .defaults = x265_defaults,
530 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
531 .wrapper_name = "libx265",