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"
36 /** The min size of an XMV header. */
37 #define XMV_MIN_HEADER_SIZE 36
39 /** Audio flag: ADPCM'd 5.1 stream, front left / right channels */
40 #define XMV_AUDIO_ADPCM51_FRONTLEFTRIGHT 1
41 /** Audio flag: ADPCM'd 5.1 stream, front center / low frequency channels */
42 #define XMV_AUDIO_ADPCM51_FRONTCENTERLOW 2
43 /** Audio flag: ADPCM'd 5.1 stream, rear left / right channels */
44 #define XMV_AUDIO_ADPCM51_REARLEFTRIGHT 4
46 /** Audio flag: Any of the ADPCM'd 5.1 stream flags. */
47 #define XMV_AUDIO_ADPCM51 (XMV_AUDIO_ADPCM51_FRONTLEFTRIGHT | \
48 XMV_AUDIO_ADPCM51_FRONTCENTERLOW | \
49 XMV_AUDIO_ADPCM51_REARLEFTRIGHT)
51 /** A video packet with an XMV file. */
52 typedef struct XMVVideoPacket {
53 int stream_index; ///< The decoder stream index for this video packet.
55 uint32_t data_size; ///< The size of the remaining video data.
56 uint64_t data_offset; ///< The offset of the video data within the file.
58 uint32_t current_frame; ///< The current frame within this video packet.
59 uint32_t frame_count; ///< The amount of frames within this video packet.
61 int has_extradata; ///< Does the video packet contain extra data?
62 uint8_t extradata[4]; ///< The extra data
64 int64_t last_pts; ///< PTS of the last video frame.
65 int64_t pts; ///< PTS of the most current video frame.
68 /** An audio packet with an XMV file. */
69 typedef struct XMVAudioPacket {
70 int stream_index; ///< The decoder stream index for this audio packet.
72 /* Stream format properties. */
73 uint16_t compression; ///< The type of compression.
74 uint16_t channels; ///< Number of channels.
75 uint32_t sample_rate; ///< Sampling rate.
76 uint16_t bits_per_sample; ///< Bits per compressed sample.
77 uint32_t bit_rate; ///< Bits of compressed data per second.
78 uint16_t flags; ///< Flags
79 uint16_t block_align; ///< Bytes per compressed block.
80 uint16_t block_samples; ///< Decompressed samples per compressed block.
82 enum CodecID codec_id; ///< The codec ID of the compression scheme.
84 uint32_t data_size; ///< The size of the remaining audio data.
85 uint64_t data_offset; ///< The offset of the audio data within the file.
87 uint32_t frame_size; ///< Number of bytes to put into an audio frame.
89 uint64_t block_count; ///< Running counter of decompressed audio block.
92 /** Context for demuxing an XMV file. */
93 typedef struct XMVDemuxContext {
94 uint16_t audio_track_count; ///< Number of audio track in this file.
96 uint32_t this_packet_size; ///< Size of the current packet.
97 uint32_t next_packet_size; ///< Size of the next packet.
99 uint64_t this_packet_offset; ///< Offset of the current packet.
100 uint64_t next_packet_offset; ///< Offset of the next packet.
102 uint16_t current_stream; ///< The index of the stream currently handling.
103 uint16_t stream_count; ///< The number of streams in this file.
