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 /** A video packet with an XMV file. */
53 typedef struct XMVVideoPacket {
54 int stream_index; ///< The decoder stream index for this video packet.
56 uint32_t data_size; ///< The size of the remaining video data.
57 uint64_t data_offset; ///< The offset of the video data within the file.
59 uint32_t current_frame; ///< The current frame within this video packet.
60 uint32_t frame_count; ///< The amount of frames within this video packet.
62 int has_extradata; ///< Does the video packet contain extra data?
63 uint8_t extradata[4]; ///< The extra data
65 int64_t last_pts; ///< PTS of the last video frame.
66 int64_t pts; ///< PTS of the most current video frame.
69 /** An audio packet with an XMV file. */
70 typedef struct XMVAudioPacket {
71 int stream_index; ///< The decoder stream index for this audio packet.
73 /* Stream format properties. */
74 uint16_t compression; ///< The type of compression.
75 uint16_t channels; ///< Number of channels.
76 uint32_t sample_rate; ///< Sampling rate.
77 uint16_t bits_per_sample; ///< Bits per compressed sample.
78 uint32_t bit_rate; ///< Bits of compressed data per second.
79 uint16_t flags; ///< Flags
80 unsigned block_align; ///< Bytes per compressed block.
81 uint16_t block_samples; ///< Decompressed samples per compressed block.
83 enum AVCodecID codec_id; ///< The codec ID of the compression scheme.
85 uint32_t data_size; ///< The size of the remaining audio data.
86 uint64_t data_offset; ///< The offset of the audio data within the file.
88 uint32_t frame_size; ///< Number of bytes to put into an audio frame.
90 uint64_t block_count; ///< Running counter of decompressed audio block.
93 /** Context for demuxing an XMV file. */
94 typedef struct XMVDemuxContext {
95 uint16_t audio_track_count; ///< Number of audio track in this file.
97 uint32_t this_packet_size; ///< Size of the current packet.
98 uint32_t next_packet_size; ///< Size of the next packet.
100 uint64_t this_packet_offset; ///< Offset of the current packet.
101 uint64_t next_packet_offset; ///< Offset of the next packet.
103 uint16_t current_stream; ///< The index of the stream currently handling.
104 uint16_t stream_count; ///< The number of streams in this file.
106 XMVVideoPacket video; ///< The video packet contained in each packet.
107 XMVAudioPacket *audio; ///< The audio packets contained in each packet.
110 static int xmv_probe(AVProbeData *p)
112 uint32_t file_version;
114 if (p->buf_size < XMV_MIN_HEADER_SIZE)
117 file_version = AV_RL32(p->buf + 16);
118 if ((file_version == 0) || (file_version > 4))
121 if (!memcmp(p->buf + 12, "xobX", 4))
122 return AVPROBE_SCORE_MAX;
127 static int xmv_read_close(AVFormatContext *s)
129 XMVDemuxContext *xmv = s->priv_data;
131 av_freep(&xmv->audio);
136 static int xmv_read_header(AVFormatContext *s)
138 XMVDemuxContext *xmv = s->priv_data;
139 AVIOContext *pb = s->pb;
140 AVStream *vst = NULL;
142 uint32_t file_version;
143 uint32_t this_packet_size;
144 uint16_t audio_track;
147 avio_skip(pb, 4); /* Next packet size */
149 this_packet_size = avio_rl32(pb);
151 avio_skip(pb, 4); /* Max packet size */
152 avio_skip(pb, 4); /* "xobX" */
154 file_version = avio_rl32(pb);
155 if ((file_version != 4) && (file_version != 2))
156 avpriv_request_sample(s, "Uncommon version %d", file_version);
161 vst = avformat_new_stream(s, NULL);
163 return AVERROR(ENOMEM);
165 avpriv_set_pts_info(vst, 32, 1, 1000);
167 vst->codec->codec_type = AVMEDIA_TYPE_VIDEO;
168 vst->codec->codec_id = AV_CODEC_ID_WMV2;
169 vst->codec->codec_tag = MKBETAG('W', 'M', 'V', '2');
170 vst->codec->width = avio_rl32(pb);
171 vst->codec->height = avio_rl32(pb);
173 vst->duration = avio_rl32(pb);
175 xmv->video.stream_index = vst->index;
179 xmv->audio_track_count = avio_rl16(pb);
