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 FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 * Microsoft XMV demuxer
30 #include "libavutil/intreadwrite.h"
35 #include "libavutil/avassert.h"
37 /** The min size of an XMV header. */
38 #define XMV_MIN_HEADER_SIZE 36
40 /** Audio flag: ADPCM'd 5.1 stream, front left / right channels */
41 #define XMV_AUDIO_ADPCM51_FRONTLEFTRIGHT 1
42 /** Audio flag: ADPCM'd 5.1 stream, front center / low frequency channels */
43 #define XMV_AUDIO_ADPCM51_FRONTCENTERLOW 2
44 /** Audio flag: ADPCM'd 5.1 stream, rear left / right channels */
45 #define XMV_AUDIO_ADPCM51_REARLEFTRIGHT 4
47 /** Audio flag: Any of the ADPCM'd 5.1 stream flags. */
48 #define XMV_AUDIO_ADPCM51 (XMV_AUDIO_ADPCM51_FRONTLEFTRIGHT | \
49 XMV_AUDIO_ADPCM51_FRONTCENTERLOW | \
50 XMV_AUDIO_ADPCM51_REARLEFTRIGHT)
52 #define XMV_BLOCK_ALIGN_SIZE 36
54 /** A video packet with an XMV file. */
55 typedef struct XMVVideoPacket
{
56 int stream_index
; ///< The decoder stream index for this video packet.
58 uint32_t data_size
; ///< The size of the remaining video data.
59 uint64_t data_offset
; ///< The offset of the video data within the file.
61 uint32_t current_frame
; ///< The current frame within this video packet.
62 uint32_t frame_count
; ///< The amount of frames within this video packet.
64 int has_extradata
; ///< Does the video packet contain extra data?
65 uint8_t extradata
[4]; ///< The extra data
67 int64_t last_pts
; ///< PTS of the last video frame.
68 int64_t pts
; ///< PTS of the most current video frame.
71 /** An audio packet with an XMV file. */
72 typedef struct XMVAudioPacket
{
73 int stream_index
; ///< The decoder stream index for this audio packet.
75 /* Stream format properties. */
76 uint16_t compression
; ///< The type of compression.
77 uint16_t channels
; ///< Number of channels.
78 uint32_t sample_rate
; ///< Sampling rate.
79 uint16_t bits_per_sample
; ///< Bits per compressed sample.
80 uint32_t bit_rate
; ///< Bits of compressed data per second.
81 uint16_t flags
; ///< Flags
82 unsigned block_align
; ///< Bytes per compressed block.
83 uint16_t block_samples
; ///< Decompressed samples per compressed block.
85 enum AVCodecID codec_id
; ///< The codec ID of the compression scheme.
87 uint32_t data_size
; ///< The size of the remaining audio data.
88 uint64_t data_offset
; ///< The offset of the audio data within the file.
90 uint32_t frame_size
; ///< Number of bytes to put into an audio frame.
92 uint64_t block_count
; ///< Running counter of decompressed audio block.
95 /** Context for demuxing an XMV file. */
96 typedef struct XMVDemuxContext
{
97 uint16_t audio_track_count
; ///< Number of audio track in this file.
99 uint32_t this_packet_size
; ///< Size of the current packet.
100 uint32_t next_packet_size
; ///< Size of the next packet.
102 uint64_t this_packet_offset
; ///< Offset of the current packet.
103 uint64_t next_packet_offset
; ///< Offset of the next packet.
105 uint16_t current_stream
; ///< The index of the stream currently handling.
106 uint16_t stream_count
; ///< The number of streams in this file.
108 XMVVideoPacket video
; ///< The video packet contained in each packet.
109 XMVAudioPacket
*audio
; ///< The audio packets contained in each packet.
