26 #define VPX_DISABLE_CTRL_TYPECHECKS 1
27 #define VPX_CODEC_DISABLE_COMPAT 1
28 #include <vpx/vpx_encoder.h>
29 #include <vpx/vp8cx.h>
85 #define VP8F_ERROR_RESILIENT 0x00000001
86 #define VP8F_AUTO_ALT_REF 0x00000002
131 [VP8E_SET_CPUUSED] =
"VP8E_SET_CPUUSED",
132 [VP8E_SET_ENABLEAUTOALTREF] =
"VP8E_SET_ENABLEAUTOALTREF",
133 [VP8E_SET_NOISE_SENSITIVITY] =
"VP8E_SET_NOISE_SENSITIVITY",
134 [VP8E_SET_STATIC_THRESHOLD] =
"VP8E_SET_STATIC_THRESHOLD",
135 [VP8E_SET_TOKEN_PARTITIONS] =
"VP8E_SET_TOKEN_PARTITIONS",
136 [VP8E_SET_ARNR_MAXFRAMES] =
"VP8E_SET_ARNR_MAXFRAMES",
137 [VP8E_SET_ARNR_STRENGTH] =
"VP8E_SET_ARNR_STRENGTH",
138 [VP8E_SET_ARNR_TYPE] =
"VP8E_SET_ARNR_TYPE",
139 [VP8E_SET_TUNING] =
"VP8E_SET_TUNING",
140 [VP8E_SET_CQ_LEVEL] =
"VP8E_SET_CQ_LEVEL",
141 [VP8E_SET_MAX_INTRA_BITRATE_PCT] =
"VP8E_SET_MAX_INTRA_BITRATE_PCT",
142 [VP8E_SET_SHARPNESS] =
"VP8E_SET_SHARPNESS",
143 [VP8E_SET_TEMPORAL_LAYER_ID] =
"VP8E_SET_TEMPORAL_LAYER_ID",
144 #if CONFIG_LIBVPX_VP9_ENCODER
145 [VP9E_SET_LOSSLESS] =
"VP9E_SET_LOSSLESS",
146 [VP9E_SET_TILE_COLUMNS] =
"VP9E_SET_TILE_COLUMNS",
147 [VP9E_SET_TILE_ROWS] =
"VP9E_SET_TILE_ROWS",
148 [VP9E_SET_FRAME_PARALLEL_DECODING] =
"VP9E_SET_FRAME_PARALLEL_DECODING",
149 [VP9E_SET_AQ_MODE] =
"VP9E_SET_AQ_MODE",
150 [VP9E_SET_COLOR_SPACE] =
"VP9E_SET_COLOR_SPACE",
151 [VP9E_SET_SVC_LAYER_ID] =
"VP9E_SET_SVC_LAYER_ID",
152 #if VPX_ENCODER_ABI_VERSION >= 12
153 [VP9E_SET_SVC_PARAMETERS] =
"VP9E_SET_SVC_PARAMETERS",
155 [VP9E_SET_SVC] =
"VP9E_SET_SVC",
156 #if VPX_ENCODER_ABI_VERSION >= 11
157 [VP9E_SET_COLOR_RANGE] =
"VP9E_SET_COLOR_RANGE",
159 #if VPX_ENCODER_ABI_VERSION >= 12
160 [VP9E_SET_TARGET_LEVEL] =
"VP9E_SET_TARGET_LEVEL",
161 [VP9E_GET_LEVEL] =
"VP9E_GET_LEVEL",
163 #ifdef VPX_CTRL_VP9E_SET_ROW_MT
164 [VP9E_SET_ROW_MT] =
"VP9E_SET_ROW_MT",
166 #ifdef VPX_CTRL_VP9E_SET_TUNE_CONTENT
167 [VP9E_SET_TUNE_CONTENT] =
"VP9E_SET_TUNE_CONTENT",
169 #ifdef VPX_CTRL_VP9E_SET_TPL
170 [VP9E_SET_TPL] =
"VP9E_SET_TPL",
178 const char *
error = vpx_codec_error(&
ctx->encoder);
179 const char *detail = vpx_codec_error_detail(&
ctx->encoder);
187 const struct vpx_codec_enc_cfg *cfg)
195 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
196 #
if CONFIG_LIBVPX_VP9_ENCODER
199 " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
200 width,
"g_usage:", cfg->g_usage,
201 width,
"g_threads:", cfg->g_threads,
202 width,
"g_profile:", cfg->g_profile,
203 width,
"g_w:", cfg->g_w,
204 width,
"g_h:", cfg->g_h,
205 #
if CONFIG_LIBVPX_VP9_ENCODER
206 width,
"g_bit_depth:", cfg->g_bit_depth,
207 width,
"g_input_bit_depth:", cfg->g_input_bit_depth,
209 width,
"g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
210 width,
"g_error_resilient:", cfg->g_error_resilient,
211 width,
"g_pass:", cfg->g_pass,
212 width,
"g_lag_in_frames:", cfg->g_lag_in_frames);
214 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
216 width,
"rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
217 width,
"rc_resize_allowed:", cfg->rc_resize_allowed,
218 width,
"rc_resize_up_thresh:", cfg->rc_resize_up_thresh,
219 width,
"rc_resize_down_thresh:", cfg->rc_resize_down_thresh,
220 width,
"rc_end_usage:", cfg->rc_end_usage,
221 width,
"rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
222 width,
"rc_target_bitrate:", cfg->rc_target_bitrate);
225 width,
"rc_min_quantizer:", cfg->rc_min_quantizer,
226 width,
"rc_max_quantizer:", cfg->rc_max_quantizer);
229 width,
"rc_undershoot_pct:", cfg->rc_undershoot_pct,
230 width,
"rc_overshoot_pct:", cfg->rc_overshoot_pct);
231 av_log(avctx,
level,
"temporal layering settings\n"
232 " %*s%u\n",
width,
"ts_number_layers:", cfg->ts_number_layers);
235 "\n %*s",
width,
"ts_target_bitrate:");
236 for (
i = 0;
i < VPX_TS_MAX_LAYERS;
i++)
238 "%u ", cfg->ts_target_bitrate[
i]);