105 XMVVideoPacket video; ///< The video packet contained in each packet.
106 XMVAudioPacket *audio; ///< The audio packets contained in each packet.
109 static int xmv_probe(AVProbeData *p)
111 uint32_t file_version;
113 if (p->buf_size < XMV_MIN_HEADER_SIZE)
116 file_version = AV_RL32(p->buf + 16);
117 if ((file_version == 0) || (file_version > 4))
120 if (!memcmp(p->buf + 12, "xobX", 4))
121 return AVPROBE_SCORE_MAX;
126 static int xmv_read_header(AVFormatContext *s)
128 XMVDemuxContext *xmv = s->priv_data;
129 AVIOContext *pb = s->pb;
130 AVStream *vst = NULL;
132 uint32_t file_version;
133 uint32_t this_packet_size;
134 uint16_t audio_track;
136 avio_skip(pb, 4); /* Next packet size */
138 this_packet_size = avio_rl32(pb);
140 avio_skip(pb, 4); /* Max packet size */
141 avio_skip(pb, 4); /* "xobX" */
143 file_version = avio_rl32(pb);
144 if ((file_version != 4) && (file_version != 2))
145 av_log_ask_for_sample(s, "Found uncommon version %d\n", file_version);
150 vst = avformat_new_stream(s, NULL);
152 return AVERROR(ENOMEM);
154 avpriv_set_pts_info(vst, 32, 1, 1000);
156 vst->codec->codec_type = AVMEDIA_TYPE_VIDEO;
157 vst->codec->codec_id = CODEC_ID_WMV2;
158 vst->codec->codec_tag = MKBETAG('W', 'M', 'V', '2');
159 vst->codec->width = avio_rl32(pb);
160 vst->codec->height = avio_rl32(pb);
162 vst->duration = avio_rl32(pb);
164 xmv->video.stream_index = vst->index;
168 xmv->audio_track_count = avio_rl16(pb);
170 avio_skip(pb, 2); /* Unknown (padding?) */
172 xmv->audio = av_malloc(xmv->audio_track_count * sizeof(XMVAudioPacket));
174 return AVERROR(ENOMEM);
176 for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
177 XMVAudioPacket *packet = &xmv->audio[audio_track];
178 AVStream *ast = NULL;
180 packet->compression = avio_rl16(pb);
181 packet->channels = avio_rl16(pb);
182 packet->sample_rate = avio_rl32(pb);
183 packet->bits_per_sample = avio_rl16(pb);
184 packet->flags = avio_rl16(pb);
186 if (!packet->channels) {
187 av_log(s, AV_LOG_ERROR, "0 channels\n");
188 return AVERROR(EINVAL);
191 packet->bit_rate = packet->bits_per_sample *
192 packet->sample_rate *
194 packet->block_align = 36 * packet->channels;
195 packet->block_samples = 64;
196 packet->codec_id = ff_wav_codec_get_id(packet->compression,
197 packet->bits_per_sample);
199 packet->stream_index = -1;
201 packet->frame_size = 0;
202 packet->block_count = 0;
204 /* TODO: ADPCM'd 5.1 sound is encoded in three separate streams.
205 * Those need to be interleaved to a proper 5.1 stream. */
206 if (packet->flags & XMV_AUDIO_ADPCM51)
207 av_log(s, AV_LOG_WARNING, "Unsupported 5.1 ADPCM audio stream "
208 "(0x%04X)\n", packet->flags);
210 ast = avformat_new_stream(s, NULL);
212 return AVERROR(ENOMEM);
214 ast->codec->codec_type = AVMEDIA_TYPE_AUDIO;
215 ast->codec->codec_id = packet->codec_id;
216 ast->codec->codec_tag = packet->compression;
217 ast->codec->channels = packet->channels;
218 ast->codec->sample_rate = packet->sample_rate;
219 ast->codec->bits_per_coded_sample = packet->bits_per_sample;
220 ast->codec->bit_rate = packet->bit_rate;
221 ast->codec->block_align = 36 * packet->channels;
223 avpriv_set_pts_info(ast, 32, packet->block_samples, packet->sample_rate);
225 packet->stream_index = ast->index;
227 ast->duration = vst->duration;
231 /* Initialize the packet context */
233 xmv->next_packet_offset = avio_tell(pb);
234 xmv->next_packet_size = this_packet_size - xmv->next_packet_offset;
235 xmv->stream_count = xmv->audio_track_count + 1;
240 static void xmv_read_extradata(uint8_t *extradata, AVIOContext *pb)
242 /* Read the XMV extradata */
244 uint32_t data = avio_rl32(pb);
246 int mspel_bit = !!(data & 0x01);
247 int loop_filter = !!(data & 0x02);