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_close(AVFormatContext
*s
)
131 XMVDemuxContext
*xmv
= s
->priv_data
;
133 av_freep(&xmv
->audio
);
138 static int xmv_read_header(AVFormatContext
*s
)
140 XMVDemuxContext
*xmv
= s
->priv_data
;
141 AVIOContext
*pb
= s
->pb
;
142 AVStream
*vst
= NULL
;
144 uint32_t file_version
;
145 uint32_t this_packet_size
;
146 uint16_t audio_track
;
149 avio_skip(pb
, 4); /* Next packet size */
151 this_packet_size
= avio_rl32(pb
);
153 avio_skip(pb
, 4); /* Max packet size */
154 avio_skip(pb
, 4); /* "xobX" */
156 file_version
= avio_rl32(pb
);
157 if ((file_version
!= 4) && (file_version
!= 2))
158 avpriv_request_sample(s
, "Uncommon version %"PRIu32
"", file_version
);
163 vst
= avformat_new_stream(s
, NULL
);
165 return AVERROR(ENOMEM
);
167 avpriv_set_pts_info(vst
, 32, 1, 1000);
169 vst
->codec
->codec_type
= AVMEDIA_TYPE_VIDEO
;
170 vst
->codec
->codec_id
= AV_CODEC_ID_WMV2
;
171 vst
->codec
->codec_tag
= MKBETAG('W', 'M', 'V', '2');
172 vst
->codec
->width
= avio_rl32(pb
);
173 vst
->codec
->height
= avio_rl32(pb
);
175 vst
->duration
= avio_rl32(pb
);
177 xmv
->video
.stream_index
= vst
->index
;
181 xmv
->audio_track_count
= avio_rl16(pb
);
183 avio_skip(pb
, 2); /* Unknown (padding?) */
185 xmv
->audio
= av_malloc_array(xmv
->audio_track_count
, sizeof(XMVAudioPacket
));
187 ret
= AVERROR(ENOMEM
);
191 for (audio_track
= 0; audio_track
< xmv
->audio_track_count
; audio_track
++) {
192 XMVAudioPacket
*packet
= &xmv
->audio
[audio_track
];
193 AVStream
*ast
= NULL
;
195 packet
->compression
= avio_rl16(pb
);
196 packet
->channels
= avio_rl16(pb
);
197 packet
->sample_rate
= avio_rl32(pb
);
198 packet
->bits_per_sample
= avio_rl16(pb
);
199 packet
->flags
= avio_rl16(pb
);
201 packet
->bit_rate
= packet
->bits_per_sample
*
202 packet
->sample_rate
*
204 packet
->block_align
= XMV_BLOCK_ALIGN_SIZE
* packet
->channels
;
205 packet
->block_samples
= 64;
206 packet
->codec_id
= ff_wav_codec_get_id(packet
->compression
,
207 packet
->bits_per_sample
);
209 packet
->stream_index
= -1;
211 packet
->frame_size
= 0;
212 packet
->block_count
= 0;
214 /* TODO: ADPCM'd 5.1 sound is encoded in three separate streams.
215 * Those need to be interleaved to a proper 5.1 stream. */
216 if (packet
->flags
& XMV_AUDIO_ADPCM51
)
217 av_log(s
, AV_LOG_WARNING
, "Unsupported 5.1 ADPCM audio stream "
218 "(0x%04X)\n", packet
->flags
);
220 if (!packet
->channels
|| !packet
->sample_rate
||
221 packet
->channels
>= UINT16_MAX
/ XMV_BLOCK_ALIGN_SIZE
) {
222 av_log(s
, AV_LOG_ERROR
, "Invalid parameters for audio track %"PRIu16
".\n",
224 ret
= AVERROR_INVALIDDATA
;
228 ast
= avformat_new_stream(s
, NULL
);
230 ret
= AVERROR(ENOMEM
);
234 ast
->codec
->codec_type
= AVMEDIA_TYPE_AUDIO
;
235 ast
->codec
->codec_id
= packet
->codec_id
;
236 ast
->codec
->codec_tag
= packet
->compression
;
237 ast
->codec
->channels
= packet
->channels
;
238 ast
->codec
->sample_rate
= packet
->sample_rate
;
239 ast
->codec
->bits_per_coded_sample
= packet
->bits_per_sample
;
240 ast
->codec
->bit_rate
= packet
->bit_rate
;
241 ast
->codec
->block_align
= 36 * packet
->channels
;
243 avpriv_set_pts_info(ast
, 32, packet
->block_samples
, packet
->sample_rate
);
245 packet
->stream_index
= ast
->index
;
247 ast
->duration
= vst
->duration
;
251 /* Initialize the packet context */
253 xmv
->next_packet_offset
= avio_tell(pb
);
254 xmv
->next_packet_size
= this_packet_size
- xmv
->next_packet_offset
;
255 xmv
->stream_count
= xmv
->audio_track_count
+ 1;
264 static void xmv_read_extradata(uint8_t *extradata
, AVIOContext
*pb
)
266 /* Read the XMV extradata */
268 uint32_t data
= avio_rl32(pb
);
270 int mspel_bit
= !!(data
& 0x01);
271 int loop_filter
= !!(data
& 0x02);
272 int abt_flag
= !!(data
& 0x04);