240 #if (VPX_ENCODER_ABI_VERSION >= 12) && CONFIG_LIBVPX_VP9_ENCODER
243 "\n %*s",
width,
"layer_target_bitrate:");
244 for (
i = 0;
i < VPX_TS_MAX_LAYERS;
i++)
246 "%u ", cfg->layer_target_bitrate[
i]);
251 "\n %*s",
width,
"ts_rate_decimator:");
252 for (
i = 0;
i < VPX_TS_MAX_LAYERS;
i++)
256 "\n %*s%u\n",
width,
"ts_periodicity:", cfg->ts_periodicity);
258 "\n %*s",
width,
"ts_layer_id:");
259 for (
i = 0;
i < VPX_TS_MAX_PERIODICITY;
i++)
263 " %*s%u\n %*s%u\n %*s%u\n",
264 width,
"rc_buf_sz:", cfg->rc_buf_sz,
265 width,
"rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
266 width,
"rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
267 av_log(avctx,
level,
"2 pass rate control settings\n"
268 " %*s%u\n %*s%u\n %*s%u\n",
269 width,
"rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
270 width,
"rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
271 width,
"rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
272 #if VPX_ENCODER_ABI_VERSION >= 14
274 width,
"rc_2pass_vbr_corpus_complexity:", cfg->rc_2pass_vbr_corpus_complexity);
277 " %*s%d\n %*s%u\n %*s%u\n",
278 width,
"kf_mode:", cfg->kf_mode,
279 width,
"kf_min_dist:", cfg->kf_min_dist,
280 width,
"kf_max_dist:", cfg->kf_max_dist);
314 enum vp8e_enc_control_id
id,
int val)
324 res = vpx_codec_control(&
ctx->encoder,
id,
val);
325 if (res != VPX_CODEC_OK) {
331 return res == VPX_CODEC_OK ? 0 :
AVERROR(EINVAL);
334 #if VPX_ENCODER_ABI_VERSION >= 12
336 enum vp8e_enc_control_id
id,
int *
val)
346 res = vpx_codec_control(&
ctx->encoder,
id,
val);
347 if (res != VPX_CODEC_OK) {
353 return res == VPX_CODEC_OK ? 0 :
AVERROR(EINVAL);
361 #if VPX_ENCODER_ABI_VERSION >= 12
365 if (!codecctl_intp(avctx, VP9E_GET_LEVEL, &level_out))
372 vpx_codec_destroy(&
ctx->encoder);
374 vpx_codec_destroy(&
ctx->encoder_alpha);
387 char *saveptr =
NULL;
390 while (token && dest_idx < max_entries) {
391 dest[dest_idx++] = strtoul(token,
NULL, 10);
397 int *layer_flags,
int *flag_periodicity)
399 switch (layering_mode) {
404 static const int ids[2] = { 0, 1 };
405 cfg->ts_periodicity = 2;
406 *flag_periodicity = 2;
407 cfg->ts_number_layers = 2;
408 cfg->ts_rate_decimator[0] = 2;
409 cfg->ts_rate_decimator[1] = 1;
410 memcpy(cfg->ts_layer_id, ids,
sizeof(ids));
413 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
414 VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
416 VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_GF |
417 VP8_EFLAG_NO_UPD_LAST |
418 VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_REF_GF;
428 static const int ids[4] = { 0, 2, 1, 2 };
429 cfg->ts_periodicity = 4;
430 *flag_periodicity = 4;
431 cfg->ts_number_layers = 3;
432 cfg->ts_rate_decimator[0] = 4;
433 cfg->ts_rate_decimator[1] = 2;
434 cfg->ts_rate_decimator[2] = 1;
435 memcpy(cfg->ts_layer_id, ids,
sizeof(ids));
442 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
443 VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
445 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
446 VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
447 VP8_EFLAG_NO_UPD_ARF;
449 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
450 VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
452 VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_REF_ARF |
453 VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
454 VP8_EFLAG_NO_UPD_ARF;
463 static const int ids[4] = { 0, 2, 1, 2 };
464 cfg->ts_periodicity = 4;
465 *flag_periodicity = 4;
466 cfg->ts_number_layers = 3;
467 cfg->ts_rate_decimator[0] = 4;
468 cfg->ts_rate_decimator[1] = 2;
469 cfg->ts_rate_decimator[2] = 1;
470 memcpy(cfg->ts_layer_id, ids,
sizeof(ids));
476 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
477 VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
479 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
