1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-23 12:43:46 +02:00

avformat/mov: add support for Immersive Audio Model and Formats in ISOBMFF

Signed-off-by: James Almer <jamrial@gmail.com>
This commit is contained in:
James Almer 2024-02-15 19:38:38 -03:00
parent 9ba327e70f
commit fe637161db
3 changed files with 290 additions and 17 deletions

2
configure vendored
View File

@ -3552,7 +3552,7 @@ matroska_demuxer_suggest="bzlib zlib"
matroska_muxer_select="mpeg4audio riffenc aac_adtstoasc_bsf pgs_frame_merge_bsf vp9_superframe_bsf"
mlp_demuxer_select="mlp_parser"
mmf_muxer_select="riffenc"
mov_demuxer_select="iso_media riffdec"
mov_demuxer_select="iso_media riffdec iamfdec"
mov_demuxer_suggest="zlib"
mov_muxer_select="iso_media riffenc rtpenc_chain vp9_superframe_bsf aac_adtstoasc_bsf ac3_parser"
mp3_demuxer_select="mpegaudio_parser"

View File

@ -168,6 +168,7 @@ typedef struct MOVStreamContext {
AVIOContext *pb;
int refcount;
int pb_is_copied;
int id; ///< AVStream id
int ffindex; ///< AVStream index
int next_chunk;
unsigned int chunk_count;
@ -264,6 +265,8 @@ typedef struct MOVStreamContext {
AVEncryptionInfo *default_encrypted_sample;
MOVEncryptionIndex *encryption_index;
} cenc;
struct IAMFDemuxContext *iamf;
} MOVStreamContext;
typedef struct HEIFItem {

View File

@ -58,6 +58,8 @@
#include "internal.h"
#include "avio_internal.h"
#include "demux.h"
#include "iamf_parse.h"
#include "iamf_reader.h"
#include "dovi_isom.h"
#include "riff.h"
#include "isom.h"
@ -212,6 +214,7 @@ static int mov_read_covr(MOVContext *c, AVIOContext *pb, int type, int len)
}
st = c->fc->streams[c->fc->nb_streams - 1];
st->priv_data = sc;
sc->id = st->id;
sc->refcount = 1;
if (st->attached_pic.size >= 8 && id != AV_CODEC_ID_BMP) {
@ -836,6 +839,183 @@ static int mov_read_dac3(MOVContext *c, AVIOContext *pb, MOVAtom atom)
return 0;
}
static int mov_read_iacb(MOVContext *c, AVIOContext *pb, MOVAtom atom)
{
AVStream *st;
MOVStreamContext *sc;
FFIOContext b;
AVIOContext *descriptor_pb;
AVDictionary *metadata;
IAMFContext *iamf;
int64_t start_time, duration;
unsigned descriptors_size;
int nb_frames, disposition;
int version, ret;
if (atom.size < 5)
return AVERROR_INVALIDDATA;
if (c->fc->nb_streams < 1)
return 0;
version = avio_r8(pb);
if (version != 1) {
av_log(c->fc, AV_LOG_ERROR, "%s configurationVersion %d",
version < 1 ? "invalid" : "unsupported", version);
return AVERROR_INVALIDDATA;
}
descriptors_size = ffio_read_leb(pb);
if (!descriptors_size || descriptors_size > INT_MAX)
return AVERROR_INVALIDDATA;
st = c->fc->streams[c->fc->nb_streams - 1];
sc = st->priv_data;
sc->iamf = av_mallocz(sizeof(*sc->iamf));
if (!sc->iamf)
return AVERROR(ENOMEM);
iamf = &sc->iamf->iamf;
st->codecpar->extradata = av_malloc(descriptors_size);
if (!st->codecpar->extradata)
return AVERROR(ENOMEM);
st->codecpar->extradata_size = descriptors_size;
ret = avio_read(pb, st->codecpar->extradata, descriptors_size);
if (ret != descriptors_size)
