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;
53 * If the encoder does not support ROI then warn the first time we
54 * encounter a frame with ROI side data.
59 static int is_keyframe(NalUnitType naltype)
62 case NAL_UNIT_CODED_SLICE_BLA_W_LP:
63 case NAL_UNIT_CODED_SLICE_BLA_W_RADL:
64 case NAL_UNIT_CODED_SLICE_BLA_N_LP:
65 case NAL_UNIT_CODED_SLICE_IDR_W_RADL:
66 case NAL_UNIT_CODED_SLICE_IDR_N_LP:
67 case NAL_UNIT_CODED_SLICE_CRA:
74 static av_cold int libx265_encode_close(AVCodecContext *avctx)
76 libx265Context *ctx = avctx->priv_data;
78 ctx->api->param_free(ctx->params);
81 ctx->api->encoder_close(ctx->encoder);
86 static av_cold int libx265_param_parse_float(AVCodecContext *avctx,
87 const char *key, float value)
89 libx265Context *ctx = avctx->priv_data;
92 snprintf(buf, sizeof(buf), "%2.2f", value);
93 if (ctx->api->param_parse(ctx->params, key, buf) == X265_PARAM_BAD_VALUE) {
94 av_log(avctx, AV_LOG_ERROR, "Invalid value %2.2f for param \"%s\".\n", value, key);
95 return AVERROR(EINVAL);
101 static av_cold int libx265_param_parse_int(AVCodecContext *avctx,
102 const char *key, int value)
104 libx265Context *ctx = avctx->priv_data;
107 snprintf(buf, sizeof(buf), "%d", value);
108 if (ctx->api->param_parse(ctx->params, key, buf) == X265_PARAM_BAD_VALUE) {
109 av_log(avctx, AV_LOG_ERROR, "Invalid value %d for param \"%s\".\n", value, key);
110 return AVERROR(EINVAL);
116 static av_cold int libx265_encode_init(AVCodecContext *avctx)
118 libx265Context *ctx = avctx->priv_data;
119 AVCPBProperties *cpb_props = NULL;
122 ctx->api = x265_api_get(av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth);
124 ctx->api = x265_api_get(0);
126 ctx->params = ctx->api->param_alloc();
128 av_log(avctx, AV_LOG_ERROR, "Could not allocate x265 param structure.\n");
129 return AVERROR(ENOMEM);
132 if (ctx->api->param_default_preset(ctx->params, ctx->preset, ctx->tune) < 0) {
135 av_log(avctx, AV_LOG_ERROR, "Error setting preset/tune %s/%s.\n", ctx->preset, ctx->tune);
136 av_log(avctx, AV_LOG_INFO, "Possible presets:");
137 for (i = 0; x265_preset_names[i]; i++)
138 av_log(avctx, AV_LOG_INFO, " %s", x265_preset_names[i]);
140 av_log(avctx, AV_LOG_INFO, "\n");
141 av_log(avctx, AV_LOG_INFO, "Possible tunes:");
142 for (i = 0; x265_tune_names[i]; i++)
143 av_log(avctx, AV_LOG_INFO, " %s", x265_tune_names[i]);
145 av_log(avctx, AV_LOG_INFO, "\n");
147 return AVERROR(EINVAL);
150 ctx->params->frameNumThreads = avctx->thread_count;
151 if (avctx->framerate.num > 0 && avctx->framerate.den > 0) {
152 ctx->params->fpsNum = avctx->framerate.num;
153 ctx->params->fpsDenom = avctx->framerate.den;
155 ctx->params->fpsNum = avctx->time_base.den;
156 ctx->params->fpsDenom = avctx->time_base.num * avctx->ticks_per_frame;
158 ctx->params->sourceWidth = avctx->width;
159 ctx->params->sourceHeight = avctx->height;
160 ctx->params->bEnablePsnr = !!(avctx->flags & AV_CODEC_FLAG_PSNR);
161 ctx->params->bOpenGOP = !(avctx->flags & AV_CODEC_FLAG_CLOSED_GOP);
163 /* Tune the CTU size based on input resolution. */
164 if (ctx->params->sourceWidth < 64 || ctx->params->sourceHeight < 64)
165 ctx->params->maxCUSize = 32;