480 VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
482 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
483 VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
485 VP8_EFLAG_NO_REF_LAST |
486 VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
487 VP8_EFLAG_NO_UPD_ARF;
503 size_t value_len = strlen(
value);
504 int ts_layering_mode = 0;
509 if (!strcmp(
key,
"ts_number_layers"))
510 enccfg->ts_number_layers = strtoul(
value, &
value, 10);
511 else if (!strcmp(
key,
"ts_target_bitrate")) {
514 #if (VPX_ENCODER_ABI_VERSION >= 12) && CONFIG_LIBVPX_VP9_ENCODER
518 }
else if (!strcmp(
key,
"ts_rate_decimator")) {
520 }
else if (!strcmp(
key,
"ts_periodicity")) {
521 enccfg->ts_periodicity = strtoul(
value, &
value, 10);
522 }
else if (!strcmp(
key,
"ts_layer_id")) {
524 }
else if (!strcmp(
key,
"ts_layering_mode")) {
529 #if (VPX_ENCODER_ABI_VERSION >= 12) && CONFIG_LIBVPX_VP9_ENCODER
530 enccfg->temporal_layering_mode = VP9E_TEMPORAL_LAYERING_MODE_BYPASS;
531 enccfg->ss_number_layers = 1;
533 if (ts_layering_mode) {
543 #if CONFIG_LIBVPX_VP9_ENCODER
545 struct vpx_codec_enc_cfg *enccfg, vpx_codec_flags_t *
flags,
546 vpx_img_fmt_t *img_fmt)
549 enccfg->g_bit_depth = enccfg->g_input_bit_depth = 8;
553 enccfg->g_profile = 0;
554 *img_fmt = VPX_IMG_FMT_I420;
557 enccfg->g_profile = 1;
558 *img_fmt = VPX_IMG_FMT_I422;
561 enccfg->g_profile = 1;
562 *img_fmt = VPX_IMG_FMT_I440;
565 ctx->vpx_cs = VPX_CS_SRGB;
567 enccfg->g_profile = 1;
568 *img_fmt = VPX_IMG_FMT_I444;
572 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
573 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
575 enccfg->g_profile = 2;
576 *img_fmt = VPX_IMG_FMT_I42016;
577 *
flags |= VPX_CODEC_USE_HIGHBITDEPTH;
583 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
584 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
586 enccfg->g_profile = 3;
587 *img_fmt = VPX_IMG_FMT_I42216;
588 *
flags |= VPX_CODEC_USE_HIGHBITDEPTH;
594 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
595 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
597 enccfg->g_profile = 3;
598 *img_fmt = VPX_IMG_FMT_I44016;
599 *
flags |= VPX_CODEC_USE_HIGHBITDEPTH;
605 ctx->vpx_cs = VPX_CS_SRGB;
608 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
609 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
612 enccfg->g_profile = 3;
613 *img_fmt = VPX_IMG_FMT_I44416;
614 *
flags |= VPX_CODEC_USE_HIGHBITDEPTH;
627 enum vpx_color_space vpx_cs;
631 vpx_cs =
ctx->vpx_cs;
651 #if VPX_ENCODER_ABI_VERSION >= 11
654 enum vpx_color_range vpx_cr;
674 struct vpx_codec_enc_cfg *enccfg)
678 avctx->
bit_rate = enccfg->rc_target_bitrate * 1000;
679 if (enccfg->rc_end_usage == VPX_CQ) {
681 "Bitrate not specified for constrained quality mode, using default of %dkbit/sec\n",
682 enccfg->rc_target_bitrate);
684 enccfg->rc_end_usage = VPX_CQ;
687 "Neither bitrate nor constrained quality specified, using default CRF of %d and bitrate of %dkbit/sec\n",
688 ctx->crf, enccfg->rc_target_bitrate);
693 #if CONFIG_LIBVPX_VP9_ENCODER
699 struct vpx_codec_enc_cfg *enccfg)
703 if (enccfg->rc_end_usage != VPX_Q &&
ctx->lossless < 0) {
704 enccfg->rc_end_usage = VPX_Q;
707 "Neither bitrate nor constrained quality specified, using default CRF of %d\n",
718 struct vpx_codec_enc_cfg *enccfg)
721 #if CONFIG_LIBVPX_VP9_ENCODER
723 set_vp9_defaults(avctx, enccfg);
731 const struct vpx_codec_iface *iface)
734 struct vpx_codec_enc_cfg enccfg = { 0 };
735 struct vpx_codec_enc_cfg enccfg_alpha;
739 vpx_img_fmt_t img_fmt = VPX_IMG_FMT_I420;
740 #if CONFIG_LIBVPX_VP9_ENCODER
741 vpx_codec_caps_t codec_caps = vpx_codec_get_caps(iface);
742 vpx_svc_extra_cfg_t svc_params;
752 if ((res = vpx_codec_enc_config_default(iface, &enccfg, 0)) != VPX_CODEC_OK) {