181 avio_skip(pb, 2); /* Unknown (padding?) */
183 xmv->audio = av_malloc(xmv->audio_track_count * sizeof(XMVAudioPacket));
185 ret = AVERROR(ENOMEM);
189 for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
190 XMVAudioPacket *packet = &xmv->audio[audio_track];
191 AVStream *ast = NULL;
193 packet->compression = avio_rl16(pb);
194 packet->channels = avio_rl16(pb);
195 packet->sample_rate = avio_rl32(pb);
196 packet->bits_per_sample = avio_rl16(pb);
197 packet->flags = avio_rl16(pb);
199 packet->bit_rate = packet->bits_per_sample *
200 packet->sample_rate *
202 packet->block_align = 36 * packet->channels;
203 packet->block_samples = 64;
204 packet->codec_id = ff_wav_codec_get_id(packet->compression,
205 packet->bits_per_sample);
207 packet->stream_index = -1;
209 packet->frame_size = 0;
210 packet->block_count = 0;
212 /* TODO: ADPCM'd 5.1 sound is encoded in three separate streams.
213 * Those need to be interleaved to a proper 5.1 stream. */
214 if (packet->flags & XMV_AUDIO_ADPCM51)
215 av_log(s, AV_LOG_WARNING, "Unsupported 5.1 ADPCM audio stream "
216 "(0x%04X)\n", packet->flags);
218 if (!packet->channels || !packet->sample_rate) {
219 av_log(s, AV_LOG_ERROR, "Invalid parameters for audio track %d.\n",
221 ret = AVERROR_INVALIDDATA;
225 ast = avformat_new_stream(s, NULL);
227 ret = AVERROR(ENOMEM);
231 ast->codec->codec_type = AVMEDIA_TYPE_AUDIO;
232 ast->codec->codec_id = packet->codec_id;
233 ast->codec->codec_tag = packet->compression;
234 ast->codec->channels = packet->channels;
235 ast->codec->sample_rate = packet->sample_rate;
236 ast->codec->bits_per_coded_sample = packet->bits_per_sample;
237 ast->codec->bit_rate = packet->bit_rate;
238 ast->codec->block_align = 36 * packet->channels;
240 avpriv_set_pts_info(ast, 32, packet->block_samples, packet->sample_rate);
242 packet->stream_index = ast->index;
244 ast->duration = vst->duration;
248 /* Initialize the packet context */
250 xmv->next_packet_offset = avio_tell(pb);
251 xmv->next_packet_size = this_packet_size - xmv->next_packet_offset;
252 xmv->stream_count = xmv->audio_track_count + 1;
261 static void xmv_read_extradata(uint8_t *extradata, AVIOContext *pb)
263 /* Read the XMV extradata */
265 uint32_t data = avio_rl32(pb);
267 int mspel_bit = !!(data & 0x01);
268 int loop_filter = !!(data & 0x02);
269 int abt_flag = !!(data & 0x04);
270 int j_type_bit = !!(data & 0x08);
271 int top_left_mv_flag = !!(data & 0x10);
272 int per_mb_rl_bit = !!(data & 0x20);
273 int slice_count = (data >> 6) & 7;
275 /* Write it back as standard WMV2 extradata */
279 data |= mspel_bit << 15;
280 data |= loop_filter << 14;
281 data |= abt_flag << 13;
282 data |= j_type_bit << 12;
283 data |= top_left_mv_flag << 11;
284 data |= per_mb_rl_bit << 10;
285 data |= slice_count << 7;
287 AV_WB32(extradata, data);
290 static int xmv_process_packet_header(AVFormatContext *s)
292 XMVDemuxContext *xmv = s->priv_data;
293 AVIOContext *pb = s->pb;
296 uint16_t audio_track;
297 uint64_t data_offset;
299 /* Next packet size */
300 xmv->next_packet_size = avio_rl32(pb);
302 /* Packet video header */
304 if (avio_read(pb, data, 8) != 8)
307 xmv->video.data_size = AV_RL32(data) & 0x007FFFFF;
309 xmv->video.current_frame = 0;
310 xmv->video.frame_count = (AV_RL32(data) >> 23) & 0xFF;
312 xmv->video.has_extradata = (data[3] & 0x80) != 0;
314 /* Adding the audio data sizes and the video data size keeps you 4 bytes
315 * short for every audio track. But as playing around with XMV files with
316 * ADPCM audio showed, taking the extra 4 bytes from the audio data gives
317 * you either completely distorted audio or click (when skipping the
318 * remaining 68 bytes of the ADPCM block). Subtracting 4 bytes for every
319 * audio track from the video data works at least for the audio. Probably
320 * some alignment thing?