166 if (ctx->params->sourceWidth < 32 || ctx->params->sourceHeight < 32)
167 ctx->params->maxCUSize = 16;
168 if (ctx->params->sourceWidth < 16 || ctx->params->sourceHeight < 16) {
169 av_log(avctx, AV_LOG_ERROR, "Image size is too small (%dx%d).\n",
170 ctx->params->sourceWidth, ctx->params->sourceHeight);
171 return AVERROR(EINVAL);
175 ctx->params->vui.bEnableVideoSignalTypePresentFlag = 1;
177 ctx->params->vui.bEnableVideoFullRangeFlag = avctx->pix_fmt == AV_PIX_FMT_YUVJ420P ||
178 avctx->pix_fmt == AV_PIX_FMT_YUVJ422P ||
179 avctx->pix_fmt == AV_PIX_FMT_YUVJ444P ||
180 avctx->color_range == AVCOL_RANGE_JPEG;
182 if ((avctx->color_primaries <= AVCOL_PRI_SMPTE432 &&
183 avctx->color_primaries != AVCOL_PRI_UNSPECIFIED) ||
184 (avctx->color_trc <= AVCOL_TRC_ARIB_STD_B67 &&
185 avctx->color_trc != AVCOL_TRC_UNSPECIFIED) ||
186 (avctx->colorspace <= AVCOL_SPC_ICTCP &&
187 avctx->colorspace != AVCOL_SPC_UNSPECIFIED)) {
189 ctx->params->vui.bEnableColorDescriptionPresentFlag = 1;
191 // x265 validates the parameters internally
192 ctx->params->vui.colorPrimaries = avctx->color_primaries;
193 ctx->params->vui.transferCharacteristics = avctx->color_trc;
194 #if X265_BUILD >= 159
195 if (avctx->color_trc == AVCOL_TRC_ARIB_STD_B67)
196 ctx->params->preferredTransferCharacteristics = ctx->params->vui.transferCharacteristics;
198 ctx->params->vui.matrixCoeffs = avctx->colorspace;
201 if (avctx->sample_aspect_ratio.num > 0 && avctx->sample_aspect_ratio.den > 0) {
203 int sar_num, sar_den;
205 av_reduce(&sar_num, &sar_den,
206 avctx->sample_aspect_ratio.num,
207 avctx->sample_aspect_ratio.den, 65535);
208 snprintf(sar, sizeof(sar), "%d:%d", sar_num, sar_den);
209 if (ctx->api->param_parse(ctx->params, "sar", sar) == X265_PARAM_BAD_VALUE) {
210 av_log(avctx, AV_LOG_ERROR, "Invalid SAR: %d:%d.\n", sar_num, sar_den);
211 return AVERROR_INVALIDDATA;
215 switch (avctx->pix_fmt) {
216 case AV_PIX_FMT_YUV420P:
217 case AV_PIX_FMT_YUV420P10:
218 case AV_PIX_FMT_YUV420P12:
219 ctx->params->internalCsp = X265_CSP_I420;
221 case AV_PIX_FMT_YUV422P:
222 case AV_PIX_FMT_YUV422P10:
223 case AV_PIX_FMT_YUV422P12:
224 ctx->params->internalCsp = X265_CSP_I422;
226 case AV_PIX_FMT_GBRP:
227 case AV_PIX_FMT_GBRP10:
228 case AV_PIX_FMT_GBRP12:
229 ctx->params->vui.matrixCoeffs = AVCOL_SPC_RGB;
230 ctx->params->vui.bEnableVideoSignalTypePresentFlag = 1;
231 ctx->params->vui.bEnableColorDescriptionPresentFlag = 1;
232 case AV_PIX_FMT_YUV444P:
233 case AV_PIX_FMT_YUV444P10:
234 case AV_PIX_FMT_YUV444P12:
235 ctx->params->internalCsp = X265_CSP_I444;
237 case AV_PIX_FMT_GRAY8:
238 case AV_PIX_FMT_GRAY10:
239 case AV_PIX_FMT_GRAY12:
240 if (ctx->api->api_build_number < 85) {
241 av_log(avctx, AV_LOG_ERROR,
242 "libx265 version is %d, must be at least 85 for gray encoding.\n",
243 ctx->api->api_build_number);
244 return AVERROR_INVALIDDATA;
246 ctx->params->internalCsp = X265_CSP_I400;
253 snprintf(crf, sizeof(crf), "%2.2f", ctx->crf);
254 if (ctx->api->param_parse(ctx->params, "crf", crf) == X265_PARAM_BAD_VALUE) {
255 av_log(avctx, AV_LOG_ERROR, "Invalid crf: %2.2f.\n", ctx->crf);
256 return AVERROR(EINVAL);
258 } else if (avctx->bit_rate > 0) {