return ret < 0 ? ret : AVERROR_INVALIDDATA;
ffio_init_read_context(&b, st->codecpar->extradata, descriptors_size);
descriptor_pb = &b.pub;
ret = ff_iamfdec_read_descriptors(iamf, descriptor_pb, descriptors_size, c->fc);
if (ret < 0)
return ret;
metadata = st->metadata;
st->metadata = NULL;
start_time = st->start_time;
nb_frames = st->nb_frames;
duration = st->duration;
disposition = st->disposition;
for (int i = 0; i < iamf->nb_audio_elements; i++) {
IAMFAudioElement *audio_element = iamf->audio_elements[i];
AVStreamGroup *stg =
avformat_stream_group_create(c->fc, AV_STREAM_GROUP_PARAMS_IAMF_AUDIO_ELEMENT, NULL);
if (!stg) {
ret = AVERROR(ENOMEM);
goto fail;
}
av_iamf_audio_element_free(&stg->params.iamf_audio_element);
stg->id = audio_element->audio_element_id;
/* Transfer ownership */
stg->params.iamf_audio_element = audio_element->element;
audio_element->element = NULL;
for (int j = 0; j < audio_element->nb_substreams; j++) {
IAMFSubStream *substream = &audio_element->substreams[j];
AVStream *stream;
if (!i && !j) {
if (audio_element->layers[0].substream_count != 1)
st->disposition &= ~AV_DISPOSITION_DEFAULT;
stream = st;
} else
stream = avformat_new_stream(c->fc, NULL);
if (!stream) {
ret = AVERROR(ENOMEM);
goto fail;
}
stream->start_time = start_time;
stream->nb_frames = nb_frames;
stream->duration = duration;
stream->disposition = disposition;
if (stream != st) {
stream->priv_data = sc;
sc->refcount++;
}
if (i || j) {
stream->disposition |= AV_DISPOSITION_DEPENDENT;
if (audio_element->layers[0].substream_count == 1)
stream->disposition &= ~AV_DISPOSITION_DEFAULT;
}
ret = avcodec_parameters_copy(stream->codecpar, substream->codecpar);
if (ret < 0)
goto fail;
stream->id = substream->audio_substream_id;
avpriv_set_pts_info(st, 64, 1, sc->time_scale);
ret = avformat_stream_group_add_stream(stg, stream);
if (ret < 0)
goto fail;
}
ret = av_dict_copy(&stg->metadata, metadata, 0);
if (ret < 0)
goto fail;
}
for (int i = 0; i < iamf->nb_mix_presentations; i++) {
IAMFMixPresentation *mix_presentation = iamf->mix_presentations[i];
const AVIAMFMixPresentation *mix = mix_presentation->cmix;
AVStreamGroup *stg =
avformat_stream_group_create(c->fc, AV_STREAM_GROUP_PARAMS_IAMF_MIX_PRESENTATION, NULL);
if (!stg) {
ret = AVERROR(ENOMEM);
goto fail;
}
av_iamf_mix_presentation_free(&stg->params.iamf_mix_presentation);
stg->id = mix_presentation->mix_presentation_id;
/* Transfer ownership */
stg->params.iamf_mix_presentation = mix_presentation->mix;
mix_presentation->mix = NULL;
for (int j = 0; j < mix->nb_submixes; j++) {
const AVIAMFSubmix *submix = mix->submixes[j];
for (int k = 0; k < submix->nb_elements; k++) {
const AVIAMFSubmixElement *submix_element = submix->elements[k];
const AVStreamGroup *audio_element = NULL;
for (int l = 0; l < c->fc->nb_stream_groups; l++)
if (c->fc->stream_groups[l]->type == AV_STREAM_GROUP_PARAMS_IAMF_AUDIO_ELEMENT &&
c->fc->stream_groups[l]->id == submix_element->audio_element_id) {