321 * The video data has (always?) lots of padding, so it should work out...
323 xmv->video.data_size -= xmv->audio_track_count * 4;
325 xmv->current_stream = 0;
326 if (!xmv->video.frame_count) {
327 xmv->video.frame_count = 1;
328 xmv->current_stream = xmv->stream_count > 1;
331 /* Packet audio header */
333 for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
334 XMVAudioPacket *packet = &xmv->audio[audio_track];
336 if (avio_read(pb, data, 4) != 4)
339 packet->data_size = AV_RL32(data) & 0x007FFFFF;
340 if ((packet->data_size == 0) && (audio_track != 0))
341 /* This happens when I create an XMV with several identical audio
342 * streams. From the size calculations, duplicating the previous
343 * stream's size works out, but the track data itself is silent.
344 * Maybe this should also redirect the offset to the previous track?
346 packet->data_size = xmv->audio[audio_track - 1].data_size;
348 /* Carve up the audio data in frame_count slices */
349 packet->frame_size = packet->data_size / xmv->video.frame_count;
350 packet->frame_size -= packet->frame_size % packet->block_align;
353 /* Packet data offsets */
355 data_offset = avio_tell(pb);
357 xmv->video.data_offset = data_offset;
358 data_offset += xmv->video.data_size;
360 for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
361 xmv->audio[audio_track].data_offset = data_offset;
362 data_offset += xmv->audio[audio_track].data_size;
365 /* Video frames header */
367 /* Read new video extra data */
368 if (xmv->video.data_size > 0) {
369 if (xmv->video.has_extradata) {
370 xmv_read_extradata(xmv->video.extradata, pb);
372 xmv->video.data_size -= 4;
373 xmv->video.data_offset += 4;
375 if (xmv->video.stream_index >= 0) {
376 AVStream *vst = s->streams[xmv->video.stream_index];
378 av_assert0(xmv->video.stream_index < s->nb_streams);
380 if (vst->codec->extradata_size < 4) {
381 av_free(vst->codec->extradata);
383 vst->codec->extradata =
384 av_malloc(4 + FF_INPUT_BUFFER_PADDING_SIZE);
385 vst->codec->extradata_size = 4;
388 memcpy(vst->codec->extradata, xmv->video.extradata, 4);
396 static int xmv_fetch_new_packet(AVFormatContext *s)
398 XMVDemuxContext *xmv = s->priv_data;
399 AVIOContext *pb = s->pb;
402 if (xmv->this_packet_offset == xmv->next_packet_offset)
406 xmv->this_packet_offset = xmv->next_packet_offset;
407 if (avio_seek(pb, xmv->this_packet_offset, SEEK_SET) != xmv->this_packet_offset)
410 /* Update the size */
411 xmv->this_packet_size = xmv->next_packet_size;
412 if (xmv->this_packet_size < (12 + xmv->audio_track_count * 4))
415 /* Process the header */
416 result = xmv_process_packet_header(s);
420 /* Update the offset */
421 xmv->next_packet_offset = xmv->this_packet_offset + xmv->this_packet_size;
426 static int xmv_fetch_audio_packet(AVFormatContext *s,
427 AVPacket *pkt, uint32_t stream)
429 XMVDemuxContext *xmv = s->priv_data;
430 AVIOContext *pb = s->pb;
431 XMVAudioPacket *audio = &xmv->audio[stream];
