From 25835e25931c28b6edb01a0e639c9be0770b519f Mon Sep 17 00:00:00 2001 From: James Almer Date: Sat, 25 Nov 2023 11:37:55 -0300 Subject: [PATCH] avformat: Immersive Audio Model and Formats muxer Signed-off-by: James Almer --- Changelog | 2 +- libavformat/Makefile | 1 + libavformat/allformats.c | 1 + libavformat/iamf_writer.c | 860 ++++++++++++++++++++++++++++++++++++++ libavformat/iamf_writer.h | 51 +++ libavformat/iamfenc.c | 387 +++++++++++++++++ libavformat/version.h | 2 +- 7 files changed, 1302 insertions(+), 2 deletions(-) create mode 100644 libavformat/iamf_writer.c create mode 100644 libavformat/iamf_writer.h create mode 100644 libavformat/iamfenc.c diff --git a/Changelog b/Changelog index 52ceb4f192..cdf1c5f09a 100644 --- a/Changelog +++ b/Changelog @@ -10,7 +10,7 @@ version : - demuxing, decoding, filtering, encoding, and muxing in the ffmpeg CLI now all run in parallel - enable gdigrab device to grab a window using the hwnd=HANDLER syntax -- IAMF raw demuxer +- IAMF raw demuxer and muxer version 6.1: - libaribcaption decoder diff --git a/libavformat/Makefile b/libavformat/Makefile index f23c22792b..581e378d95 100644 --- a/libavformat/Makefile +++ b/libavformat/Makefile @@ -259,6 +259,7 @@ OBJS-$(CONFIG_HLS_DEMUXER) += hls.o hls_sample_encryption.o OBJS-$(CONFIG_HLS_MUXER) += hlsenc.o hlsplaylist.o avc.o OBJS-$(CONFIG_HNM_DEMUXER) += hnm.o OBJS-$(CONFIG_IAMF_DEMUXER) += iamfdec.o iamf_parse.o iamf.o +OBJS-$(CONFIG_IAMF_MUXER) += iamfenc.o iamf_writer.o iamf.o OBJS-$(CONFIG_ICO_DEMUXER) += icodec.o OBJS-$(CONFIG_ICO_MUXER) += icoenc.o OBJS-$(CONFIG_IDCIN_DEMUXER) += idcin.o diff --git a/libavformat/allformats.c b/libavformat/allformats.c index 6e520b78a6..ce6be5f04d 100644 --- a/libavformat/allformats.c +++ b/libavformat/allformats.c @@ -213,6 +213,7 @@ extern const AVInputFormat ff_hls_demuxer; extern const FFOutputFormat ff_hls_muxer; extern const AVInputFormat ff_hnm_demuxer; extern const AVInputFormat ff_iamf_demuxer; +extern const FFOutputFormat ff_iamf_muxer; extern const AVInputFormat ff_ico_demuxer; extern const FFOutputFormat ff_ico_muxer; extern const AVInputFormat ff_idcin_demuxer; diff --git a/libavformat/iamf_writer.c b/libavformat/iamf_writer.c new file mode 100644 index 0000000000..9962845049 --- /dev/null +++ b/libavformat/iamf_writer.c @@ -0,0 +1,860 @@ +/* + * Immersive Audio Model and Formats muxing helpers and structs + * Copyright (c) 2023 James Almer + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/channel_layout.h" +#include "libavutil/intreadwrite.h" +#include "libavutil/iamf.h" +#include "libavutil/mem.h" +#include "libavcodec/get_bits.h" +#include "libavcodec/flac.h" +#include "libavcodec/mpeg4audio.h" +#include "libavcodec/put_bits.h" +#include "avformat.h" +#include "avio_internal.h" +#include "iamf.h" +#include "iamf_writer.h" + + +static int update_extradata(IAMFCodecConfig *codec_config) +{ + GetBitContext gb; + PutBitContext pb; + int ret; + + switch(codec_config->codec_id) { + case AV_CODEC_ID_OPUS: + if (codec_config->extradata_size < 19) + return AVERROR_INVALIDDATA; + codec_config->extradata_size -= 8; + memmove(codec_config->extradata, codec_config->extradata + 8, codec_config->extradata_size); + AV_WB8(codec_config->extradata + 1, 2); // set channels to stereo + break; + case AV_CODEC_ID_FLAC: { + uint8_t buf[13]; + + init_put_bits(&pb, buf, sizeof(buf)); + ret = init_get_bits8(&gb, codec_config->extradata, codec_config->extradata_size); + if (ret < 0) + return ret; + + put_bits32(&pb, get_bits_long(&gb, 32)); // min/max blocksize + put_bits64(&pb, 48, get_bits64(&gb, 48)); // min/max framesize + put_bits(&pb, 20, get_bits(&gb, 20)); // samplerate + skip_bits(&gb, 3); + put_bits(&pb, 3, 1); // set channels to stereo + ret = put_bits_left(&pb); + put_bits(&pb, ret, get_bits(&gb, ret)); + flush_put_bits(&pb); + + memcpy(codec_config->extradata, buf, sizeof(buf)); + break; + } + default: + break; + } + + return 0; +} + +static int fill_codec_config(IAMFContext *iamf, const AVStreamGroup *stg, + IAMFCodecConfig *codec_config) +{ + const AVStream *st = stg->streams[0]; + IAMFCodecConfig **tmp; + int j, ret = 0; + + codec_config->codec_id = st->codecpar->codec_id; + codec_config->sample_rate = st->codecpar->sample_rate; + codec_config->codec_tag = st->codecpar->codec_tag; + codec_config->nb_samples = st->codecpar->frame_size; + codec_config->seek_preroll = st->codecpar->seek_preroll; + if (st->codecpar->extradata_size) { + codec_config->extradata = av_memdup(st->codecpar->extradata, st->codecpar->extradata_size); + if (!codec_config->extradata) + return AVERROR(ENOMEM); + codec_config->extradata_size = st->codecpar->extradata_size; + ret = update_extradata(codec_config); + if (ret < 0) + goto fail; + } + + for (j = 0; j < iamf->nb_codec_configs; j++) { + if (!memcmp(iamf->codec_configs[j], codec_config, offsetof(IAMFCodecConfig, extradata)) && + (!codec_config->extradata_size || !memcmp(iamf->codec_configs[j]->extradata, + codec_config->extradata, codec_config->extradata_size))) + break; + } + + if (j < iamf->nb_codec_configs) { + av_free(iamf->codec_configs[j]->extradata); + av_free(iamf->codec_configs[j]); + iamf->codec_configs[j] = codec_config; + return j; + } + + tmp = av_realloc_array(iamf->codec_configs, iamf->nb_codec_configs + 1, sizeof(*iamf->codec_configs)); + if (!tmp) { + ret = AVERROR(ENOMEM); + goto