259 ctx->params->rc.bitrate = avctx->bit_rate / 1000;
260 ctx->params->rc.rateControlMode = X265_RC_ABR;
261 } else if (ctx->cqp >= 0) {
262 ret = libx265_param_parse_int(avctx, "qp", ctx->cqp);
268 if (avctx->qmin >= 0) {
269 ret = libx265_param_parse_int(avctx, "qpmin", avctx->qmin);
273 if (avctx->qmax >= 0) {
274 ret = libx265_param_parse_int(avctx, "qpmax", avctx->qmax);
279 if (avctx->max_qdiff >= 0) {
280 ret = libx265_param_parse_int(avctx, "qpstep", avctx->max_qdiff);
284 if (avctx->qblur >= 0) {
285 ret = libx265_param_parse_float(avctx, "qblur", avctx->qblur);
289 if (avctx->qcompress >= 0) {
290 ret = libx265_param_parse_float(avctx, "qcomp", avctx->qcompress);
294 if (avctx->i_quant_factor >= 0) {
295 ret = libx265_param_parse_float(avctx, "ipratio", avctx->i_quant_factor);
299 if (avctx->b_quant_factor >= 0) {
300 ret = libx265_param_parse_float(avctx, "pbratio", avctx->b_quant_factor);
305 ctx->params->rc.vbvBufferSize = avctx->rc_buffer_size / 1000;
306 ctx->params->rc.vbvMaxBitrate = avctx->rc_max_rate / 1000;
308 cpb_props = ff_add_cpb_side_data(avctx);
310 return AVERROR(ENOMEM);
311 cpb_props->buffer_size = ctx->params->rc.vbvBufferSize * 1000;
312 cpb_props->max_bitrate = ctx->params->rc.vbvMaxBitrate * 1000;
313 cpb_props->avg_bitrate = ctx->params->rc.bitrate * 1000;
315 if (!(avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER))
316 ctx->params->bRepeatHeaders = 1;
318 if (avctx->gop_size >= 0) {
319 ret = libx265_param_parse_int(avctx, "keyint", avctx->gop_size);
323 if (avctx->keyint_min > 0) {
324 ret = libx265_param_parse_int(avctx, "min-keyint", avctx->keyint_min);
328 if (avctx->max_b_frames >= 0) {
329 ret = libx265_param_parse_int(avctx, "bframes", avctx->max_b_frames);
333 if (avctx->refs >= 0) {
334 ret = libx265_param_parse_int(avctx, "ref", avctx->refs);
339 if (ctx->x265_opts) {
340 AVDictionary *dict = NULL;
341 AVDictionaryEntry *en = NULL;
343 if (!av_dict_parse_string(&dict, ctx->x265_opts, "=", ":", 0)) {
344 while ((en = av_dict_get(dict, "", en, AV_DICT_IGNORE_SUFFIX))) {
345 int parse_ret = ctx->api->param_parse(ctx->params, en->key, en->value);
348 case X265_PARAM_BAD_NAME:
349 av_log(avctx, AV_LOG_WARNING,
350 "Unknown option: %s.\n", en->key);
352 case X265_PARAM_BAD_VALUE:
353 av_log(avctx, AV_LOG_WARNING,
354 "Invalid value for %s: %s.\n", en->key, en->value);
364 if (ctx->params->rc.vbvBufferSize && avctx->rc_initial_buffer_occupancy > 1000 &&
365 ctx->params->rc.vbvBufferInit == 0.9) {
366 ctx->params->rc.vbvBufferInit = (float)avctx->rc_initial_buffer_occupancy / 1000;
370 if (ctx->api->param_apply_profile(ctx->params, ctx->profile) < 0) {
372 av_log(avctx, AV_LOG_ERROR, "Invalid or incompatible profile set: %s.\n", ctx->profile);
373 av_log(avctx, AV_LOG_INFO, "Possible profiles:");
374 for (i = 0; x265_profile_names[i]; i++)
375 av_log(avctx, AV_LOG_INFO, " %s", x265_profile_names[i]);
376 av_log(avctx, AV_LOG_INFO, "\n");
377 return AVERROR(EINVAL);
381 ctx->encoder = ctx->api->encoder_open(ctx->params);
383 av_log(avctx, AV_LOG_ERROR, "Cannot open libx265 encoder.\n");
384 libx265_encode_close(avctx);
385 return AVERROR_INVALIDDATA;