audio_element = c->fc->stream_groups[l];
break;
}
av_assert0(audio_element);
for (int l = 0; l < audio_element->nb_streams; l++) {
ret = avformat_stream_group_add_stream(stg, audio_element->streams[l]);
if (ret < 0 && ret != AVERROR(EEXIST))
goto fail;
}
}
}
ret = av_dict_copy(&stg->metadata, metadata, 0);
if (ret < 0)
goto fail;
}
ret = 0;
fail:
av_dict_free(&metadata);
return ret;
}
static int mov_read_dec3(MOVContext *c, AVIOContext *pb, MOVAtom atom)
{
AVStream *st;
@ -1331,7 +1511,7 @@ static int64_t get_frag_time(AVFormatContext *s, AVStream *dst_st,
// If the stream is referenced by any sidx, limit the search
// to fragments that referenced this stream in the sidx
if (sc->has_sidx) {
frag_stream_info = get_frag_stream_info(frag_index, index, dst_st->id);
frag_stream_info = get_frag_stream_info(frag_index, index, sc->id);
if (frag_stream_info->sidx_pts != AV_NOPTS_VALUE)
return frag_stream_info->sidx_pts;
if (frag_stream_info->first_tfra_pts != AV_NOPTS_VALUE)
@ -1342,9 +1522,11 @@ static int64_t get_frag_time(AVFormatContext *s, AVStream *dst_st,
for (i = 0; i < frag_index->item[index].nb_stream_info; i++) {
AVStream *frag_stream = NULL;
frag_stream_info = &frag_index->item[index].stream_info[i];
for (j = 0; j < s->nb_streams; j++)
if (s->streams[j]->id == frag_stream_info->id)
for (j = 0; j < s->nb_streams; j++) {
MOVStreamContext *sc2 = s->streams[j]->priv_data;
if (sc2->id == frag_stream_info->id)
frag_stream = s->streams[j];
}
if (!frag_stream) {
av_log(s, AV_LOG_WARNING, "No stream matching sidx ID found.\n");
continue;
@ -1410,12 +1592,13 @@ static int update_frag_index(MOVContext *c, int64_t offset)
for (i = 0; i < c->fc->nb_streams; i++) {
// Avoid building frag index if streams lack track id.
if (c->fc->streams[i]->id < 0) {
MOVStreamContext *sc = c->fc->streams[i]->priv_data;
if (sc->id < 0) {
av_free(frag_stream_info);
return AVERROR_INVALIDDATA;
}
frag_stream_info[i].id = c->fc->streams[i]->id;
frag_stream_info[i].id = sc->id;
frag_stream_info[i].sidx_pts = AV_NOPTS_VALUE;
frag_stream_info[i].tfdt_dts = AV_NOPTS_VALUE;
frag_stream_info[i].next_trun_dts = AV_NOPTS_VALUE;
@ -3210,7 +3393,7 @@ static int mov_read_stts(MOVContext *c, AVIOContext *pb, MOVAtom atom)
"All samples in data stream index:id [%d:%d] have zero "
"duration, stream set to be discarded by default. Override "
"using AVStream->discard or -discard for ffmpeg command.\n",
st->index, st->id);
st->index, sc->id);
st->discard = AVDISCARD_ALL;
}
sc->track_end = duration;
@ -4590,6 +4773,50 @@ static void fix_timescale(MOVContext *c, MOVStreamContext *sc)
}
}
static int mov_update_iamf_streams(MOVContext *c, const AVStream *st)
{
const MOVStreamContext *sc = st->priv_data;
const IAMFContext *iamf = &sc->iamf->iamf;
for (int i = 0; i < iamf->nb_audio_elements; i++) {
const AVStreamGroup *stg = NULL;
for (int j = 0; j < c->fc->nb_stream_groups; j++)
if (c->fc->stream_groups[j]->id == iamf->audio_elements[i]->audio_element_id)
stg = c->fc->stream_groups[j];