273 int j_type_bit
= !!(data
& 0x08);
274 int top_left_mv_flag
= !!(data
& 0x10);
275 int per_mb_rl_bit
= !!(data
& 0x20);
276 int slice_count
= (data
>> 6) & 7;
278 /* Write it back as standard WMV2 extradata */
282 data
|= mspel_bit
<< 15;
283 data
|= loop_filter
<< 14;
284 data
|= abt_flag
<< 13;
285 data
|= j_type_bit
<< 12;
286 data
|= top_left_mv_flag
<< 11;
287 data
|= per_mb_rl_bit
<< 10;
288 data
|= slice_count
<< 7;
290 AV_WB32(extradata
, data
);
293 static int xmv_process_packet_header(AVFormatContext
*s
)
295 XMVDemuxContext
*xmv
= s
->priv_data
;
296 AVIOContext
*pb
= s
->pb
;
299 uint16_t audio_track
;
300 uint64_t data_offset
;
302 /* Next packet size */
303 xmv
->next_packet_size
= avio_rl32(pb
);
305 /* Packet video header */
307 if (avio_read(pb
, data
, 8) != 8)
310 xmv
->video
.data_size
= AV_RL32(data
) & 0x007FFFFF;
312 xmv
->video
.current_frame
= 0;
313 xmv
->video
.frame_count
= (AV_RL32(data
) >> 23) & 0xFF;
315 xmv
->video
.has_extradata
= (data
[3] & 0x80) != 0;
317 /* Adding the audio data sizes and the video data size keeps you 4 bytes
318 * short for every audio track. But as playing around with XMV files with
319 * ADPCM audio showed, taking the extra 4 bytes from the audio data gives
320 * you either completely distorted audio or click (when skipping the
321 * remaining 68 bytes of the ADPCM block). Subtracting 4 bytes for every
322 * audio track from the video data works at least for the audio. Probably
323 * some alignment thing?
324 * The video data has (always?) lots of padding, so it should work out...
326 xmv
->video
.data_size
-= xmv
->audio_track_count
* 4;
328 xmv
->current_stream
= 0;
329 if (!xmv
->video
.frame_count
) {
330 xmv
->video
.frame_count
= 1;
331 xmv
->current_stream
= xmv
->stream_count
> 1;
334 /* Packet audio header */
336 for (audio_track
= 0; audio_track
< xmv
->audio_track_count
; audio_track
++) {
337 XMVAudioPacket
*packet
= &xmv
->audio
[audio_track
];
339 if (avio_read(pb
, data
, 4) != 4)
342 packet
->data_size
= AV_RL32(data
) & 0x007FFFFF;
343 if ((packet
->data_size
== 0) && (audio_track
!= 0))
344 /* This happens when I create an XMV with several identical audio
345 * streams. From the size calculations, duplicating the previous
346 * stream's size works out, but the track data itself is silent.
347 * Maybe this should also redirect the offset to the previous track?
349 packet
->data_size
= xmv
->audio
[audio_track
- 1].data_size
;
351 /* Carve up the audio data in frame_count slices */
352 packet
->frame_size
= packet
->data_size
/ xmv
->video
.frame_count
;
353 packet
->frame_size
-= packet
->frame_size
% packet
->block_align
;
356 /* Packet data offsets */
358 data_offset
= avio_tell(pb
);
360 xmv
->video
.data_offset
= data_offset
;
361 data_offset
+= xmv
->video
.data_size
;
363 for (audio_track
= 0; audio_track
< xmv
->audio_track_count
; audio_track
++) {
364 xmv
->audio
[audio_track
].data_offset
= data_offset
;
365 data_offset
+= xmv
->audio
[audio_track
].data_size
;
368 /* Video frames header */
370 /* Read new video extra data */
371 if (xmv
->video
.data_size
> 0) {
372 if (xmv
->video
.has_extradata
) {
373 xmv_read_extradata(xmv
->video
.extradata
, pb
);
375 xmv
->video
.data_size
-= 4;
376 xmv
->video
.data_offset
+= 4;
378 if (xmv
->video
.stream_index
>= 0) {
379 AVStream
*vst
= s
->streams
[xmv
->video
.stream_index
];
381 av_assert0(xmv
->video
.stream_index
< s
->nb_streams
);
383 if (vst
->codec
->extradata_size
< 4) {
384 av_free(vst
->codec
->extradata
);
386 ff_alloc_extradata(vst
->codec
, 4);
389 memcpy(vst
->codec
->extradata
, xmv
->video
.extradata
, 4);
397 static int xmv_fetch_new_packet(AVFormatContext
*s
)
399 XMVDemuxContext
*xmv
= s
->priv_data
;
400 AVIOContext
*pb
= s
->pb
;
403 if (xmv