388 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
392 avctx->extradata_size = ctx->api->encoder_headers(ctx->encoder, &nal, &nnal);
393 if (avctx->extradata_size <= 0) {
394 av_log(avctx, AV_LOG_ERROR, "Cannot encode headers.\n");
395 libx265_encode_close(avctx);
396 return AVERROR_INVALIDDATA;
399 avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
400 if (!avctx->extradata) {
401 av_log(avctx, AV_LOG_ERROR,
402 "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size);
403 libx265_encode_close(avctx);
404 return AVERROR(ENOMEM);
407 memcpy(avctx->extradata, nal[0].payload, avctx->extradata_size);
413 static av_cold int libx265_encode_set_roi(libx265Context *ctx, const AVFrame *frame, x265_picture* pic)
415 AVFrameSideData *sd = av_frame_get_side_data(frame, AV_FRAME_DATA_REGIONS_OF_INTEREST);
417 if (ctx->params->rc.aqMode == X265_AQ_NONE) {
418 if (!ctx->roi_warned) {
420 av_log(ctx, AV_LOG_WARNING, "Adaptive quantization must be enabled to use ROI encoding, skipping ROI.\n");
423 /* 8x8 block when qg-size is 8, 16*16 block otherwise. */
424 int mb_size = (ctx->params->rc.qgSize == 8) ? 8 : 16;
425 int mbx = (frame->width + mb_size - 1) / mb_size;
426 int mby = (frame->height + mb_size - 1) / mb_size;
427 int qp_range = 51 + 6 * (pic->bitDepth - 8);
429 const AVRegionOfInterest *roi;
431 float *qoffsets; /* will be freed after encode is called. */
433 roi = (const AVRegionOfInterest*)sd->data;
434 roi_size = roi->self_size;
435 if (!roi_size || sd->size % roi_size != 0) {
436 av_log(ctx, AV_LOG_ERROR, "Invalid AVRegionOfInterest.self_size.\n");
437 return AVERROR(EINVAL);
439 nb_rois = sd->size / roi_size;
441 qoffsets = av_mallocz_array(mbx * mby, sizeof(*qoffsets));
443 return AVERROR(ENOMEM);
445 // This list must be iterated in reverse because the first
446 // region in the list applies when regions overlap.
447 for (int i = nb_rois - 1; i >= 0; i--) {
448 int startx, endx, starty, endy;
451 roi = (const AVRegionOfInterest*)(sd->data + roi_size * i);
453 starty = FFMIN(mby, roi->top / mb_size);
454 endy = FFMIN(mby, (roi->bottom + mb_size - 1)/ mb_size);
455 startx = FFMIN(mbx, roi->left / mb_size);
456 endx = FFMIN(mbx, (roi->right + mb_size - 1)/ mb_size);
458 if (roi->qoffset.den == 0) {
460 av_log(ctx, AV_LOG_ERROR, "AVRegionOfInterest.qoffset.den must not be zero.\n");
461 return AVERROR(EINVAL);
463 qoffset = roi->qoffset.num * 1.0f / roi->qoffset.den;
464 qoffset = av_clipf(qoffset * qp_range, -qp_range, +qp_range);
466 for (int y = starty; y < endy; y++)
467 for (int x = startx; x < endx; x++)
468 qoffsets[x + y*mbx] = qoffset;
471 pic->quantOffsets = qoffsets;
477 static int libx265_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
478 const AVFrame *pic, int *got_packet)
480 libx265Context *ctx = avctx->priv_data;
481 x265_picture x265pic;
482 x265_picture x265pic_out = { 0 };
490 ctx->api->picture_init(ctx->params, &x265pic);
493 for (i = 0; i < 3; i++) {
494 x265pic.planes[i] = pic->data[i];
495 x265pic.stride[i] = pic->linesize[i];
498 x265pic.pts = pic->pts;
499 x265pic.bitDepth = av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth;
501 x265pic.sliceType = pic->pict_type == AV_PICTURE_TYPE_I ?