fail; + } + + iamf->codec_configs = tmp; + iamf->codec_configs[iamf->nb_codec_configs] = codec_config; + codec_config->codec_config_id = iamf->nb_codec_configs; + + return iamf->nb_codec_configs++; + +fail: + av_freep(&codec_config->extradata); + return ret; +} + +static IAMFParamDefinition *add_param_definition(IAMFContext *iamf, AVIAMFParamDefinition *param, + const IAMFAudioElement *audio_element, void *log_ctx) +{ + IAMFParamDefinition **tmp, *param_definition; + IAMFCodecConfig *codec_config = NULL; + + tmp = av_realloc_array(iamf->param_definitions, iamf->nb_param_definitions + 1, + sizeof(*iamf->param_definitions)); + if (!tmp) + return NULL; + + iamf->param_definitions = tmp; + + param_definition = av_mallocz(sizeof(*param_definition)); + if (!param_definition) + return NULL; + + if (audio_element) + codec_config = iamf->codec_configs[audio_element->codec_config_id]; + + if (!param->parameter_rate) { + if (!codec_config) { + av_log(log_ctx, AV_LOG_ERROR, "parameter_rate needed but not set for parameter_id %u\n", + param->parameter_id); + return NULL; + } + param->parameter_rate = codec_config->sample_rate; + } + if (codec_config) { + if (!param->duration) + param->duration = codec_config->nb_samples; + if (!param->constant_subblock_duration) + param->constant_subblock_duration = codec_config->nb_samples; + } + + param_definition->mode = !!param->duration; + param_definition->param = param; + param_definition->audio_element = audio_element; + iamf->param_definitions[iamf->nb_param_definitions++] = param_definition; + + return param_definition; +} + +int ff_iamf_add_audio_element(IAMFContext *iamf, const AVStreamGroup *stg, void *log_ctx) +{ + const AVIAMFAudioElement *iamf_audio_element; + IAMFAudioElement **tmp, *audio_element; + IAMFCodecConfig *codec_config; + int ret; + + if (stg->type != AV_STREAM_GROUP_PARAMS_IAMF_AUDIO_ELEMENT) + return AVERROR(EINVAL); + + iamf_audio_element = stg->params.iamf_audio_element; + if (iamf_audio_element->audio_element_type == AV_IAMF_AUDIO_ELEMENT_TYPE_SCENE) { + const AVIAMFLayer *layer = iamf_audio_element->layers[0]; + if (iamf_audio_element->nb_layers != 1) { + av_log(log_ctx, AV_LOG_ERROR, "Invalid amount of layers for SCENE_BASED audio element. Must be 1\n"); + return AVERROR(EINVAL); + } + if (layer->ch_layout.order != AV_CHANNEL_ORDER_CUSTOM && + layer->ch_layout.order != AV_CHANNEL_ORDER_AMBISONIC) { + av_log(log_ctx, AV_LOG_ERROR, "Invalid channel layout for SCENE_BASED audio element\n"); + return AVERROR(EINVAL); + } + if (layer->ambisonics_mode >= AV_IAMF_AMBISONICS_MODE_PROJECTION) { + av_log(log_ctx, AV_LOG_ERROR, "Unsuported ambisonics mode %d\n", layer->ambisonics_mode); + return AVERROR_PATCHWELCOME; + } + for (int i = 0; i < stg->nb_streams; i++) { + if (stg->streams[i]->codecpar->ch_layout.nb_channels > 1) { + av_log(log_ctx, AV_LOG_ERROR, "Invalid amount of channels in a stream for MONO mode ambisonics\n"); + return AVERROR(EINVAL); + } + } + } else + for (int j, i = 0; i < iamf_audio_element->nb_layers; i++) { + const AVIAMFLayer *layer = iamf_audio_element->layers[i]; + for (j = 0; j < FF_ARRAY_ELEMS(ff_iamf_scalable_ch_layouts); j++) + if (!av_channel_layout_compare(&layer->ch_layout, &ff_iamf_scalable_ch_layouts[j])) + break; + + if (j >= FF_ARRAY_ELEMS(ff_iamf_scalable_ch_layouts)) { + av_log(log_ctx, AV_LOG_ERROR, "Unsupported channel layout in stream group #%d\n", i); + return AVERROR(EINVAL); + } + } + + for (int i = 0; i < iamf->nb_audio_elements; i++) { + if (stg->id == iamf->audio_elements[i]->audio_element_id) { + av_log(log_ctx, AV_LOG_ERROR, "Duplicated Audio Element id %"PRId64"\n", stg->id); + return AVERROR(EINVAL); + } + } + + codec_config = av_mallocz(sizeof(*codec_config)); + if (!codec_config) + return AVERROR(ENOMEM); + + ret = fill_codec_config(iamf, stg, codec_config); + if (ret < 0) { + av_free(codec_config); + return ret; + } + + audio_element = av_mallocz(sizeof(*audio_element)); + if (!audio_element) + return AVERROR(ENOMEM); + + audio_element->element = stg->params.iamf_audio_element; + audio_element->audio_element_id = stg->id; + audio_element->codec_config_id = ret; + + audio_element->substreams = av_calloc(stg->nb_streams, sizeof(*audio_element->substreams)); + if (!audio_element->substreams) + return AVERROR(ENOMEM); + audio_element->nb_substreams = stg->nb_streams; + + audio_element->layers = av_calloc(iamf_audio_element->nb_layers, sizeof(*audio_element->layers)); + if (!audio_element->layers) + return AVERROR(ENOMEM); + + for (int i = 0, j = 0; i < iamf_audio_element->nb_layers; i++) { + int nb_channels = iamf_audio_element->layers[i]->ch_layout.nb_channels; + + IAMFLayer *layer = &audio_element->layers[i]; + if (!layer) + return AVERROR(ENOMEM); + memset(layer, 0, sizeof(*layer)); + + if (i) + nb_channels -= iamf_audio_element->layers[i - 1]->ch_layout.nb_channels; + for (; nb_channels > 0 && j < stg->nb_streams; j++) { + const AVStream *st = stg->streams[j]; + IAMFSubStream *substream = &audio_element->substreams[j]; + + substream->audio_substream_id = st->id; + layer->substream_count++; + layer->coupled_substream_count += st->codecpar->ch_layout.nb_channels == 2; + nb_channels -= st->codecpar->ch_layout.nb_channels; + } + if (nb_channels) { + av_log(log_ctx, AV_LOG_ERROR, "Invalid channel count across substreams in layer %u from stream group %u\n", + i, stg->index); + return