754 vpx_codec_err_to_string(res));
758 #if CONFIG_LIBVPX_VP9_ENCODER
773 enccfg.g_w = avctx->
width;
774 enccfg.g_h = avctx->
height;
779 enccfg.g_lag_in_frames=
ctx->lag_in_frames;
782 enccfg.g_pass = VPX_RC_FIRST_PASS;
784 enccfg.g_pass = VPX_RC_LAST_PASS;
786 enccfg.g_pass = VPX_RC_ONE_PASS;
790 enccfg.rc_end_usage = VPX_CBR;
791 }
else if (
ctx->crf >= 0) {
792 enccfg.rc_end_usage = VPX_CQ;
793 #if CONFIG_LIBVPX_VP9_ENCODER
795 enccfg.rc_end_usage = VPX_Q;
802 #if CONFIG_LIBVPX_VP9_ENCODER
803 enccfg.ss_target_bitrate[0] = enccfg.rc_target_bitrate;
811 enccfg.rc_min_quantizer =
812 enccfg.rc_max_quantizer = 0;
814 if (avctx->
qmin >= 0)
815 enccfg.rc_min_quantizer = avctx->
qmin;
816 if (avctx->
qmax >= 0)
817 enccfg.rc_max_quantizer = avctx->
qmax;
820 if (enccfg.rc_end_usage == VPX_CQ
821 #
if CONFIG_LIBVPX_VP9_ENCODER
822 || enccfg.rc_end_usage == VPX_Q
825 if (
ctx->crf < enccfg.rc_min_quantizer ||
ctx->crf > enccfg.rc_max_quantizer) {
827 "CQ level %d must be between minimum and maximum quantizer value (%d-%d)\n",
828 ctx->crf, enccfg.rc_min_quantizer, enccfg.rc_max_quantizer);
833 #if FF_API_PRIVATE_OPT
839 enccfg.rc_dropframe_thresh =
ctx->drop_threshold;
844 enccfg.rc_2pass_vbr_minsection_pct =
847 enccfg.rc_2pass_vbr_maxsection_pct =
849 #if CONFIG_LIBVPX_VP9_ENCODER
851 #if VPX_ENCODER_ABI_VERSION >= 14
852 if (
ctx->corpus_complexity >= 0)
853 enccfg.rc_2pass_vbr_corpus_complexity =
ctx->corpus_complexity;
862 enccfg.rc_buf_initial_sz =
864 enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
865 if (
ctx->rc_undershoot_pct >= 0)
866 enccfg.rc_undershoot_pct =
ctx->rc_undershoot_pct;
867 if (
ctx->rc_overshoot_pct >= 0)
868 enccfg.rc_overshoot_pct =
ctx->rc_overshoot_pct;
874 enccfg.kf_max_dist = avctx->
gop_size;
876 if (enccfg.g_pass == VPX_RC_FIRST_PASS)
877 enccfg.g_lag_in_frames = 0;
878 else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
879 int decode_size,
ret;
886 ctx->twopass_stats.sz = strlen(avctx->
stats_in) * 3 / 4;
891 ctx->twopass_stats.sz);
892 ctx->twopass_stats.sz = 0;
896 ctx->twopass_stats.sz);
897 if (decode_size < 0) {
902 ctx->twopass_stats.sz = decode_size;
903 enccfg.rc_twopass_stats_in =
ctx->twopass_stats;
910 enccfg.g_profile = avctx->
profile;
917 "Error parsing option '%s = %s'.\n",
923 res = vpx_codec_enc_init(&
ctx->encoder, iface, &enccfg,
flags);
924 if (res != VPX_CODEC_OK) {
928 #if CONFIG_LIBVPX_VP9_ENCODER
930 memset(&svc_params, 0,
sizeof(svc_params));
931 for (
int i = 0;
i < enccfg.ts_number_layers; ++
i) {
932 svc_params.max_quantizers[
i] = enccfg.rc_max_quantizer;
933 svc_params.min_quantizers[
i] = enccfg.rc_min_quantizer;
935 svc_params.scaling_factor_num[0] = enccfg.g_h;
936 svc_params.scaling_factor_den[0] = enccfg.g_h;
937 #if VPX_ENCODER_ABI_VERSION >= 12
939 codecctl_intp(avctx, VP9E_SET_SVC_PARAMETERS, (
int *)&svc_params);
944 enccfg_alpha = enccfg;
945 res = vpx_codec_enc_init(&
ctx->encoder_alpha, iface, &enccfg_alpha,
flags);
946 if (res != VPX_CODEC_OK) {
956 ctx->auto_alt_ref = 1;
957 if (
ctx->auto_alt_ref >= 0)
960 if (
ctx->arnr_max_frames >= 0)
962 if (
ctx->arnr_strength >= 0)
964 if (
ctx->arnr_type >= 0)
970 av_log(avctx,
AV_LOG_ERROR,
"Transparency encoding with auto_alt_ref does not work\n");
974 if (
ctx->sharpness >= 0)
978 #if FF_API_PRIVATE_OPT
984 codecctl_int(avctx, VP8E_SET_NOISE_SENSITIVITY,
ctx->noise_sensitivity);
990 if (
ctx->max_intra_rate >= 0)
991 codecctl_int(avctx, VP8E_SET_MAX_INTRA_BITRATE_PCT,
ctx->max_intra_rate);
993 #if CONFIG_LIBVPX_VP9_ENCODER
995 if (
ctx->lossless >= 0)
997 if (
ctx->tile_columns >= 0)
999 if (
ctx->tile_rows >= 0)
1001 if (
ctx->frame_parallel >= 0)
1002 codecctl_int(avctx, VP9E_SET_FRAME_PARALLEL_DECODING,
ctx->frame_parallel);
1003 if (
ctx->aq_mode >= 0)
1005 set_colorspace(avctx);