434 uint32_t block_count;
438 if (avio_seek(pb, audio->data_offset, SEEK_SET) != audio->data_offset)
441 if ((xmv->video.current_frame + 1) < xmv->video.frame_count)
442 /* Not the last frame, get at most frame_size bytes. */
443 data_size = FFMIN(audio->frame_size, audio->data_size);
445 /* Last frame, get the rest. */
446 data_size = audio->data_size;
448 /* Read the packet */
449 result = av_get_packet(pb, pkt, data_size);
453 pkt->stream_index = audio->stream_index;
455 /* Calculate the PTS */
457 block_count = data_size / audio->block_align;
459 pkt->duration = block_count;
460 pkt->pts = audio->block_count;
461 pkt->dts = AV_NOPTS_VALUE;
463 audio->block_count += block_count;
466 audio->data_size -= data_size;
467 audio->data_offset += data_size;
472 static int xmv_fetch_video_packet(AVFormatContext *s,
475 XMVDemuxContext *xmv = s->priv_data;
476 AVIOContext *pb = s->pb;
477 XMVVideoPacket *video = &xmv->video;
480 uint32_t frame_header;
481 uint32_t frame_size, frame_timestamp;
485 if (avio_seek(pb, video->data_offset, SEEK_SET) != video->data_offset)
488 /* Read the frame header */
489 frame_header = avio_rl32(pb);
491 frame_size = (frame_header & 0x1FFFF) * 4 + 4;
492 frame_timestamp = (frame_header >> 17);
494 if ((frame_size + 4) > video->data_size)
497 /* Get the packet data */
498 result = av_get_packet(pb, pkt, frame_size);
499 if (result != frame_size)
502 /* Contrary to normal WMV2 video, the bit stream in XMV's
503 * WMV2 is little-endian.
504 * TODO: This manual swap is of course suboptimal.
506 for (data = pkt->data, end = pkt->data + frame_size; data < end; data += 4)
507 AV_WB32(data, AV_RL32(data));
509 pkt->stream_index = video->stream_index;
511 /* Calculate the PTS */
513 video->last_pts = frame_timestamp + video->pts;
516 pkt->pts = video->last_pts;
517 pkt->dts = AV_NOPTS_VALUE;
519 video->pts += frame_timestamp;
522 pkt->flags = (pkt->data[0] & 0x80) ? 0 : AV_PKT_FLAG_KEY;
525 video->data_size -= frame_size + 4;
526 video->data_offset += frame_size + 4;
531 static int xmv_read_packet(AVFormatContext *s,
534 XMVDemuxContext *xmv = s->priv_data;
537 if (xmv->video.current_frame == xmv->video.frame_count) {
538 /* No frames left in this packet, so we fetch a new one */
540 result = xmv_fetch_new_packet(s);
545 if (xmv->current_stream == 0) {
546 /* Fetch a video frame */
548 result = xmv_fetch_video_packet(s, pkt);
553 /* Fetch an audio frame */
555 result = xmv_fetch_audio_packet(s, pkt, xmv->current_stream - 1);
560 /* Increase our counters */
561 if (++xmv->current_stream >= xmv->stream_count) {
562 xmv->current_stream = 0;
563 xmv->video.current_frame += 1;
569 AVInputFormat ff_xmv_demuxer = {
571 .long_name = NULL_IF_CONFIG_SMALL("Microsoft XMV"),
572 .priv_data_size = sizeof(XMVDemuxContext),
573 .read_probe = xmv_probe,
574 .read_header = xmv_read_header,
575 .read_packet = xmv_read_packet,
576 .read_close = xmv_read_close,