av_assert0(stg);
for (int j = 0; j < stg->nb_streams; j++) {
const FFStream *sti = cffstream(st);
AVStream *out = stg->streams[j];
FFStream *out_sti = ffstream(stg->streams[j]);
out->codecpar->bit_rate = 0;
if (out == st)
continue;
out->time_base = st->time_base;
out->start_time = st->start_time;
out->duration = st->duration;
out->nb_frames = st->nb_frames;
out->discard = st->discard;
av_assert0(!out_sti->index_entries);
out_sti->index_entries = av_malloc(sti->index_entries_allocated_size);
if (!out_sti->index_entries)
return AVERROR(ENOMEM);
out_sti->index_entries_allocated_size = sti->index_entries_allocated_size;
out_sti->nb_index_entries = sti->nb_index_entries;
out_sti->skip_samples = sti->skip_samples;
memcpy(out_sti->index_entries, sti->index_entries, sti->index_entries_allocated_size);
}
}
return 0;
}
static int mov_read_trak(MOVContext *c, AVIOContext *pb, MOVAtom atom)
{
AVStream *st;
@ -4670,6 +4897,12 @@ static int mov_read_trak(MOVContext *c, AVIOContext *pb, MOVAtom atom)
mov_build_index(c, st);
if (sc->iamf) {
ret = mov_update_iamf_streams(c, st);
if (ret < 0)
return ret;
}
if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
MOVDref *dref = &sc->drefs[sc->dref_id - 1];
if (c->enable_drefs) {
@ -4902,6 +5135,7 @@ static int heif_add_stream(MOVContext *c, HEIFItem *item)
st->priv_data = sc;
st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
st->codecpar->codec_id = mov_codec_id(st, item->type);
sc->id = st->id;
sc->ffindex = st->index;
c->trak_index = st->index;
st->avg_frame_rate.num = st->avg_frame_rate.den = 1;
@ -5000,6 +5234,7 @@ static int mov_read_tkhd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
avio_rb32(pb); /* modification time */
}
st->id = (int)avio_rb32(pb); /* track id (NOT 0 !)*/
sc->id = st->id;
avio_rb32(pb); /* reserved */
/* highlevel (considering edits) duration in movie timebase */
@ -5174,7 +5409,8 @@ static int mov_read_tfdt(MOVContext *c, AVIOContext *pb, MOVAtom atom)
int64_t base_media_decode_time;
for (i = 0; i < c->fc->nb_streams; i++) {
if (c->fc->streams[i]->id == frag->track_id) {
sc = c->fc->streams[i]->priv_data;
if (sc->id == frag->track_id) {
st = c->fc->streams[i];
break;
}
@ -5227,7 +5463,8 @@ static int mov_read_trun(MOVContext *c, AVIOContext *pb, MOVAtom atom)
}
for (i = 0; i < c->fc->nb_streams; i++) {
if (c->fc->streams[i]->id == frag->track_id) {
sc = c->fc->streams[i]->priv_data;
if (sc->id == frag->track_id) {
st = c->fc->streams[i];
sti = ffstream(st);
break;
@ -5530,7 +5767,8 @@ static int mov_read_sidx(MOVContext *c, AVIOContext *pb, MOVAtom atom)
track_id = avio_rb32(pb); // Reference ID
for (i = 0; i < c->fc->nb_streams; i++) {
if (c->fc->streams[i]->id == track_id) {
sc = c->fc->streams[i]->priv_data;
if (sc->id == track_id) {
st = c->fc->streams[i];
break;
}
@ -6447,7 +6685,8 @@ static int get_current_encryption_info(MOVContext *c, MOVEncryptionIndex **encry
frag_stream_info = get_current_frag_stream_info(&c->frag_index);