1006 #if VPX_ENCODER_ABI_VERSION >= 11
1009 #if VPX_ENCODER_ABI_VERSION >= 12
1012 #ifdef VPX_CTRL_VP9E_SET_ROW_MT
1013 if (
ctx->row_mt >= 0)
1016 #ifdef VPX_CTRL_VP9E_SET_TUNE_CONTENT
1017 if (
ctx->tune_content >= 0)
1020 #ifdef VPX_CTRL_VP9E_SET_TPL
1021 if (
ctx->tpl_model >= 0)
1030 vpx_img_wrap(&
ctx->rawimg, img_fmt, avctx->
width, avctx->
height, 1,
1032 #if CONFIG_LIBVPX_VP9_ENCODER
1034 ctx->rawimg.bit_depth = enccfg.g_bit_depth;
1041 if (enccfg.rc_end_usage == VPX_CBR ||
1042 enccfg.g_pass != VPX_RC_ONE_PASS) {
1053 const struct vpx_codec_cx_pkt *
src,
1054 const struct vpx_codec_cx_pkt *src_alpha,
1057 dst->
pts =
src->data.frame.pts;
1059 dst->
flags =
src->data.frame.flags;
1060 dst->
sz =
src->data.frame.sz;
1061 dst->
buf =
src->data.frame.buf;
1064 if (!(dst->
flags & VPX_FRAME_IS_INVISIBLE)) {
1067 if (
ctx->have_sse) {
1072 memcpy(dst->
sse,
ctx->sse,
sizeof(dst->
sse));
1079 dst->
buf_alpha = src_alpha->data.frame.buf;
1080 dst->
sz_alpha = src_alpha->data.frame.sz;
1103 #if FF_API_CODED_FRAME
1110 if (!!(cx_frame->
flags & VPX_FRAME_IS_KEY)) {
1112 #if FF_API_CODED_FRAME
1120 #if FF_API_CODED_FRAME
1128 cx_frame->
have_sse ? 3 : 0, pict_type);
1133 #if FF_API_CODED_FRAME
1141 for (
i = 0;
i < 3; ++
i) {
1174 const struct vpx_codec_cx_pkt *
pkt;
1175 const struct vpx_codec_cx_pkt *pkt_alpha =
NULL;
1176 const void *iter =
NULL;
1177 const void *iter_alpha =
NULL;
1180 if (
ctx->coded_frame_list) {
1186 ctx->coded_frame_list = cx_frame->
next;
1192 while ((
pkt = vpx_codec_get_cx_data(&
ctx->encoder, &iter)) &&
1194 (pkt_alpha = vpx_codec_get_cx_data(&
ctx->encoder_alpha, &iter_alpha)))) {
1195 switch (
pkt->kind) {
1196 case VPX_CODEC_CX_FRAME_PKT:
1212 "Frame queue element alloc failed\n");
1218 if (!cx_frame->
buf) {
1226 if (
ctx->is_alpha) {
1235 memcpy(cx_frame->
buf_alpha, pkt_alpha->data.frame.buf, pkt_alpha->data.frame.sz);
1240 case VPX_CODEC_STATS_PKT: {
1241 struct vpx_fixed_buf *
stats = &
ctx->twopass_stats;
1245 pkt->
data.twopass_stats.sz)) < 0) {
1255 case VPX_CODEC_PSNR_PKT:
1263 case VPX_CODEC_CUSTOM_PKT:
1273 vpx_roi_map_t *roi_map,
int block_size,
int segment_cnt)
1278 #define MAX_DELTA_Q 63
1293 memset(roi_map, 0,
sizeof(*roi_map));
1304 if (!self_size || sd->
size % self_size) {
1308 nb_rois = sd->
size / self_size;
1314 for (
int i = 0;
i < nb_rois;
i++) {
1328 if (!segment_mapping[mapping_index]) {
1329 if (segment_id == segment_cnt) {
1331 "ROI only supports %d segments (and segment 0 is reserved for non-ROIs), skipping the left ones.\n",
1336 segment_mapping[mapping_index] = segment_id + 1;
1337 roi_map->delta_q[segment_id] =
delta_q;
1342 roi_map->rows = (frame_height + block_size - 1) / block_size;
1343 roi_map->cols = (frame_width + block_size - 1) / block_size;
1344 roi_map->roi_map =
av_mallocz_array(roi_map->rows * roi_map->cols,
sizeof(*roi_map->roi_map));
1345 if (!roi_map->roi_map) {
1353 for (
int i = nb_rois - 1;
i >= 0;
i--) {
1356 int starty, endy, startx, endx;
1360 starty = av_clip(roi->
top / block_size, 0, roi_map->rows);
1361 endy = av_clip((roi->
bottom + block_size - 1) / block_size, 0, roi_map->rows);
1362 startx = av_clip(roi->
left / block_size, 0, roi_map->cols);
1363 endx = av_clip((roi->
right + block_size - 1) / block_size, 0, roi_map->cols);
1369 if (mapping_value) {
1370 for (
int y = starty; y < endy; y++)
1371 for (
int x = startx; x < endx; x++)
1372 roi_map->roi_map[x + y * roi_map->cols] = mapping_value - 1;
1383 #ifdef VPX_CTRL_VP9E_SET_ROI_MAP
1384 int version = vpx_codec_version();
1387 int patch = VPX_VERSION_PATCH(
version);
1390 vpx_roi_map_t roi_map;
1391 const int segment_cnt = 8;
1392 const int block_size = 8;
1395 if (
ctx->aq_mode > 0 ||
ctx->cpu_used < 5 ||
ctx->deadline != VPX_DL_REALTIME) {
1396 if (!
ctx->roi_warned) {