248 int abt_flag = !!(data & 0x04);
249 int j_type_bit = !!(data & 0x08);
250 int top_left_mv_flag = !!(data & 0x10);
251 int per_mb_rl_bit = !!(data & 0x20);
252 int slice_count = (data >> 6) & 7;
254 /* Write it back as standard WMV2 extradata */
258 data |= mspel_bit << 15;
259 data |= loop_filter << 14;
260 data |= abt_flag << 13;
261 data |= j_type_bit << 12;
262 data |= top_left_mv_flag << 11;
263 data |= per_mb_rl_bit << 10;
264 data |= slice_count << 7;
266 AV_WB32(extradata, data);
269 static int xmv_process_packet_header(AVFormatContext *s)
271 XMVDemuxContext *xmv = s->priv_data;
272 AVIOContext *pb = s->pb;
275 uint16_t audio_track;
276 uint64_t data_offset;
278 /* Next packet size */
279 xmv->next_packet_size = avio_rl32(pb);
281 /* Packet video header */
283 if (avio_read(pb, data, 8) != 8)
286 xmv->video.data_size = AV_RL32(data) & 0x007FFFFF;
288 xmv->video.current_frame = 0;
289 xmv->video.frame_count = (AV_RL32(data) >> 23) & 0xFF;
291 xmv->video.has_extradata = (data[3] & 0x80) != 0;
293 /* Adding the audio data sizes and the video data size keeps you 4 bytes
294 * short for every audio track. But as playing around with XMV files with
295 * ADPCM audio showed, taking the extra 4 bytes from the audio data gives
296 * you either completely distorted audio or click (when skipping the
297 * remaining 68 bytes of the ADPCM block). Substracting 4 bytes for every
298 * audio track from the video data works at least for the audio. Probably
299 * some alignment thing?
300 * The video data has (always?) lots of padding, so it should work out...
302 xmv->video.data_size -= xmv->audio_track_count * 4;
304 xmv->current_stream = 0;
305 if (!xmv->video.frame_count) {
306 xmv->video.frame_count = 1;
307 xmv->current_stream = xmv->stream_count > 1;
310 /* Packet audio header */
312 for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
313 XMVAudioPacket *packet = &xmv->audio[audio_track];
315 if (avio_read(pb, data, 4) != 4)
318 packet->data_size = AV_RL32(data) & 0x007FFFFF;
319 if ((packet->data_size == 0) && (audio_track != 0))
320 /* This happens when I create an XMV with several identical audio
321 * streams. From the size calculations, duplicating the previous
322 * stream's size works out, but the track data itself is silent.
323 * Maybe this should also redirect the offset to the previous track?
325 packet->data_size = xmv->audio[audio_track - 1].data_size;
327 /* Carve up the audio data in frame_count slices */
328 packet->frame_size = packet->data_size / xmv->video.frame_count;
329 packet->frame_size -= packet->frame_size % packet->block_align;
332 /* Packet data offsets */
334 data_offset = avio_tell(pb);
336 xmv->video.data_offset = data_offset;
337 data_offset += xmv->video.data_size;
339 for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
340 xmv->audio[audio_track].data_offset = data_offset;
341 data_offset += xmv->audio[audio_track].data_size;
344 /* Video frames header */
346 /* Read new video extra data */
347 if (xmv->video.data_size > 0) {
348 if (xmv->video.has_extradata) {
349 xmv_read_extradata(xmv->video.extradata, pb);
351 xmv->video.data_size -= 4;
352 xmv->video.data_offset += 4;
354 if (xmv->video.stream_index >= 0) {
355 AVStream *vst = s->streams[xmv->video.stream_index];
357 assert(xmv->video.stream_index < s->nb_streams);
359 if (vst->codec->extradata_size < 4) {
360 av_free(vst->codec->extradata);
362 vst->codec->extradata =
363 av_malloc(4 + FF_INPUT_BUFFER_PADDING_SIZE);
364 vst->codec->extradata_size = 4;
367 memcpy(vst->codec->extradata, xmv->video.extradata, 4);
375 static int xmv_fetch_new_packet(AVFormatContext *s)
377 XMVDemuxContext *xmv = s->priv_data;
378 AVIOContext *pb = s->pb;