502 (ctx->forced_idr ? X265_TYPE_IDR : X265_TYPE_I) :
503 pic->pict_type == AV_PICTURE_TYPE_P ? X265_TYPE_P :
504 pic->pict_type == AV_PICTURE_TYPE_B ? X265_TYPE_B :
507 ret = libx265_encode_set_roi(ctx, pic, &x265pic);
512 ret = ctx->api->encoder_encode(ctx->encoder, &nal, &nnal,
513 pic ? &x265pic : NULL, &x265pic_out);
515 av_freep(&x265pic.quantOffsets);
518 return AVERROR_EXTERNAL;
523 for (i = 0; i < nnal; i++)
524 payload += nal[i].sizeBytes;
526 ret = ff_alloc_packet2(avctx, pkt, payload, payload);
528 av_log(avctx, AV_LOG_ERROR, "Error getting output packet.\n");
533 for (i = 0; i < nnal; i++) {
534 memcpy(dst, nal[i].payload, nal[i].sizeBytes);
535 dst += nal[i].sizeBytes;
537 if (is_keyframe(nal[i].type))
538 pkt->flags |= AV_PKT_FLAG_KEY;
541 pkt->pts = x265pic_out.pts;
542 pkt->dts = x265pic_out.dts;
544 #if FF_API_CODED_FRAME
545 FF_DISABLE_DEPRECATION_WARNINGS
546 switch (x265pic_out.sliceType) {
549 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
552 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
555 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_B;
558 FF_ENABLE_DEPRECATION_WARNINGS
561 #if X265_BUILD >= 130
562 if (x265pic_out.sliceType == X265_TYPE_B)
564 if (x265pic_out.frameData.sliceType == 'b')
566 pkt->flags |= AV_PKT_FLAG_DISPOSABLE;
572 static const enum AVPixelFormat x265_csp_eight[] = {
584 static const enum AVPixelFormat x265_csp_ten[] = {
592 AV_PIX_FMT_YUV420P10,
593 AV_PIX_FMT_YUV422P10,
594 AV_PIX_FMT_YUV444P10,
601 static const enum AVPixelFormat x265_csp_twelve[] = {
609 AV_PIX_FMT_YUV420P10,
610 AV_PIX_FMT_YUV422P10,
611 AV_PIX_FMT_YUV444P10,
613 AV_PIX_FMT_YUV420P12,
614 AV_PIX_FMT_YUV422P12,
615 AV_PIX_FMT_YUV444P12,
623 static av_cold void libx265_encode_init_csp(AVCodec *codec)
625 if (x265_api_get(12))
626 codec->pix_fmts = x265_csp_twelve;
627 else if (x265_api_get(10))
628 codec->pix_fmts = x265_csp_ten;
629 else if (x265_api_get(8))
630 codec->pix_fmts = x265_csp_eight;
633 #define OFFSET(x) offsetof(libx265Context, x)
634 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
635 static const AVOption options[] = {
636 { "crf", "set the x265 crf", OFFSET(crf), AV_OPT_TYPE_FLOAT, { .dbl = -1 }, -1, FLT_MAX, VE },
637 { "qp", "set the x265 qp", OFFSET(cqp), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, INT_MAX, VE },
638 { "forced-idr", "if forcing keyframes, force them as IDR frames", OFFSET(forced_idr),AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
639 { "preset", "set the x265 preset", OFFSET(preset), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
640 { "tune", "set the x265 tune parameter", OFFSET(tune), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
641 { "profile", "set the x265 profile", OFFSET(profile), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
642 { "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 },
646 static const AVClass class = {
647 .class_name = "libx265",
648 .item_name = av_default_item_name,
650 .version = LIBAVUTIL_VERSION_INT,
653 static const AVCodecDefault x265_defaults[] = {
657 { "keyint_min", "-1" },
664 { "i_qfactor", "-1" },
665 { "b_qfactor", "-1" },
669 AVCodec ff_libx265_encoder = {
671 .long_name = NULL_IF_CONFIG_SMALL("libx265 H.265 / HEVC"),
672 .type = AVMEDIA_TYPE_VIDEO,
673 .id = AV_CODEC_ID_HEVC,
674 .init = libx265_encode_init,
675 .init_static_data = libx265_encode_init_csp,
676 .encode2 = libx265_encode_frame,
677 .close = libx265_encode_close,
678 .priv_data_size = sizeof(libx265Context),
679 .priv_class = &class,
680 .defaults = x265_defaults,
681 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
682 .wrapper_name = "libx265",