1397 ctx->roi_warned = 1;
1399 "and deadline is REALTIME, so skipping ROI.\n");
1404 ret =
set_roi_map(avctx, sd, frame_width, frame_height, &roi_map, block_size, segment_cnt);
1410 memset(roi_map.ref_frame, -1,
sizeof(roi_map.ref_frame));
1412 if (vpx_codec_control(&
ctx->encoder, VP9E_SET_ROI_MAP, &roi_map)) {
1421 if (!
ctx->roi_warned) {
1422 ctx->roi_warned = 1;
1423 av_log(avctx,
AV_LOG_WARNING,
"ROI is not supported, please upgrade libvpx to version >= 1.8.1. "
1424 "You may need to rebuild ffmpeg.\n");
1431 vpx_roi_map_t roi_map;
1432 const int segment_cnt = 4;
1433 const int block_size = 16;
1436 int ret =
set_roi_map(avctx, sd, frame_width, frame_height, &roi_map, block_size, segment_cnt);
1442 if (vpx_codec_control(&
ctx->encoder, VP8E_SET_ROI_MAP, &roi_map)) {
1454 struct vpx_image *rawimg_alpha = &
ctx->rawimg_alpha;
1455 unsigned char **
planes = rawimg_alpha->planes;
1456 int *
stride = rawimg_alpha->stride;
1458 if (!
planes[VPX_PLANE_U] ||
1460 width != (
int)rawimg_alpha->d_w ||
1461 height != (
int)rawimg_alpha->d_h) {
1465 vpx_img_wrap(rawimg_alpha, VPX_IMG_FMT_I420,
width,
height, 1,
1483 struct vpx_image *rawimg =
NULL;
1484 struct vpx_image *rawimg_alpha =
NULL;
1485 int64_t timestamp = 0;
1486 int res, coded_size;
1487 vpx_enc_frame_flags_t
flags = 0;
1488 const struct vpx_codec_enc_cfg *enccfg =
ctx->encoder.config.enc;
1489 vpx_svc_layer_id_t layer_id;
1490 int layer_id_valid = 0;
1494 rawimg = &
ctx->rawimg;
1495 rawimg->planes[VPX_PLANE_Y] =
frame->data[0];
1496 rawimg->planes[VPX_PLANE_U] =
frame->data[1];
1497 rawimg->planes[VPX_PLANE_V] =
frame->data[2];
1498 rawimg->stride[VPX_PLANE_Y] =
frame->linesize[0];
1499 rawimg->stride[VPX_PLANE_U] =
frame->linesize[1];
1500 rawimg->stride[VPX_PLANE_V] =
frame->linesize[2];
1501 if (
ctx->is_alpha) {
1502 rawimg_alpha = &
ctx->rawimg_alpha;
1506 rawimg_alpha->planes[VPX_PLANE_Y] =
frame->data[3];
1507 rawimg_alpha->stride[VPX_PLANE_Y] =
frame->linesize[3];
1509 timestamp =
frame->pts;
1510 #if VPX_IMAGE_ABI_VERSION >= 4
1511 switch (
frame->color_range) {
1513 rawimg->range = VPX_CR_STUDIO_RANGE;
1516 rawimg->range = VPX_CR_FULL_RANGE;
1521 flags |= VPX_EFLAG_FORCE_KF;
1522 if (
frame->metadata) {
1528 memset(&layer_id, 0,
sizeof(layer_id));
1532 layer_id.temporal_layer_id = strtoul(en->
value,
NULL, 10);
1533 #ifdef VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT
1534 layer_id.temporal_layer_id_per_spatial[0] = layer_id.temporal_layer_id;
1551 if (enccfg->ts_number_layers > 1 &&
ctx->ts_layer_flags) {
1552 if (
flags & VPX_EFLAG_FORCE_KF) {
1554 ctx->current_temporal_idx = 0;
1555 flags = VPX_EFLAG_FORCE_KF;
1561 flags |=
ctx->ts_layer_flags[
ctx->current_temporal_idx];
1563 memset(&layer_id, 0,
sizeof(layer_id));
1564 #if VPX_ENCODER_ABI_VERSION >= 12
1565 layer_id.spatial_layer_id = 0;
1567 layer_id.temporal_layer_id = enccfg->ts_layer_id[
ctx->current_temporal_idx];
1568 #ifdef VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT
1569 layer_id.temporal_layer_id_per_spatial[0] = layer_id.temporal_layer_id;
1574 if (layer_id_valid) {
1576 codecctl_int(avctx, VP8E_SET_TEMPORAL_LAYER_ID, layer_id.temporal_layer_id);
1578 #if CONFIG_LIBVPX_VP9_ENCODER && VPX_ENCODER_ABI_VERSION >= 12
1580 codecctl_intp(avctx, VP9E_SET_SVC_LAYER_ID, (
int *)&layer_id);
1585 res = vpx_codec_encode(&
ctx->encoder, rawimg, timestamp,
1587 if (res != VPX_CODEC_OK) {
1592 if (
ctx->is_alpha) {
1593 res = vpx_codec_encode(&
ctx->encoder_alpha, rawimg_alpha, timestamp,
1595 if (res != VPX_CODEC_OK) {
1613 ctx->twopass_stats.sz);
1614 }
else if (enccfg->ts_number_layers > 1 &&
ctx->ts_layer_flags) {
1615 ctx->current_temporal_idx = (
ctx->current_temporal_idx + 1) % enccfg->ts_periodicity;
1618 *got_packet = !!coded_size;
1622 #define OFFSET(x) offsetof(VPxContext, x)