if (frag_stream_info) {
for (i = 0; i < c->fc->nb_streams; i++) {
if (c->fc->streams[i]->id == frag_stream_info->id) {
*sc = c->fc->streams[i]->priv_data;
if ((*sc)->id == frag_stream_info->id) {
st = c->fc->streams[i];
break;
}
@ -7391,7 +7630,7 @@ static int cenc_filter(MOVContext *mov, AVStream* st, MOVStreamContext *sc, AVPa
AVEncryptionInfo *encrypted_sample;
int encrypted_index, ret;
frag_stream_info = get_frag_stream_info_from_pkt(&mov->frag_index, pkt, st->id);
frag_stream_info = get_frag_stream_info_from_pkt(&mov->frag_index, pkt, sc->id);
encrypted_index = current_index;
encryption_index = NULL;
if (frag_stream_info) {
@ -8194,6 +8433,7 @@ static const MOVParseTableEntry mov_default_parse_table[] = {
{ MKTAG('i','p','r','p'), mov_read_iprp },
{ MKTAG('i','i','n','f'), mov_read_iinf },
{ MKTAG('a','m','v','e'), mov_read_amve }, /* ambient viewing environment box */
{ MKTAG('i','a','c','b'), mov_read_iacb },
{ 0, NULL }
};
@ -8425,11 +8665,13 @@ static void mov_read_chapters(AVFormatContext *s)
AVStream *st = NULL;
FFStream *sti = NULL;
chapter_track = mov->chapter_tracks[j];
for (i = 0; i < s->nb_streams; i++)
if (s->streams[i]->id == chapter_track) {
for (i = 0; i < s->nb_streams; i++) {
sc = mov->fc->streams[i]->priv_data;
if (sc->id == chapter_track) {
st = s->streams[i];
break;
}
}
if (!st) {
av_log(s, AV_LOG_ERROR, "Referenced QT chapter track not found\n");
continue;
@ -8662,6 +8904,10 @@ static void mov_free_stream_context(AVFormatContext *s, AVStream *st)
av_freep(&sc->mastering);
av_freep(&sc->coll);
av_freep(&sc->ambient);
if (sc->iamf)
ff_iamf_read_deinit(sc->iamf);
av_freep(&sc->iamf);
}
static int mov_read_close(AVFormatContext *s)
@ -8916,9 +9162,11 @@ static int mov_read_header(AVFormatContext *s)
AVDictionaryEntry *tcr;
int tmcd_st_id = -1;
for (j = 0; j < s->nb_streams; j++)
if (s->streams[j]->id == sc->timecode_track)
for (j = 0; j < s->nb_streams; j++) {
MOVStreamContext *sc2 = s->streams[j]->priv_data;
if (sc2->id == sc->timecode_track)
tmcd_st_id = j;
}
if (tmcd_st_id < 0 || tmcd_st_id == i)
continue;
@ -9298,7 +9546,29 @@ static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
if (st->codecpar->codec_id == AV_CODEC_ID_EIA_608 && sample->size > 8)
ret = get_eia608_packet(sc->pb, pkt, sample->size);
else
else if (sc->iamf) {
int64_t pts, dts, pos, duration;
int flags, size = sample->size;
ret = mov_finalize_packet(s, st, sample, current_index, pkt);
pts = pkt->pts; dts = pkt->dts;
pos = pkt->pos; flags = pkt->flags;
duration = pkt->duration;
while (!ret && size > 0) {
ret = ff_iamf_read_packet(s, sc->iamf, sc->pb, size, pkt);
if (ret < 0) {
if (should_retry(sc->pb, ret))
mov_current_sample_dec(sc);
return ret;
}
size -= ret;
pkt->pts = pts; pkt->dts = dts;
pkt->pos = pos; pkt->flags |= flags;
pkt->duration = duration;
ret = ff_buffer_packet(s, pkt);
}
if (!ret)
return FFERROR_REDO;
} else
ret = av_get_packet(sc->pb, pkt, sample->size);
if (ret < 0) {
if (should_retry(sc->pb, ret)) {