AVERROR(EINVAL); + } + } + + if (iamf_audio_element->demixing_info) { + AVIAMFParamDefinition *param = iamf_audio_element->demixing_info; + IAMFParamDefinition *param_definition = ff_iamf_get_param_definition(iamf, param->parameter_id); + + if (param->nb_subblocks != 1) { + av_log(log_ctx, AV_LOG_ERROR, "nb_subblocks in demixing_info for stream group %u is not 1\n", stg->index); + return AVERROR(EINVAL); + } + + if (!param_definition) { + param_definition = add_param_definition(iamf, param, audio_element, log_ctx); + if (!param_definition) + return AVERROR(ENOMEM); + } + } + if (iamf_audio_element->recon_gain_info) { + AVIAMFParamDefinition *param = iamf_audio_element->recon_gain_info; + IAMFParamDefinition *param_definition = ff_iamf_get_param_definition(iamf, param->parameter_id); + + if (param->nb_subblocks != 1) { + av_log(log_ctx, AV_LOG_ERROR, "nb_subblocks in recon_gain_info for stream group %u is not 1\n", stg->index); + return AVERROR(EINVAL); + } + + if (!param_definition) { + param_definition = add_param_definition(iamf, param, audio_element, log_ctx); + if (!param_definition) + return AVERROR(ENOMEM); + } + } + + tmp = av_realloc_array(iamf->audio_elements, iamf->nb_audio_elements + 1, sizeof(*iamf->audio_elements)); + if (!tmp) + return AVERROR(ENOMEM); + + iamf->audio_elements = tmp; + iamf->audio_elements[iamf->nb_audio_elements++] = audio_element; + + return 0; +} + +int ff_iamf_add_mix_presentation(IAMFContext *iamf, const AVStreamGroup *stg, void *log_ctx) +{ + IAMFMixPresentation **tmp, *mix_presentation; + + if (stg->type != AV_STREAM_GROUP_PARAMS_IAMF_MIX_PRESENTATION) + return AVERROR(EINVAL); + + for (int i = 0; i < iamf->nb_mix_presentations; i++) { + if (stg->id == iamf->mix_presentations[i]->mix_presentation_id) { + av_log(log_ctx, AV_LOG_ERROR, "Duplicate Mix Presentation id %"PRId64"\n", stg->id); + return AVERROR(EINVAL); + } + } + + mix_presentation = av_mallocz(sizeof(*mix_presentation)); + if (!mix_presentation) + return AVERROR(ENOMEM); + + mix_presentation->mix = stg->params.iamf_mix_presentation; + mix_presentation->mix_presentation_id = stg->id; + + for (int i = 0; i < mix_presentation->mix->nb_submixes; i++) { + const AVIAMFSubmix *submix = mix_presentation->mix->submixes[i]; + AVIAMFParamDefinition *param = submix->output_mix_config; + IAMFParamDefinition *param_definition; + + if (!param) { + av_log(log_ctx, AV_LOG_ERROR, "output_mix_config is not present in submix %u from " + "Mix Presentation ID %"PRId64"\n", i, stg->id); + return AVERROR(EINVAL); + } + + param_definition = ff_iamf_get_param_definition(iamf, param->parameter_id); + if (!param_definition) { + param_definition = add_param_definition(iamf, param, NULL, log_ctx); + if (!param_definition) + return AVERROR(ENOMEM); + } + + for (int j = 0; j < submix->nb_elements; j++) { + const AVIAMFSubmixElement *element = submix->elements[j]; + param = element->element_mix_config; + + if (!param) { + av_log(log_ctx, AV_LOG_ERROR, "element_mix_config is not present for element %u in submix %u from " + "Mix Presentation ID %"PRId64"\n", j, i, stg->id); + return AVERROR(EINVAL); + } + param_definition = ff_iamf_get_param_definition(iamf, param->parameter_id); + if (!param_definition) { + param_definition = add_param_definition(iamf, param, NULL, log_ctx); + if (!param_definition) + return AVERROR(ENOMEM); + } + } + } + + tmp = av_realloc_array(iamf->mix_presentations, iamf->nb_mix_presentations + 1, sizeof(*iamf->mix_presentations)); + if (!tmp) + return AVERROR(ENOMEM); + + iamf->mix_presentations = tmp; + iamf->mix_presentations[iamf->nb_mix_presentations++] = mix_presentation; + + return 0; +} + +static int iamf_write_codec_config(const IAMFContext *iamf, + const IAMFCodecConfig *codec_config, + AVIOContext *pb) +{ + uint8_t header[MAX_IAMF_OBU_HEADER_SIZE]; + AVIOContext *dyn_bc; + uint8_t *dyn_buf = NULL; + PutBitContext pbc; + int dyn_size; + + int ret = avio_open_dyn_buf(&dyn_bc); + if (ret < 0) + return ret; + + ffio_write_leb(dyn_bc, codec_config->codec_config_id); + avio_wl32(dyn_bc, codec_config->codec_tag); + + ffio_write_leb(dyn_bc, codec_config->nb_samples); + avio_wb16(dyn_bc, codec_config->seek_preroll); + + switch(codec_config->codec_id) { + case AV_CODEC_ID_OPUS: + avio_write(dyn_bc, codec_config->extradata, codec_config->extradata_size); + break; + case AV_CODEC_ID_AAC: + return AVERROR_PATCHWELCOME; + case AV_CODEC_ID_FLAC: + avio_w8(dyn_bc, 0x80); + avio_wb24(dyn_bc, codec_config->extradata_size); + avio_write(dyn_bc, codec_config->extradata, codec_config->extradata_size); + break; + case AV_CODEC_ID_PCM_S16LE: + avio_w8(dyn_bc, 0); + avio_w8(dyn_bc, 16); + avio_wb32(dyn_bc, codec_config->sample_rate); + break; + case AV_CODEC_ID_PCM_S24LE: + avio_w8(dyn_bc, 0); + avio_w8(dyn_bc, 24); + avio_wb32(dyn_bc, codec_config->sample_rate); + break; + case AV_CODEC_ID_PCM_S32LE: + avio_w8(dyn_bc, 0); + avio_w8(dyn_bc, 32); + avio_wb32(dyn_bc, codec_config->sample_rate); + break; + case AV_CODEC_ID_PCM_S16BE: + avio_w8(dyn_bc, 1); + avio_w8(dyn_bc, 16); + avio_wb32(dyn_bc, codec_config->sample_rate); + break; + case AV_CODEC_ID_PCM_S24BE: + avio_w8(dyn_bc, 1); + avio_w8(dyn_bc, 24); + avio_wb32(dyn_bc, codec_config->sample_rate); + break; + case AV_CODEC_ID_PCM_S32BE: + avio_w8(dyn_bc, 1); + avio_w8(dyn_bc, 32); + avio_wb32(dyn_bc, codec_config->sample_rate); + break; + default: + break; + } + + init_put_bits(&pbc, header, sizeof(header)); + put_bits(&pbc, 5, IAMF_OBU_IA_CODEC_CONFIG); + put_bits(&pbc, 3, 0); + flush_put_bits(&pbc); + + dyn_size = avio_close_dyn_buf(dyn_bc, &dyn_buf); + avio_write(pb, header, put_bytes_count(&pbc, 1)); + ffio_write_leb(pb, dyn_size); + avio_write(pb, dyn_buf, dyn_size); + av_free(dyn_buf); + + return 0; +} + +static inline int rescale_rational(AVRational q, int b) +{ + return av_clip_int16(av_rescale(q.num, b, q.den)); +} + +static int scalable_channel_layout_config(const IAMFAudioElement *audio_element, + AVIOContext *dyn_bc) +{ + const AVIAMFAudioElement *element = audio_element->element; + uint8_t header[MAX_IAMF_OBU_HEADER_SIZE]; + PutBitContext pb; + + init_put_bits(&pb, header, sizeof(header)); + put_bits(&pb, 3, element->nb_layers); + put_bits(&pb, 5, 0); + flush_put_bits(&pb); + avio_write(dyn_bc, header, put_bytes_count(&pb, 1)); + for (int i = 0; i < element->nb_layers; i++) { + AVIAMFLayer *layer = element->layers[i]; + int layout; + for (layout = 0; layout < FF_ARRAY_ELEMS(ff_iamf_scalable_ch_layouts); layout++) { + if (!av_channel_layout_compare(&layer->ch_layout, &ff_iamf_scalable_ch_layouts[layout])) + break; + } + init_put_bits(&pb, header, sizeof(header)); + put_bits(&pb, 4, layout); + put_bits(&pb, 1, !!layer->output_gain_flags); + put_bits(&pb, 1, !!(layer->flags & AV_IAMF_LAYER_FLAG_RECON_GAIN)); + put_bits(&pb, 2, 0); // reserved + put_bits(&pb, 8, audio_element->layers[i].substream_count); + put_bits(&pb, 8, audio_element->layers[i].coupled_substream_count); + if (layer->output_gain_flags) { + put_bits(&pb, 6, layer->output_gain_flags); + put_bits(&pb, 2, 0); + put_bits(&pb, 16, rescale_rational(layer->output_gain, 1 << 8)); + } + flush_put_bits(&pb); + avio_write(dyn_bc, header, put_bytes_count(&pb, 1)); + } + + return 0; +} + +static int ambisonics_config(const IAMFAudioElement *audio_element, + AVIOContext *dyn_bc) +{ + const AVIAMFAudioElement *element = audio_element->element; + AVIAMFLayer *layer = element->layers[0]; + + ffio_write_leb(dyn_bc, 0); // ambisonics_mode + ffio_write_leb(dyn_bc, layer->ch_layout.nb_channels); // output_channel_count + ffio_write_leb(dyn_bc, audio_element->nb_substreams); // substream_count + + if (layer->ch_layout.order == AV_CHANNEL_ORDER_AMBISONIC) + for (int i = 0; i < layer->ch_layout.nb_channels; i++) + avio_w8(dyn_bc, i); + else + for (int i = 0; i < layer->ch_layout.nb_channels; i++) + avio_w8(dyn_bc, layer->ch_layout.u.map[i].id); + + return 0; +} + +static int param_definition(const IAMFContext *iamf, + const IAMFParamDefinition *param_def, + AVIOContext *dyn_bc, void *log_ctx) +{ + const AVIAMFParamDefinition *param = param_def->param; + + ffio_write_leb(dyn_bc, param->parameter_id); + ffio_write_leb(dyn_bc, param->parameter_rate); + avio_w8(dyn_bc, param->duration ? 0 : 1 << 7); + if (param->duration) { + ffio_write_leb(dyn_bc, param->duration); + ffio_write_leb(dyn_bc, param->constant_subblock_duration); + if (param->constant_subblock_duration == 0) { + ffio_write_leb(dyn_bc, param->nb_subblocks); + for (int i = 0; i < param->nb_subblocks; i++) { + const void *subblock = av_iamf_param_definition_get_subblock(param, i); + + switch (param->type) { + case AV_IAMF_PARAMETER_DEFINITION_MIX_GAIN: { + const AVIAMFMixGain *mix = subblock; + ffio_write_leb(dyn_bc, mix->subblock_duration); + break; + } + case AV_IAMF_PARAMETER_DEFINITION_DEMIXING: { + const AVIAMFDemixingInfo *demix = subblock; + ffio_write_leb(dyn_bc, demix->subblock_duration); + break; + } + case AV_IAMF_PARAMETER_DEFINITION_RECON_GAIN: { + const AVIAMFReconGain *recon = subblock; + ffio_write_leb(dyn_bc, recon->subblock_duration); + break; + } + } + } + } + } + + return 0; +} + +static int iamf_write_audio_element(const IAMFContext *iamf, + const IAMFAudioElement *audio_element, + AVIOContext *pb, void *log_ctx) +{ + const AVIAMFAudioElement *element = audio_element->element; + const IAMFCodecConfig *codec_config = iamf->codec_configs[audio_element->codec_config_id]; + uint8_t header[MAX_IAMF_OBU_HEADER_SIZE]; + AVIOContext *dyn_bc; + uint8_t *dyn_buf = NULL; + PutBitContext pbc; + int param_definition_types = AV_IAMF_PARAMETER_DEFINITION_DEMIXING, dyn_size; + + int ret = avio_open_dyn_buf(&dyn_bc); + if (ret < 0) + return ret; + + ffio_write_leb(dyn_bc, audio_element->audio_element_id); + + init_put_bits(&pbc, header, sizeof(header)); + put_bits(&pbc, 3, element->audio_element_type); + put_bits(&pbc, 5, 0); + flush_put_bits(&pbc); + avio_write(dyn_bc, header, put_bytes_count(&pbc, 1)); + + ffio_write_leb(dyn_bc, audio_element->codec_config_id); + ffio_write_leb(dyn_bc, audio_element->nb_substreams); + + for (int i = 0; i < audio_element->nb_substreams; i++) + ffio_write_leb(dyn_bc, audio_element->substreams[i].audio_substream_id); + + if (element->nb_layers == 1) + param_definition_types &= ~AV_IAMF_PARAMETER_DEFINITION_DEMIXING; + if (element->nb_layers > 1) + param_definition_types |= AV_IAMF_PARAMETER_DEFINITION_RECON_GAIN; + if (codec_config->codec_tag == MKTAG('f','L','a','C') || + codec_config->codec_tag == MKTAG('i','p','c','m')) + param_definition_types &= ~AV_IAMF_PARAMETER_DEFINITION_RECON_GAIN; + + ffio_write_leb(dyn_bc, av_popcount(param_definition_types)); // num_parameters + + if (param_definition_types & 1) { + const AVIAMFParamDefinition *param = element->demixing_info; + const IAMFParamDefinition *param_def; + const AVIAMFDemixingInfo *demix; + + if (!param) { + av_log(log_ctx, AV_LOG_ERROR, "demixing_info needed but not set in Stream Group #%u\n", + audio_element->audio_element_id); + return AVERROR(EINVAL); + } + + demix = av_iamf_param_definition_get_subblock(param, 0); + ffio_write_leb(dyn_bc, AV_IAMF_PARAMETER_DEFINITION_DEMIXING); // type + + param_def = ff_iamf_get_param_definition(iamf, param->parameter_id); + ret = param_definition(iamf, param_def, dyn_bc, log_ctx); + if (ret < 0) + return ret; + + avio_w8(dyn_bc, demix->dmixp_mode << 5); // dmixp_mode + avio_w8(dyn_bc, element->default_w << 4); // default_w + } + if (param_definition_types & 2) { + const AVIAMFParamDefinition *param = element->recon_gain_info; + const IAMFParamDefinition *param_def; + + if (!param) { + av_log(log_ctx, AV_LOG_ERROR, "recon_gain_info needed but not set in Stream Group #%u\n", + audio_element->audio_element_id); + return AVERROR(EINVAL); + } + ffio_write_leb(dyn_bc, AV_IAMF_PARAMETER_DEFINITION_RECON_GAIN); // type + + param_def = ff_iamf_get_param_definition(iamf, param->parameter_id); + ret = param_definition(iamf, param_def, dyn_bc, log_ctx); + if (ret < 0) + return ret; + } + + if (element->audio_element_type == AV_IAMF_AUDIO_ELEMENT_TYPE_CHANNEL) { + ret = scalable_channel_layout_config(audio_element, dyn_bc); + if (ret < 0) + return ret; + } else { + ret = ambisonics_config(audio_element, dyn_bc); + if (ret < 0) + return ret; + } + + init_put_bits(&pbc, header, sizeof(header)); + put_bits(&pbc, 5, IAMF_OBU_IA_AUDIO_ELEMENT); + put_bits(&pbc, 3, 0); + flush_put_bits(&pbc); + + dyn_size = avio_close_dyn_buf(dyn_bc, &dyn_buf); + avio_write(pb, header, put_bytes_count(&pbc, 1)); + ffio_write_leb(pb, dyn_size); + avio_write(pb, dyn_buf, dyn_size); + av_free(dyn_buf); + + return 0; +} + +static int iamf_write_mixing_presentation(const IAMFContext *iamf, + const IAMFMixPresentation *mix_presentation, + AVIOContext *pb, void *log_ctx) +{ + uint8_t header[MAX_IAMF_OBU_HEADER_SIZE]; + const AVIAMFMixPresentation *mix = mix_presentation->mix; + const AVDictionaryEntry *tag = NULL; + PutBitContext pbc; + AVIOContext *dyn_bc; + uint8_t *dyn_buf = NULL; + int dyn_size; + + int ret = avio_open_dyn_buf(&dyn_bc); + if (ret < 0) + return ret; + + ffio_write_leb(dyn_bc, mix_presentation->mix_presentation_id); // mix_presentation_id + ffio_write_leb(dyn_bc, av_dict_count(mix->annotations)); // count_label + + while ((tag = av_dict_iterate(mix->annotations, tag))) + avio_put_str(dyn_bc, tag->key); + while ((tag = av_dict_iterate(mix->annotations, tag))) + avio_put_str(dyn_bc, tag->value); + + ffio_write_leb(dyn_bc, mix->nb_submixes); + for (int i = 0; i < mix->nb_submixes; i++) { + const AVIAMFSubmix *sub_mix = mix->submixes[i]; + const IAMFParamDefinition *param_def; + + ffio_write_leb(dyn_bc, sub_mix->nb_elements); + for (int j = 0; j < sub_mix->nb_elements; j++) { + const IAMFAudioElement *audio_element = NULL; + const AVIAMFSubmixElement *submix_element = sub_mix->elements[j]; + + for (int k = 0; k < iamf->nb_audio_elements; k++) + if (iamf->audio_elements[k]->audio_element_id == submix_element->audio_element_id) { + audio_element = iamf->audio_elements[k]; + break; + } + + av_assert0(audio_element); + ffio_write_leb(dyn_bc, submix_element->audio_element_id); + + if (av_dict_count(submix_element->annotations) != av_dict_count(mix->annotations)) { + av_log(log_ctx, AV_LOG_ERROR, "Inconsistent amount of labels in submix %d from Mix Presentation id #%u\n", + j, audio_element->audio_element_id); + return AVERROR(EINVAL); + } + while ((tag = av_dict_iterate(submix_element->annotations, tag))) + avio_put_str(dyn_bc, tag->value); + + init_put_bits(&pbc, header, sizeof(header)); + put_bits(&pbc, 2, submix_element->headphones_rendering_mode); + put_bits(&pbc, 6, 0); // reserved + flush_put_bits(&pbc); + avio_write(dyn_bc, header, put_bytes_count(&pbc, 1)); + ffio_write_leb(dyn_bc, 0); // rendering_config_extension_size + + param_def = ff_iamf_get_param_definition(iamf, submix_element->element_mix_config->parameter_id); + ret = param_definition(iamf, param_def, dyn_bc, log_ctx); + if (ret < 0) + return ret; + + avio_wb16(dyn_bc, rescale_rational(submix_element->default_mix_gain, 1 << 8)); + } + + param_def = ff_iamf_get_param_definition(iamf, sub_mix->output_mix_config->parameter_id); + ret = param_definition(iamf, param_def, dyn_bc, log_ctx); + if (ret < 0) + return ret; + avio_wb16(dyn_bc, rescale_rational(sub_mix->default_mix_gain, 1 << 8)); + + ffio_write_leb(dyn_bc, sub_mix->nb_layouts); // nb_layouts + for (int i = 0; i < sub_mix->nb_layouts; i++) { + const AVIAMFSubmixLayout *submix_layout = sub_mix->layouts[i]; + int layout, info_type; + int dialogue = submix_layout->dialogue_anchored_loudness.num && + submix_layout->dialogue_anchored_loudness.den; + int album = submix_layout->album_anchored_loudness.num && + submix_layout->album_anchored_loudness.den; + + if (layout == FF_ARRAY_ELEMS(ff_iamf_sound_system_map)) { + av_log(log_ctx, AV_LOG_ERROR, "Invalid Sound System value in a submix\n"); + return AVERROR(EINVAL); + } + + if (submix_layout->layout_type == AV_IAMF_SUBMIX_LAYOUT_TYPE_LOUDSPEAKERS) { + for (layout = 0; layout < FF_ARRAY_ELEMS(ff_iamf_sound_system_map); layout++) { + if (!av_channel_layout_compare(&submix_layout->sound_system, &ff_iamf_sound_system_map[layout].layout)) + break; + } + if (layout == FF_ARRAY_ELEMS(ff_iamf_sound_system_map)) { + av_log(log_ctx, AV_LOG_ERROR, "Invalid Sound System value in a submix\n"); + return AVERROR(EINVAL); + } + } + init_put_bits(&pbc, header, sizeof(header)); + put_bits(&pbc, 2, submix_layout->layout_type); // layout_type + if (submix_layout->layout_type == AV_IAMF_SUBMIX_LAYOUT_TYPE_LOUDSPEAKERS) { + put_bits(&pbc, 4, ff_iamf_sound_system_map[layout].id); // sound_system + put_bits(&pbc, 2, 0); // reserved + } else + put_bits(&pbc, 6, 0); // reserved + flush_put_bits(&pbc); + avio_write(dyn_bc, header, put_bytes_count(&pbc, 1)); + + info_type = (submix_layout->true_peak.num && submix_layout->true_peak.den); + info_type |= (dialogue || album) << 1; + avio_w8(dyn_bc, info_type); + avio_wb16(dyn_bc, rescale_rational(submix_layout->integrated_loudness, 1 << 8)); + avio_wb16(dyn_bc, rescale_rational(submix_layout->digital_peak, 1 << 8)); + if (info_type & 1) + avio_wb16(dyn_bc, rescale_rational(submix_layout->true_peak, 1 << 8)); + if (info_type & 2) { + avio_w8(dyn_bc, dialogue + album); // num_anchored_loudness + if (dialogue) { + avio_w8(dyn_bc, IAMF_ANCHOR_ELEMENT_DIALOGUE); + avio_wb16(dyn_bc, rescale_rational(submix_layout->dialogue_anchored_loudness, 1 << 8)); + } + if (album) { + avio_w8(dyn_bc, IAMF_ANCHOR_ELEMENT_ALBUM); + avio_wb16(dyn_bc, rescale_rational(submix_layout->album_anchored_loudness, 1 << 8)); + } + } + } + } + + init_put_bits(&pbc, header, sizeof(header)); + put_bits(&pbc, 5, IAMF_OBU_IA_MIX_PRESENTATION); + put_bits(&pbc, 3, 0); + flush_put_bits(&pbc); + + dyn_size = avio_close_dyn_buf(dyn_bc, &dyn_buf); + avio_write(pb, header, put_bytes_count(&pbc, 1)); + ffio_write_leb(pb, dyn_size); + avio_write(pb, dyn_buf, dyn_size); + av_free(dyn_buf); + + return 0; +} + +int ff_iamf_write_descriptors(const IAMFContext *iamf, AVIOContext *pb, void *log_ctx) +{ + uint8_t header[MAX_IAMF_OBU_HEADER_SIZE]; + PutBitContext pbc; + AVIOContext *dyn_bc; + uint8_t *dyn_buf = NULL; + int dyn_size; + + int ret = avio_open_dyn_buf(&dyn_bc); + if (ret < 0) + return ret; + + // Sequence Header + init_put_bits(&pbc, header, sizeof(header)); + put_bits(&pbc, 5, IAMF_OBU_IA_SEQUENCE_HEADER); + put_bits(&pbc, 3, 0); + flush_put_bits(&pbc); + + avio_write(dyn_bc, header, put_bytes_count(&pbc, 1)); + ffio_write_leb(dyn_bc, 6); + avio_wb32(dyn_bc, MKBETAG('i','a','m','f')); + avio_w8(dyn_bc, iamf->nb_audio_elements > 1); // primary_profile + avio_w8(dyn_bc, iamf->nb_audio_elements > 1); // additional_profile + + dyn_size = avio_close_dyn_buf(dyn_bc, &dyn_buf); + avio_write(pb, dyn_buf, dyn_size); + av_free(dyn_buf); + + for (int i = 0; i < iamf->nb_codec_configs; i++) { + ret = iamf_write_codec_config(iamf, iamf->codec_configs[i], pb); + if (ret < 0) + return ret; + } + + for (int i = 0; i < iamf->nb_audio_elements; i++) { + ret = iamf_write_audio_element(iamf, iamf->audio_elements[i], pb, log_ctx); + if (ret < 0) + return ret; + } + + for (int i = 0; i < iamf->nb_mix_presentations; i++) { + ret = iamf_write_mixing_presentation(iamf, iamf->mix_presentations[i], pb, log_ctx); + if (ret < 0) + return ret; + } + + return 0; +} diff --git a/libavformat/iamf_writer.h b/libavformat/iamf_writer.h new file mode 100644 index 0000000000..93354670b8 --- /dev/null +++ b/libavformat/iamf_writer.h @@ -0,0 +1,51 @@ +/* + * Immersive Audio Model and Formats muxing helpers and structs + * Copyright (c) 2023 James Almer + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AVFORMAT_IAMF_WRITER_H +#define AVFORMAT_IAMF_WRITER_H + +#include + +#include "libavutil/common.h" +#include "avformat.h" +#include "avio.h" +#include "iamf.h" + +static inline IAMFParamDefinition *ff_iamf_get_param_definition(const IAMFContext *iamf, + unsigned int parameter_id) +{ + IAMFParamDefinition *param_definition = NULL; + + for (int i = 0; i < iamf->nb_param_definitions; i++) + if (iamf->param_definitions[i]->param->parameter_id == parameter_id) { + param_definition = iamf->param_definitions[i]; + break; + } + + return param_definition; +} + +int ff_iamf_add_audio_element(IAMFContext *iamf, const AVStreamGroup *stg, void *log_ctx); +int ff_iamf_add_mix_presentation(IAMFContext *iamf, const AVStreamGroup *stg, void *log_ctx); + +int ff_iamf_write_descriptors(const IAMFContext *iamf, AVIOContext *pb, void *log_ctx); + +#endif /* AVFORMAT_IAMF_WRITER_H */ diff --git a/libavformat/iamfenc.c b/libavformat/iamfenc.c new file mode 100644 index 0000000000..0a043ce3a0 --- /dev/null +++ b/libavformat/iamfenc.c @@ -0,0 +1,387 @@ +/* + * IAMF muxer + * Copyright (c) 2023 James Almer + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include + +#include "libavutil/avassert.h" +#include "libavutil/common.h" +#include "libavutil/iamf.h" +#include "libavcodec/get_bits.h" +#include "libavcodec/put_bits.h" +#include "avformat.h" +#include "avio_internal.h" +#include "iamf.h" +#include "iamf_writer.h" +#include "internal.h" +#include "mux.h" + +typedef struct IAMFMuxContext { + IAMFContext iamf; + + int first_stream_id; +} IAMFMuxContext; + +static int iamf_init(AVFormatContext *s) +{ + IAMFMuxContext *const c = s->priv_data; + IAMFContext *const iamf = &c->iamf; + int nb_audio_elements = 0, nb_mix_presentations = 0; + int ret; + + if (!s->nb_streams) { + av_log(s, AV_LOG_ERROR, "There must be at least one stream\n"); + return