1623 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
1625 #define COMMON_OPTIONS \
1626 { "lag-in-frames", "Number of frames to look ahead for " \
1627 "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
1628 { "arnr-maxframes", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
1629 { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
1630 { "arnr-type", "altref noise reduction filter type", OFFSET(arnr_type), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, "arnr_type"}, \
1631 { "backward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "arnr_type" }, \
1632 { "forward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "arnr_type" }, \
1633 { "centered", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "arnr_type" }, \
1634 { "tune", "Tune the encoding to a specific scenario", OFFSET(tune), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, "tune"}, \
1635 { "psnr", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VP8_TUNE_PSNR}, 0, 0, VE, "tune"}, \
1636 { "ssim", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VP8_TUNE_SSIM}, 0, 0, VE, "tune"}, \
1637 { "deadline", "Time to spend encoding, in microseconds.", OFFSET(deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
1638 { "best", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_BEST_QUALITY}, 0, 0, VE, "quality"}, \
1639 { "good", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_GOOD_QUALITY}, 0, 0, VE, "quality"}, \
1640 { "realtime", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_REALTIME}, 0, 0, VE, "quality"}, \
1641 { "error-resilient", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX, VE, "er"}, \
1642 { "max-intra-rate", "Maximum I-frame bitrate (pct) 0=unlimited", OFFSET(max_intra_rate), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
1643 { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, "er"}, \
1644 { "partitions", "The frame partitions are independently decodable " \
1645 "by the bool decoder, meaning that partitions can be decoded even " \
1646 "though earlier partitions have been lost. Note that intra prediction" \
1647 " is still done over the partition boundary.", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0, VE, "er"}, \
1648 { "crf", "Select the quality for constant quality mode", offsetof(VPxContext, crf), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 63, VE }, \
1649 { "static-thresh", "A change threshold on blocks below which they will be skipped by the encoder", OFFSET(static_thresh), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE }, \
1650 { "drop-threshold", "Frame drop threshold", offsetof(VPxContext, drop_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, VE }, \
1651 { "noise-sensitivity", "Noise sensitivity", OFFSET(noise_sensitivity), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 4, VE}, \
1652 { "undershoot-pct", "Datarate undershoot (min) target (%)", OFFSET(rc_undershoot_pct), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 100, VE }, \
1653 { "overshoot-pct", "Datarate overshoot (max) target (%)", OFFSET(rc_overshoot_pct), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 1000, VE }, \
1654 { "ts-parameters", "Temporal scaling configuration using a :-separated list of key=value parameters", OFFSET(vpx_ts_parameters), AV_OPT_TYPE_DICT, {.str=NULL}, 0, 0, VE}, \
1656 #define LEGACY_OPTIONS \
1657 {"speed", "", offsetof(VPxContext, cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
1658 {"quality", "", offsetof(VPxContext, deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
1659 {"vp8flags", "", offsetof(VPxContext, flags), AV_OPT_TYPE_FLAGS, {.i64 = 0}, 0, UINT_MAX, VE, "flags"}, \
1660 {"error_resilient", "enable error resilience", 0, AV_OPT_TYPE_CONST, {.i64 = VP8F_ERROR_RESILIENT}, INT_MIN, INT_MAX, VE, "flags"}, \