->this_packet_offset
== xmv
->next_packet_offset
)
407 xmv
->this_packet_offset
= xmv
->next_packet_offset
;
408 if (avio_seek(pb
, xmv
->this_packet_offset
, SEEK_SET
) != xmv
->this_packet_offset
)
411 /* Update the size */
412 xmv
->this_packet_size
= xmv
->next_packet_size
;
413 if (xmv
->this_packet_size
< (12 + xmv
->audio_track_count
* 4))
416 /* Process the header */
417 result
= xmv_process_packet_header(s
);
421 /* Update the offset */
422 xmv
->next_packet_offset
= xmv
->this_packet_offset
+ xmv
->this_packet_size
;
427 static int xmv_fetch_audio_packet(AVFormatContext
*s
,
428 AVPacket
*pkt
, uint32_t stream
)
430 XMVDemuxContext
*xmv
= s
->priv_data
;
431 AVIOContext
*pb
= s
->pb
;
432 XMVAudioPacket
*audio
= &xmv
->audio
[stream
];
435 uint32_t block_count
;
439 if (avio_seek(pb
, audio
->data_offset
, SEEK_SET
) != audio
->data_offset
)
442 if ((xmv
->video
.current_frame
+ 1) < xmv
->video
.frame_count
)
443 /* Not the last frame, get at most frame_size bytes. */
444 data_size
= FFMIN(audio
->frame_size
, audio
->data_size
);
446 /* Last frame, get the rest. */
447 data_size
= audio
->data_size
;
449 /* Read the packet */
450 result
= av_get_packet(pb
, pkt
, data_size
);
454 pkt
->stream_index
= audio
->stream_index
;
456 /* Calculate the PTS */
458 block_count
= data_size
/ audio
->block_align
;
460 pkt
->duration
= block_count
;
461 pkt
->pts
= audio
->block_count
;
462 pkt
->dts
= AV_NOPTS_VALUE
;
464 audio
->block_count
+= block_count
;
467 audio
->data_size
-= data_size
;
468 audio
->data_offset
+= data_size
;
473 static int xmv_fetch_video_packet(AVFormatContext
*s
,
476 XMVDemuxContext
*xmv
= s
->priv_data
;
477 AVIOContext
*pb
= s
->pb
;
478 XMVVideoPacket
*video
= &xmv
->video
;
481 uint32_t frame_header
;
482 uint32_t frame_size
, frame_timestamp
;
486 if (avio_seek(pb
, video
->data_offset
, SEEK_SET
) != video
->data_offset
)
489 /* Read the frame header */
490 frame_header
= avio_rl32(pb
);
492 frame_size
= (frame_header
& 0x1FFFF) * 4 + 4;
493 frame_timestamp
= (frame_header
>> 17);
495 if ((frame_size
+ 4) > video
->data_size
)
498 /* Get the packet data */
499 result
= av_get_packet(pb
, pkt
, frame_size
);
500 if (result
!= frame_size
)
503 /* Contrary to normal WMV2 video, the bit stream in XMV's
504 * WMV2 is little-endian.
505 * TODO: This manual swap is of course suboptimal.
507 for (data
= pkt
->data
, end
= pkt
->data
+ frame_size
; data
< end
; data
+= 4)
508 AV_WB32(data
, AV_RL32(data
));
510 pkt
->stream_index
= video
->stream_index
;
512 /* Calculate the PTS */
514 video
->last_pts
= frame_timestamp
+ video
->pts
;
517 pkt
->pts
= video
->last_pts
;
518 pkt
->dts
= AV_NOPTS_VALUE
;
520 video
->pts
+= frame_timestamp
;
523 pkt
->flags
= (pkt
->data
[0] & 0x80) ? 0 : AV_PKT_FLAG_KEY
;
526 video
->data_size
-= frame_size
+ 4;
527 video
->data_offset
+= frame_size
+ 4;
532 static int xmv_read_packet(AVFormatContext
*s
,
535 XMVDemuxContext
*xmv
= s
->priv_data
;
538 if (xmv
->video
.current_frame
== xmv
->video
.frame_count
) {
539 /* No frames left in this packet, so we fetch a new one */
541 result
= xmv_fetch_new_packet(s
);
546 if (xmv
->current_stream
== 0) {
547 /* Fetch a video frame */
549 result
= xmv_fetch_video_packet(s
, pkt
);
554 /* Fetch an audio frame */
556 result
= xmv_fetch_audio_packet(s
, pkt
, xmv
->current_stream
- 1);
561 /* Increase our counters */
562 if (++xmv
->current_stream
>= xmv
->stream_count
) {
563 xmv
->current_stream
= 0;
564 xmv
->video
.current_frame
+= 1;
570 AVInputFormat ff_xmv_demuxer
= {
572 .long_name
= NULL_IF_CONFIG_SMALL("Microsoft XMV"),
573 .priv_data_size
= sizeof(XMVDemuxContext
),
574 .read_probe
= xmv_probe
,
575 .read_header
= xmv_read_header
,
576 .read_packet
= xmv_read_packet
,
577 .read_close
= xmv_read_close
,