AVERROR(EINVAL); + } + + for (int i = 0; i < s->nb_streams; i++) { + if (s->streams[i]->codecpar->codec_type != AVMEDIA_TYPE_AUDIO || + (s->streams[i]->codecpar->codec_tag != MKTAG('m','p','4','a') && + s->streams[i]->codecpar->codec_tag != MKTAG('O','p','u','s') && + s->streams[i]->codecpar->codec_tag != MKTAG('f','L','a','C') && + s->streams[i]->codecpar->codec_tag != MKTAG('i','p','c','m'))) { + av_log(s, AV_LOG_ERROR, "Unsupported codec id %s\n", + avcodec_get_name(s->streams[i]->codecpar->codec_id)); + return AVERROR(EINVAL); + } + + if (s->streams[i]->codecpar->ch_layout.nb_channels > 2) { + av_log(s, AV_LOG_ERROR, "Unsupported channel layout on stream #%d\n", i); + return AVERROR(EINVAL); + } + + for (int j = 0; j < i; j++) { + if (s->streams[i]->id == s->streams[j]->id) { + av_log(s, AV_LOG_ERROR, "Duplicated stream id %d\n", s->streams[j]->id); + return AVERROR(EINVAL); + } + } + } + + if (!s->nb_stream_groups) { + av_log(s, AV_LOG_ERROR, "There must be at least two stream groups\n"); + return AVERROR(EINVAL); + } + + for (int i = 0; i < s->nb_stream_groups; i++) { + const AVStreamGroup *stg = s->stream_groups[i]; + + if (stg->type == AV_STREAM_GROUP_PARAMS_IAMF_AUDIO_ELEMENT) + nb_audio_elements++; + if (stg->type == AV_STREAM_GROUP_PARAMS_IAMF_MIX_PRESENTATION) + nb_mix_presentations++; + } + if ((nb_audio_elements < 1 && nb_audio_elements > 2) || nb_mix_presentations < 1) { + av_log(s, AV_LOG_ERROR, "There must be >= 1 and <= 2 IAMF_AUDIO_ELEMENT and at least " + "one IAMF_MIX_PRESENTATION stream groups\n"); + return AVERROR(EINVAL); + } + + for (int i = 0; i < s->nb_stream_groups; i++) { + const AVStreamGroup *stg = s->stream_groups[i]; + if (stg->type != AV_STREAM_GROUP_PARAMS_IAMF_AUDIO_ELEMENT) + continue; + + ret = ff_iamf_add_audio_element(iamf, stg, s); + if (ret < 0) + return ret; + } + + for (int i = 0; i < s->nb_stream_groups; i++) { + const AVStreamGroup *stg = s->stream_groups[i]; + if (stg->type != AV_STREAM_GROUP_PARAMS_IAMF_MIX_PRESENTATION) + continue; + + ret = ff_iamf_add_mix_presentation(iamf, stg, s); + if (ret < 0) + return ret; + } + + c->first_stream_id = s->streams[0]->id; + + return 0; +} + +static int iamf_write_header(AVFormatContext *s) +{ + IAMFMuxContext *const c = s->priv_data; + IAMFContext *const iamf = &c->iamf; + int ret; + + ret = ff_iamf_write_descriptors(iamf, s->pb, s); + if (ret < 0) + return ret; + + c->first_stream_id = s->streams[0]->id; + + return 0; +} + +static inline int rescale_rational(AVRational q, int b) +{ + return av_clip_int16(av_rescale(q.num, b, q.den)); +} + +static int write_parameter_block(AVFormatContext *s, const AVIAMFParamDefinition *param) +{ + const IAMFMuxContext *const c = s->priv_data; + const IAMFContext *const iamf = &c->iamf; + uint8_t header[MAX_IAMF_OBU_HEADER_SIZE]; + IAMFParamDefinition *param_definition = ff_iamf_get_param_definition(iamf, param->parameter_id); + PutBitContext pb; + AVIOContext *dyn_bc; + uint8_t *dyn_buf = NULL; + int dyn_size, ret; + + if (param->type > AV_IAMF_PARAMETER_DEFINITION_RECON_GAIN) { + av_log(s, AV_LOG_DEBUG, "Ignoring side data with unknown type %u\n", + param->type); + return 0; + } + + if (!param_definition) { + av_log(s, AV_LOG_ERROR, "Non-existent Parameter Definition with ID %u referenced by a packet\n", + param->parameter_id); + return AVERROR(EINVAL); + } + + if (param->type != param_definition->param->type) { + av_log(s, AV_LOG_ERROR, "Inconsistent values for Parameter Definition " + "with ID %u in a packet\n", + param->parameter_id); + return AVERROR(EINVAL); + } + + ret = avio_open_dyn_buf(&dyn_bc); + if (ret < 0) + return ret; + + // Sequence Header + init_put_bits(&pb, header, sizeof(header)); + put_bits(&pb, 5, IAMF_OBU_IA_PARAMETER_BLOCK); + put_bits(&pb, 3, 0); + flush_put_bits(&pb); + avio_write(s->pb, header, put_bytes_count(&pb, 1)); + + ffio_write_leb(dyn_bc, param->parameter_id); + if (!param_definition->mode) { + ffio_write_leb(dyn_bc, param->duration); + ffio_write_leb(dyn_bc, param->constant_subblock_duration); + if (param->constant_subblock_duration == 0) + ffio_write_leb(dyn_bc, param->nb_subblocks); + } + + for (int i = 0; i < param->nb_subblocks; i++) { + const void *subblock = av_iamf_param_definition_get_subblock(param, i); + + switch (param->type) { + case AV_IAMF_PARAMETER_DEFINITION_MIX_GAIN: { + const AVIAMFMixGain *mix = subblock; + if (!param_definition->mode && param->constant_subblock_duration == 0) + ffio_write_leb(dyn_bc, mix->subblock_duration); + + ffio_write_leb(dyn_bc, mix->animation_type); + + avio_wb16(dyn_bc, rescale_rational(mix->start_point_value, 1 << 8)); + if (mix->animation_type >= AV_IAMF_ANIMATION_TYPE_LINEAR) + avio_wb16(dyn_bc, rescale_rational(mix->end_point_value, 1 << 8)); + if (mix->animation_type == AV_IAMF_ANIMATION_TYPE_BEZIER) { + avio_wb16(dyn_bc, rescale_rational(mix->control_point_value, 1 << 8)); + avio_w8(dyn_bc, av_clip_uint8(av_rescale(mix->control_point_relative_time.num, 1 << 8, + mix->control_point_relative_time.den))); + } + break; + } + case AV_IAMF_PARAMETER_DEFINITION_DEMIXING: { + const AVIAMFDemixingInfo *demix = subblock; + if (!param_definition->mode && param->constant_subblock_duration == 0) + ffio_write_leb(dyn_bc, demix->subblock_duration); + + avio_w8(dyn_bc, demix->dmixp_mode << 5); + break; + } + case AV_IAMF_PARAMETER_DEFINITION_RECON_GAIN: { + const AVIAMFReconGain *recon = subblock; + const AVIAMFAudioElement *audio_element = param_definition->audio_element->element; + + if (!param_definition->mode && param->constant_subblock_duration == 0) + ffio_write_leb(dyn_bc, recon->subblock_duration); + + if (!audio_element) { + av_log(s, AV_LOG_ERROR, "Invalid Parameter Definition with ID %u referenced by a packet\n", param->parameter_id); + return AVERROR(EINVAL); + } + + for (int j = 0; j < audio_element->nb_layers; j++) { + const AVIAMFLayer *layer = audio_element->layers[j]; + + if (layer->flags & AV_IAMF_LAYER_FLAG_RECON_GAIN) { + unsigned int recon_gain_flags = 0; + int k = 0; + + for (; k < 7; k++) + recon_gain_flags |= (1 << k) * !!recon->recon_gain[j][k]; + for (; k < 12; k++) + recon_gain_flags |= (2 << k) * !!recon->recon_gain[j][k]; + if (recon_gain_flags >> 8) + recon_gain_flags |= (1 << k); + + ffio_write_leb(dyn_bc, recon_gain_flags); + for (k = 0; k < 12; k++) { + if (recon->recon_gain[j][k]) + avio_w8(dyn_bc, recon->recon_gain[j][k]); + } + } + } + break; + } + default: + av_assert0(0); + } + } + + dyn_size = avio_close_dyn_buf(dyn_bc, &dyn_buf); + ffio_write_leb(s->pb, dyn_size); + avio_write(s->pb, dyn_buf, dyn_size); + av_free(dyn_buf); + + return 0; +} + +static int iamf_write_packet(AVFormatContext *s, AVPacket *pkt) +{ + const IAMFMuxContext *const c = s->priv_data; + AVStream *st = s->streams[pkt->stream_index]; + uint8_t header[MAX_IAMF_OBU_HEADER_SIZE]; + PutBitContext pb; + AVIOContext *dyn_bc; + uint8_t *side_data, *dyn_buf = NULL; + unsigned int skip_samples = 0, discard_padding = 0; + size_t side_data_size; + int dyn_size, type = st->id <= 17 ? st->id + IAMF_OBU_IA_AUDIO_FRAME_ID0 : IAMF_OBU_IA_AUDIO_FRAME; + int ret; + + if (s->nb_stream_groups && st->id == c->first_stream_id) { + AVIAMFParamDefinition *mix = + (AVIAMFParamDefinition *)av_packet_get_side_data(pkt, AV_PKT_DATA_IAMF_MIX_GAIN_PARAM, NULL); + AVIAMFParamDefinition *demix = + (AVIAMFParamDefinition *)av_packet_get_side_data(pkt, AV_PKT_DATA_IAMF_DEMIXING_INFO_PARAM, NULL); + AVIAMFParamDefinition *recon = + (AVIAMFParamDefinition *)av_packet_get_side_data(pkt, AV_PKT_DATA_IAMF_RECON_GAIN_INFO_PARAM, NULL); + + if (mix) { + ret = write_parameter_block(s, mix); + if (ret < 0) + return ret; + } + if (demix) { + ret = write_parameter_block(s, demix); + if (ret < 0) + return ret; + } + if (recon) { + ret = write_parameter_block(s, recon); + if (ret < 0) + return ret; + } + } + side_data = av_packet_get_side_data(pkt, AV_PKT_DATA_SKIP_SAMPLES, + &side_data_size); + + if (side_data && side_data_size >= 10) { + skip_samples = AV_RL32(side_data); + discard_padding = AV_RL32(side_data + 4); + } + + ret = avio_open_dyn_buf(&dyn_bc); + if (ret < 0) + return ret; + + init_put_bits(&pb, header, sizeof(header)); + put_bits(&pb, 5, type); + put_bits(&pb, 1, 0); // obu_redundant_copy + put_bits(&pb, 1, skip_samples || discard_padding); + put_bits(&pb, 1, 0); // obu_extension_flag + flush_put_bits(&pb); + avio_write(s->pb, header, put_bytes_count(&pb, 1)); + + if (skip_samples || discard_padding) { + ffio_write_leb(dyn_bc, discard_padding); + ffio_write_leb(dyn_bc, skip_samples); + } + + if (st->id > 17) + ffio_write_leb(dyn_bc, st->id); + + dyn_size = avio_close_dyn_buf(dyn_bc, &dyn_buf); + ffio_write_leb(s->pb, dyn_size + pkt->size); + avio_write(s->pb, dyn_buf, dyn_size); + av_free(dyn_buf); + avio_write(s->pb, pkt->data, pkt->size); + + return 0; +} + +static void iamf_deinit(AVFormatContext *s) +{ + IAMFMuxContext *const c = s->priv_data; + IAMFContext *const iamf = &c->iamf; + + for (int i = 0; i < iamf->nb_audio_elements; i++) { + IAMFAudioElement *audio_element = iamf->audio_elements[i]; + audio_element->element = NULL; + } + + for (int i = 0; i < iamf->nb_mix_presentations; i++) { + IAMFMixPresentation *mix_presentation = iamf->mix_presentations[i]; + mix_presentation->mix = NULL; + } + + ff_iamf_uninit_context(iamf); + + return; +} + +static const AVCodecTag iamf_codec_tags[] = { + { AV_CODEC_ID_AAC, MKTAG('m','p','4','a') }, + { AV_CODEC_ID_FLAC, MKTAG('f','L','a','C') }, + { AV_CODEC_ID_OPUS, MKTAG('O','p','u','s') }, + { AV_CODEC_ID_PCM_S16LE, MKTAG('i','p','c','m') }, + { AV_CODEC_ID_PCM_S16BE, MKTAG('i','p','c','m') }, + { AV_CODEC_ID_PCM_S24LE, MKTAG('i','p','c','m') }, + { AV_CODEC_ID_PCM_S24BE, MKTAG('i','p','c','m') }, + { AV_CODEC_ID_PCM_S32LE, MKTAG('i','p','c','m') }, + { AV_CODEC_ID_PCM_S32BE, MKTAG('i','p','c','m') }, + { AV_CODEC_ID_NONE, MKTAG('i','p','c','m') } +}; + +const FFOutputFormat ff_iamf_muxer = { + .p.name = "iamf", + .p.long_name = NULL_IF_CONFIG_SMALL("Raw Immersive Audio Model and Formats"), + .p.extensions = "iamf", + .priv_data_size = sizeof(IAMFMuxContext), + .p.audio_codec = AV_CODEC_ID_OPUS, + .init = iamf_init, + .deinit = iamf_deinit, + .write_header = iamf_write_header, + .write_packet = iamf_write_packet, + .p.codec_tag = (const AVCodecTag* const []){ iamf_codec_tags, NULL }, + .p.flags = AVFMT_GLOBALHEADER | AVFMT_NOTIMESTAMPS, +}; diff --git a/libavformat/version.h b/libavformat/version.h index e063e12b98..de9cc8e31d 100644 --- a/libavformat/version.h +++ b/libavformat/version.h @@ -31,7 +31,7 @@ #include "version_major.h" -#define LIBAVFORMAT_VERSION_MINOR 19 +#define LIBAVFORMAT_VERSION_MINOR 20 #define LIBAVFORMAT_VERSION_MICRO 100 #define LIBAVFORMAT_VERSION_INT AV_VERSION_INT(LIBAVFORMAT_VERSION_MAJOR, \