1661 {"altref", "enable use of alternate reference frames (VP8/2-pass only)", 0, AV_OPT_TYPE_CONST, {.i64 = VP8F_AUTO_ALT_REF}, INT_MIN, INT_MAX, VE, "flags"}, \
1662 {"arnr_max_frames", "altref noise reduction max frame count", offsetof(VPxContext, arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 15, VE}, \
1663 {"arnr_strength", "altref noise reduction filter strength", offsetof(VPxContext, arnr_strength), AV_OPT_TYPE_INT, {.i64 = 3}, 0, 6, VE}, \
1664 {"arnr_type", "altref noise reduction filter type", offsetof(VPxContext, arnr_type), AV_OPT_TYPE_INT, {.i64 = 3}, 1, 3, VE}, \
1665 {"rc_lookahead", "Number of frames to look ahead for alternate reference frame selection", offsetof(VPxContext, lag_in_frames), AV_OPT_TYPE_INT, {.i64 = 25}, 0, 25, VE}, \
1666 {"sharpness", "Increase sharpness at the expense of lower PSNR", offsetof(VPxContext, sharpness), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 7, VE},
1668 #if CONFIG_LIBVPX_VP8_ENCODER
1669 static const AVOption vp8_options[] = {
1671 {
"auto-alt-ref",
"Enable use of alternate reference "
1679 #if CONFIG_LIBVPX_VP9_ENCODER
1680 static const AVOption vp9_options[] = {
1682 {
"auto-alt-ref",
"Enable use of alternate reference "
1686 {
"tile-columns",
"Number of tile columns to use, log2",
OFFSET(tile_columns),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 6,
VE},
1688 {
"frame-parallel",
"Enable frame parallel decodability features",
OFFSET(frame_parallel),
AV_OPT_TYPE_BOOL,{.i64 = -1}, -1, 1,
VE},
1689 #if VPX_ENCODER_ABI_VERSION >= 12
1690 {
"aq-mode",
"adaptive quantization mode",
OFFSET(aq_mode),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 4,
VE,
"aq_mode"},
1692 {
"aq-mode",
"adaptive quantization mode",
OFFSET(aq_mode),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 3,
VE,
"aq_mode"},
1695 {
"variance",
"Variance based Aq", 0,
AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0,
VE,
"aq_mode" },
1696 {
"complexity",
"Complexity based Aq", 0,
AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0,
VE,
"aq_mode" },
1698 #if VPX_ENCODER_ABI_VERSION >= 12
1702 #ifdef VPX_CTRL_VP9E_SET_ROW_MT
1705 #ifdef VPX_CTRL_VP9E_SET_TUNE_CONTENT
1706 #if VPX_ENCODER_ABI_VERSION >= 14
1707 {
"tune-content",
"Tune content type",
OFFSET(tune_content),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2,
VE,
"tune_content" },
1709 {
"tune-content",
"Tune content type",
OFFSET(tune_content),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1,
VE,
"tune_content" },
1711 {
"default",
"Regular video content", 0,
AV_OPT_TYPE_CONST, {.i64 = 0}, 0, 0,
VE,
"tune_content" },
1712 {
"screen",
"Screen capture content", 0,
AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0,
VE,
"tune_content" },
1713 #if VPX_ENCODER_ABI_VERSION >= 14
1714 {
"film",
"Film content; improves grain retention", 0,
AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0,
VE,
"tune_content" },
1717 #if VPX_ENCODER_ABI_VERSION >= 14
1718 {
"corpus-complexity",
"corpus vbr complexity midpoint",
OFFSET(corpus_complexity),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 10000,
VE },
1720 #ifdef VPX_CTRL_VP9E_SET_TPL
1728 #undef COMMON_OPTIONS
1729 #undef LEGACY_OPTIONS
1736 {
"keyint_min",
"-1" },
1740 #if CONFIG_LIBVPX_VP8_ENCODER
1743 return vpx_init(avctx, vpx_codec_vp8_cx());
1746 static const AVClass class_vp8 = {
1749 .option = vp8_options,
1758 .priv_data_size =
sizeof(VPxContext),
1764 .priv_class = &class_vp8,
1766 .wrapper_name =
"libvpx",
1770 #if CONFIG_LIBVPX_VP9_ENCODER
1773 return vpx_init(avctx, vpx_codec_vp9_cx());
1776 static const AVClass class_vp9 = {
1779 .option = vp9_options,
1784 .
name =
"libvpx-vp9",
1788 .priv_data_size =
sizeof(VPxContext),
1794 .priv_class = &class_vp9,
1797 .wrapper_name =
"libvpx",