2
* Microsoft XMV demuxer
3
* Copyright (c) 2011 Sven Hesse <drmccoy@drmccoy.de>
4
* Copyright (c) 2011 Matthew Hoops <clone2727@gmail.com>
6
* This file is part of Libav.
8
* Libav 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
* Libav 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 Libav; if not, write to the Free Software
20
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25
* Microsoft XMV demuxer
30
#include "libavutil/intreadwrite.h"
36
#define XMV_MIN_HEADER_SIZE 36
38
#define XMV_AUDIO_ADPCM51_FRONTLEFTRIGHT 1
39
#define XMV_AUDIO_ADPCM51_FRONTCENTERLOW 2
40
#define XMV_AUDIO_ADPCM51_REARLEFTRIGHT 4
42
#define XMV_AUDIO_ADPCM51 (XMV_AUDIO_ADPCM51_FRONTLEFTRIGHT | \
43
XMV_AUDIO_ADPCM51_FRONTCENTERLOW | \
44
XMV_AUDIO_ADPCM51_REARLEFTRIGHT)
46
typedef struct XMVAudioTrack {
50
uint16_t bits_per_sample;
54
uint16_t block_samples;
56
enum CodecID codec_id;
59
typedef struct XMVVideoPacket {
60
/* The decoder stream index for this video packet. */
66
uint32_t current_frame;
69
/* Does the video packet contain extra data? */
79
typedef struct XMVAudioPacket {
80
/* The decoder stream index for this audio packet. */
83
/* The audio track this packet encodes. */
94
typedef struct XMVDemuxContext {
95
uint16_t audio_track_count;
97
XMVAudioTrack *audio_tracks;
99
uint32_t this_packet_size;
100
uint32_t next_packet_size;
102
uint32_t this_packet_offset;
103
uint32_t next_packet_offset;
105
uint16_t current_stream;
106
uint16_t stream_count;
108
XMVVideoPacket video;
109
XMVAudioPacket *audio;
112
static int xmv_probe(AVProbeData *p)
114
uint32_t file_version;
116
if (p->buf_size < XMV_MIN_HEADER_SIZE)
119
file_version = AV_RL32(p->buf + 16);
120
if ((file_version == 0) || (file_version > 4))
123
if (!memcmp(p->buf + 12, "xobX", 4))
124
return AVPROBE_SCORE_MAX;
129
static int xmv_read_header(AVFormatContext *s,
130
AVFormatParameters *ap)
132
XMVDemuxContext *xmv = s->priv_data;
133
AVIOContext *pb = s->pb;
134
AVStream *vst = NULL;
136
uint32_t file_version;
137
uint32_t this_packet_size;
138
uint16_t audio_track;
140
avio_skip(pb, 4); /* Next packet size */
142
this_packet_size = avio_rl32(pb);
144
avio_skip(pb, 4); /* Max packet size */
145
avio_skip(pb, 4); /* "xobX" */
147
file_version = avio_rl32(pb);
148
if ((file_version != 4) && (file_version != 2))
149
av_log_ask_for_sample(s, "Found uncommon version %d\n", file_version);
154
vst = avformat_new_stream(s, NULL);
156
return AVERROR(ENOMEM);
158
avpriv_set_pts_info(vst, 32, 1, 1000);
160
vst->codec->codec_type = AVMEDIA_TYPE_VIDEO;
161
vst->codec->codec_id = CODEC_ID_WMV2;
162
vst->codec->codec_tag = MKBETAG('W', 'M', 'V', '2');
163
vst->codec->width = avio_rl32(pb);
164
vst->codec->height = avio_rl32(pb);
166
vst->duration = avio_rl32(pb);
168
xmv->video.stream_index = vst->index;
172
xmv->audio_track_count = avio_rl16(pb);
174
avio_skip(pb, 2); /* Unknown (padding?) */
176
xmv->audio_tracks = av_malloc(xmv->audio_track_count * sizeof(XMVAudioTrack));
177
if (!xmv->audio_tracks)
178
return AVERROR(ENOMEM);
180
xmv->audio = av_malloc(xmv->audio_track_count * sizeof(XMVAudioPacket));
182
return AVERROR(ENOMEM);
184
for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
185
XMVAudioTrack *track = &xmv->audio_tracks[audio_track];
186
XMVAudioPacket *packet = &xmv->audio [audio_track];
187
AVStream *ast = NULL;
189
track->compression = avio_rl16(pb);
190
track->channels = avio_rl16(pb);
191
track->sample_rate = avio_rl32(pb);
192
track->bits_per_sample = avio_rl16(pb);
193
track->flags = avio_rl16(pb);
195
track->bit_rate = track->bits_per_sample *
198
track->block_align = 36 * track->channels;
199
track->block_samples = 64;
200
track->codec_id = ff_wav_codec_get_id(track->compression,
201
track->bits_per_sample);
203
packet->track = track;
204
packet->stream_index = -1;
206
packet->frame_size = 0;
207
packet->block_count = 0;
209
/* TODO: ADPCM'd 5.1 sound is encoded in three separate streams.