382 xmv->this_packet_offset = xmv->next_packet_offset;
383 if (avio_seek(pb, xmv->this_packet_offset, SEEK_SET) != xmv->this_packet_offset)
386 /* Update the size */
387 xmv->this_packet_size = xmv->next_packet_size;
388 if (xmv->this_packet_size < (12 + xmv->audio_track_count * 4))
391 /* Process the header */
392 result = xmv_process_packet_header(s);
396 /* Update the offset */
397 xmv->next_packet_offset = xmv->this_packet_offset + xmv->this_packet_size;
402 static int xmv_fetch_audio_packet(AVFormatContext *s,
403 AVPacket *pkt, uint32_t stream)
405 XMVDemuxContext *xmv = s->priv_data;
406 AVIOContext *pb = s->pb;
407 XMVAudioPacket *audio = &xmv->audio[stream];
410 uint32_t block_count;
414 if (avio_seek(pb, audio->data_offset, SEEK_SET) != audio->data_offset)
417 if ((xmv->video.current_frame + 1) < xmv->video.frame_count)
418 /* Not the last frame, get at most frame_size bytes. */
419 data_size = FFMIN(audio->frame_size, audio->data_size);
421 /* Last frame, get the rest. */
422 data_size = audio->data_size;
424 /* Read the packet */
425 result = av_get_packet(pb, pkt, data_size);
429 pkt->stream_index = audio->stream_index;
431 /* Calculate the PTS */
433 block_count = data_size / audio->block_align;
435 pkt->duration = block_count;
436 pkt->pts = audio->block_count;
437 pkt->dts = AV_NOPTS_VALUE;
439 audio->block_count += block_count;
442 audio->data_size -= data_size;
443 audio->data_offset += data_size;
448 static int xmv_fetch_video_packet(AVFormatContext *s,
451 XMVDemuxContext *xmv = s->priv_data;
452 AVIOContext *pb = s->pb;
453 XMVVideoPacket *video = &xmv->video;
456 uint32_t frame_header;
457 uint32_t frame_size, frame_timestamp;
461 if (avio_seek(pb, video->data_offset, SEEK_SET) != video->data_offset)
464 /* Read the frame header */
465 frame_header = avio_rl32(pb);
467 frame_size = (frame_header & 0x1FFFF) * 4 + 4;
468 frame_timestamp = (frame_header >> 17);
470 if ((frame_size + 4) > video->data_size)
473 /* Get the packet data */
474 result = av_get_packet(pb, pkt, frame_size);
475 if (result != frame_size)
478 /* Contrary to normal WMV2 video, the bit stream in XMV's
479 * WMV2 is little-endian.
480 * TODO: This manual swap is of course suboptimal.
482 for (data = pkt->data, end = pkt->data + frame_size; data < end; data += 4)
483 AV_WB32(data, AV_RL32(data));
485 pkt->stream_index = video->stream_index;
487 /* Calculate the PTS */
489 video->last_pts = frame_timestamp + video->pts;
492 pkt->pts = video->last_pts;
493 pkt->dts = AV_NOPTS_VALUE;
495 video->pts += frame_timestamp;
498 pkt->flags = (pkt->data[0] & 0x80) ? 0 : AV_PKT_FLAG_KEY;
501 video->data_size -= frame_size + 4;
502 video->data_offset += frame_size + 4;
507 static int xmv_read_packet(AVFormatContext *s,
510 XMVDemuxContext *xmv = s->priv_data;
513 if (xmv->video.current_frame == xmv->video.frame_count) {
514 /* No frames left in this packet, so we fetch a new one */
516 result = xmv_fetch_new_packet(s);
521 if (xmv->current_stream == 0) {
522 /* Fetch a video frame */
524 result = xmv_fetch_video_packet(s, pkt);
529 /* Fetch an audio frame */
531 result = xmv_fetch_audio_packet(s, pkt, xmv->current_stream - 1);
536 /* Increase our counters */
537 if (++xmv->current_stream >= xmv->stream_count) {
538 xmv->current_stream = 0;
539 xmv->video.current_frame += 1;
545 static int xmv_read_close(AVFormatContext *s)
547 XMVDemuxContext *xmv = s->priv_data;
549 av_freep(&xmv->audio);
554 AVInputFormat ff_xmv_demuxer = {
556 .long_name = NULL_IF_CONFIG_SMALL("Microsoft XMV"),
557 .priv_data_size = sizeof(XMVDemuxContext),
558 .read_probe = xmv_probe,
559 .read_header = xmv_read_header,
560 .read_packet = xmv_read_packet,
561 .read_close = xmv_read_close,