210
* Those need to be interleaved to a proper 5.1 stream. */
211
if (track->flags & XMV_AUDIO_ADPCM51)
212
av_log(s, AV_LOG_WARNING, "Unsupported 5.1 ADPCM audio stream "
213
"(0x%04X)\n", track->flags);
215
ast = avformat_new_stream(s, NULL);
217
return AVERROR(ENOMEM);
219
ast->codec->codec_type = AVMEDIA_TYPE_AUDIO;
220
ast->codec->codec_id = track->codec_id;
221
ast->codec->codec_tag = track->compression;
222
ast->codec->channels = track->channels;
223
ast->codec->sample_rate = track->sample_rate;
224
ast->codec->bits_per_coded_sample = track->bits_per_sample;
225
ast->codec->bit_rate = track->bit_rate;
226
ast->codec->block_align = 36 * track->channels;
228
avpriv_set_pts_info(ast, 32, track->block_samples, track->sample_rate);
230
packet->stream_index = ast->index;
232
ast->duration = vst->duration;
236
/** Initialize the packet context */
238
xmv->next_packet_offset = avio_tell(pb);
239
xmv->next_packet_size = this_packet_size - xmv->next_packet_offset;
240
xmv->stream_count = xmv->audio_track_count + 1;
245
static void xmv_read_extradata(uint8_t *extradata, AVIOContext *pb)
247
/* Read the XMV extradata */
249
uint32_t data = avio_rl32(pb);
251
int mspel_bit = !!(data & 0x01);
252
int loop_filter = !!(data & 0x02);
253
int abt_flag = !!(data & 0x04);
254
int j_type_bit = !!(data & 0x08);
255
int top_left_mv_flag = !!(data & 0x10);
256
int per_mb_rl_bit = !!(data & 0x20);
257
int slice_count = (data >> 6) & 7;
259
/* Write it back as standard WMV2 extradata */
263
data |= mspel_bit << 15;
264
data |= loop_filter << 14;
265
data |= abt_flag << 13;
266
data |= j_type_bit << 12;
267
data |= top_left_mv_flag << 11;
268
data |= per_mb_rl_bit << 10;
269
data |= slice_count << 7;
271
AV_WB32(extradata, data);
274
static int xmv_process_packet_header(AVFormatContext *s)
276
XMVDemuxContext *xmv = s->priv_data;
277
AVIOContext *pb = s->pb;
280
uint16_t audio_track;
281
uint32_t data_offset;
283
/* Next packet size */
284
xmv->next_packet_size = avio_rl32(pb);
286
/* Packet video header */
288
if (avio_read(pb, data, 8) != 8)
291
xmv->video.data_size = AV_RL32(data) & 0x007FFFFF;
293
xmv->video.current_frame = 0;
294
xmv->video.frame_count = (AV_RL32(data) >> 23) & 0xFF;
296
xmv->video.has_extradata = (data[3] & 0x80) != 0;
298
/* Adding the audio data sizes and the video data size keeps you 4 bytes
299
* short for every audio track. But as playing around with XMV files with
300
* ADPCM audio showed, taking the extra 4 bytes from the audio data gives
301
* you either completely distorted audio or click (when skipping the
302
* remaining 68 bytes of the ADPCM block). Substracting 4 bytes for every
303
* audio track from the video data works at least for the audio. Probably
304
* some alignment thing?
305
* The video data has (always?) lots of padding, so it should work out...
307
xmv->video.data_size -= xmv->audio_track_count * 4;
309
xmv->current_stream = 0;
310
if (!xmv->video.frame_count) {
311
xmv->video.frame_count = 1;
312
xmv->current_stream = 1;
315
/* Packet audio header */
317
for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
318
XMVAudioPacket *packet = &xmv->audio[audio_track];
320
if (avio_read(pb, data, 4) != 4)
323
packet->data_size = AV_RL32(data) & 0x007FFFFF;
324
if ((packet->data_size == 0) && (audio_track != 0))
325
/* This happens when I create an XMV with several identical audio
326
* streams. From the size calculations, duplicating the previous
327
* stream's size works out, but the track data itself is silent.
328
* Maybe this should also redirect the offset to the previous track?
330
packet->data_size = xmv->audio[audio_track - 1].data_size;
332
/** Carve up the audio data in frame_count slices */
333
packet->frame_size = packet->data_size / xmv->video.frame_count;
334
packet->frame_size -= packet->frame_size % packet->track->block_align;
337
/* Packet data offsets */
339
data_offset = avio_tell(pb);
341
xmv->video.data_offset = data_offset;
342
data_offset += xmv->video.data_size;
344
for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
345
xmv->audio[audio_track].data_offset = data_offset;
346
data_offset += xmv->audio[audio_track].data_size;
349
/* Video frames header */
351
/* Read new video extra data */
352
if (xmv->video.data_size > 0) {
353
if (xmv->video.has_extradata) {
354
xmv_read_extradata(xmv->video.extradata, pb);
356
xmv->video.data_size -= 4;
357
xmv->video.data_offset += 4;
359
if (xmv->video.stream_index >= 0) {
360
AVStream *vst = s->streams[xmv->video.stream_index];
362
assert(xmv->video.stream_index < s->nb_streams);
364
if (vst->codec->extradata_size < 4) {
365
av_free(vst->codec->extradata);
367
vst->codec->extradata =
368
av_malloc(4 + FF_INPUT_BUFFER_PADDING_SIZE);
369
vst->codec->extradata_size = 4;
372
memcpy(vst->codec->extradata, xmv->video.extradata, 4);
380
static int xmv_fetch_new_packet(AVFormatContext *s)
382
XMVDemuxContext *xmv = s->priv_data;
383
AVIOContext *pb = s->pb;
387
xmv->this_packet_offset = xmv->next_packet_offset;
388
if (avio_seek(pb, xmv->this_packet_offset, SEEK_SET) != xmv->this_packet_offset)
391
/* Update the size */
392
xmv->this_packet_size = xmv->next_packet_size;
393
if (xmv->this_packet_size < (12 + xmv->audio_track_count * 4))
396
/* Process the header */
397
result = xmv_process_packet_header(s);
401
/* Update the offset */
402
xmv->next_packet_offset = xmv->this_packet_offset + xmv->this_packet_size;
407
static int xmv_fetch_audio_packet(AVFormatContext *s,
408
AVPacket *pkt, uint32_t stream)
410
XMVDemuxContext *xmv = s->priv_data;
411
AVIOContext *pb = s->pb;
412
XMVAudioPacket *audio = &xmv->audio[stream];
415
uint32_t block_count;
419
if (avio_seek(pb, audio->data_offset, SEEK_SET) != audio->data_offset)
422
if ((xmv->video.current_frame + 1) < xmv->video.frame_count)
423
/* Not the last frame, get at most frame_size bytes. */
424
data_size = FFMIN(audio->frame_size, audio->data_size);
426
/* Last frame, get the rest. */
427
data_size = audio->data_size;
429
/* Read the packet */
430
result = av_get_packet(pb, pkt, data_size);
434
pkt->stream_index = audio->stream_index;
436
/* Calculate the PTS */
438
block_count = data_size / audio->track->block_align;
440
pkt->duration = block_count;
441
pkt->pts = audio->block_count;
442
pkt->dts = AV_NOPTS_VALUE;
444
audio->block_count += block_count;
447
audio->data_size -= data_size;
448
audio->data_offset += data_size;
453
static int xmv_fetch_video_packet(AVFormatContext *s,
456
XMVDemuxContext *xmv = s->priv_data;
457
AVIOContext *pb = s->pb;
458
XMVVideoPacket *video = &xmv->video;
461
uint32_t frame_header;
462
uint32_t frame_size, frame_timestamp;
466
if (avio_seek(pb, video->data_offset, SEEK_SET) != video->data_offset)
469
/* Read the frame header */
470
frame_header = avio_rl32(pb);
472
frame_size = (frame_header & 0x1FFFF) * 4 + 4;
473
frame_timestamp = (frame_header >> 17);
475
if ((frame_size + 4) > video->data_size)
478
/* Create the packet */
479
result = av_new_packet(pkt, frame_size);
483
/* Contrary to normal WMV2 video, the bit stream in XMV's
484
* WMV2 is little-endian.
485
* TODO: This manual swap is of course suboptimal.
487
for (i = 0; i < frame_size; i += 4)
488
AV_WB32(pkt->data + i, avio_rl32(pb));
490
pkt->stream_index = video->stream_index;
492
/* Calculate the PTS */
494
video->last_pts = frame_timestamp + video->pts;
497
pkt->pts = video->last_pts;
498
pkt->dts = AV_NOPTS_VALUE;
500
video->pts += frame_timestamp;
503
pkt->flags = (pkt->data[0] & 0x80) ? 0 : AV_PKT_FLAG_KEY;
506
video->data_size -= frame_size + 4;
507
video->data_offset += frame_size + 4;
512
static int xmv_read_packet(AVFormatContext *s,
515
XMVDemuxContext *xmv = s->priv_data;
518
if (xmv->video.current_frame == xmv->video.frame_count) {
519
/* No frames left in this packet, so we fetch a new one */
521
result = xmv_fetch_new_packet(s);
526
if (xmv->current_stream == 0) {
527
/* Fetch a video frame */
529
result = xmv_fetch_video_packet(s, pkt);
534
/* Fetch an audio frame */
536
result = xmv_fetch_audio_packet(s, pkt, xmv->current_stream - 1);
541
/* Increase our counters */
542
if (++xmv->current_stream >= xmv->stream_count) {
543
xmv->current_stream = 0;
544
xmv->video.current_frame += 1;
550
static int xmv_read_close(AVFormatContext *s)
552
XMVDemuxContext *xmv = s->priv_data;
555
av_free(xmv->audio_tracks);
560
AVInputFormat ff_xmv_demuxer = {
562
.long_name = NULL_IF_CONFIG_SMALL("Microsoft XMV"),
563
.priv_data_size = sizeof(XMVDemuxContext),
564
.read_probe = xmv_probe,
565
.read_header = xmv_read_header,
566
.read_packet = xmv_read_packet,
567
.read_close = xmv_read_close,