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;
300 uint16_t audio_track;
301 uint64_t data_offset;
303 /* Next packet size */
304 xmv->next_packet_size = avio_rl32(pb);
306 /* Packet video header */
308 if (avio_read(pb, data, 8) != 8)
311 xmv->video.data_size = AV_RL32(data) & 0x007FFFFF;
313 xmv->video.current_frame = 0;
314 xmv->video.frame_count = (AV_RL32(data) >> 23) & 0xFF;
316 xmv->video.has_extradata = (data[3] & 0x80) != 0;
318 /* Adding the audio data sizes and the video data size keeps you 4 bytes
319 * short for every audio track. But as playing around with XMV files with
320 * ADPCM audio showed, taking the extra 4 bytes from the audio data gives
321 * you either completely distorted audio or click (when skipping the
322 * remaining 68 bytes of the ADPCM block). Subtracting 4 bytes for every
323 * audio track from the video data works at least for the audio. Probably
324 * some alignment thing?
325 * The video data has (always?) lots of padding, so it should work out...
327 xmv->video.data_size -= xmv->audio_track_count * 4;
329 xmv->current_stream = 0;
330 if (!xmv->video.frame_count) {
331 xmv->video.frame_count = 1;
332 xmv->current_stream = xmv->stream_count > 1;
335 /* Packet audio header */
337 for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
338 XMVAudioPacket *packet = &xmv->audio[audio_track];
340 if (avio_read(pb, data, 4) != 4)
343 packet->data_size = AV_RL32(data) & 0x007FFFFF;
344 if ((packet->data_size == 0) && (audio_track != 0))
345 /* This happens when I create an XMV with several identical audio
346 * streams. From the size calculations, duplicating the previous
347 * stream's size works out, but the track data itself is silent.
348 * Maybe this should also redirect the offset to the previous track?
350 packet->data_size = xmv->audio[audio_track - 1].data_size;
352 /* Carve up the audio data in frame_count slices */
353 packet->frame_size = packet->data_size / xmv->video.frame_count;
354 packet->frame_size -= packet->frame_size % packet->block_align;
357 /* Packet data offsets */
359 data_offset = avio_tell(pb);
361 xmv->video.data_offset = data_offset;
362 data_offset += xmv->video.data_size;
364 for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
365 xmv->audio[audio_track].data_offset = data_offset;
366 data_offset += xmv->audio[audio_track].data_size;
369 /* Video frames header */
371 /* Read new video extra data */
372 if (xmv->video.data_size > 0) {
373 if (xmv->video.has_extradata) {
374 xmv_read_extradata(xmv->video.extradata, pb);
376 xmv->video.data_size -= 4;
377 xmv->video.data_offset += 4;
379 if (xmv->video.stream_index >= 0) {
380 AVStream *vst = s->streams[xmv->video.stream_index];
382 av_assert0(xmv->video.stream_index < s->nb_streams);
384 if (vst->codec->extradata_size < 4) {
385 av_freep(&vst->codec->extradata);
387 if ((ret = ff_alloc_extradata(vst->codec, 4)) < 0)
391 memcpy(vst->codec->extradata, xmv->video.extradata, 4);
399 static int xmv_fetch_new_packet(AVFormatContext *s)
401 XMVDemuxContext *xmv = s->priv_data;
402 AVIOContext *pb = s->pb;
405 if (xmv->this_packet_offset == xmv->next_packet_offset)
409 xmv->this_packet_offset = xmv->next_packet_offset;
410 if (avio_seek(pb, xmv->this_packet_offset, SEEK_SET) != xmv->this_packet_offset)
413 /* Update the size */
414 xmv->this_packet_size = xmv->next_packet_size;
415 if (xmv->this_packet_size < (12 + xmv->audio_track_count * 4))
418 /* Process the header */
419 result = xmv_process_packet_header(s);
423 /* Update the offset */
424 xmv->next_packet_offset = xmv->this_packet_offset + xmv->this_packet_size;
429 static int xmv_fetch_audio_packet(AVFormatContext *s,
430 AVPacket *pkt, uint32_t stream)
432 XMVDemuxContext *xmv = s->priv_data;
433 AVIOContext *pb = s->pb;
434 XMVAudioPacket *audio = &xmv->audio[stream];
437 uint32_t block_count;
441 if (avio_seek(pb, audio->data_offset, SEEK_SET) != audio->data_offset)
444 if ((xmv->video.current_frame + 1) < xmv->video.frame_count)
445 /* Not the last frame, get at most frame_size bytes. */
446 data_size = FFMIN(audio->frame_size, audio->data_size);
448 /* Last frame, get the rest. */
449 data_size = audio->data_size;
451 /* Read the packet */
452 result = av_get_packet(pb, pkt, data_size);
456 pkt->stream_index = audio->stream_index;
458 /* Calculate the PTS */
460 block_count = data_size / audio->block_align;
462 pkt->duration = block_count;
463 pkt->pts = audio->block_count;
464 pkt->dts = AV_NOPTS_VALUE;
466 audio->block_count += block_count;
469 audio->data_size -= data_size;
470 audio->data_offset += data_size;
475 static int xmv_fetch_video_packet(AVFormatContext *s,
478 XMVDemuxContext *xmv = s->priv_data;
479 AVIOContext *pb = s->pb;
480 XMVVideoPacket *video = &xmv->video;
483 uint32_t frame_header;
484 uint32_t frame_size, frame_timestamp;
488 if (avio_seek(pb, video->data_offset, SEEK_SET) != video->data_offset)
491 /* Read the frame header */
492 frame_header = avio_rl32(pb);
494 frame_size = (frame_header & 0x1FFFF) * 4 + 4;
495 frame_timestamp = (frame_header >> 17);
497 if ((frame_size + 4) > video->data_size)
500 /* Get the packet data */
501 result = av_get_packet(pb, pkt, frame_size);
502 if (result != frame_size)
505 /* Contrary to normal WMV2 video, the bit stream in XMV's
506 * WMV2 is little-endian.
507 * TODO: This manual swap is of course suboptimal.
509 for (data = pkt->data, end = pkt->data + frame_size; data < end; data += 4)
510 AV_WB32(data, AV_RL32(data));
512 pkt->stream_index = video->stream_index;
514 /* Calculate the PTS */
516 video->last_pts = frame_timestamp + video->pts;
519 pkt->pts = video->last_pts;
520 pkt->dts = AV_NOPTS_VALUE;
522 video->pts += frame_timestamp;
525 pkt->flags = (pkt->data[0] & 0x80) ? 0 : AV_PKT_FLAG_KEY;
528 video->data_size -= frame_size + 4;
529 video->data_offset += frame_size + 4;
534 static int xmv_read_packet(AVFormatContext *s,
537 XMVDemuxContext *xmv = s->priv_data;
540 if (xmv->video.current_frame == xmv->video.frame_count) {
541 /* No frames left in this packet, so we fetch a new one */
543 result = xmv_fetch_new_packet(s);
548 if (xmv->current_stream == 0) {
549 /* Fetch a video frame */
551 result = xmv_fetch_video_packet(s, pkt);
556 /* Fetch an audio frame */
558 result = xmv_fetch_audio_packet(s, pkt, xmv->current_stream - 1);
563 /* Increase our counters */
564 if (++xmv->current_stream >= xmv->stream_count) {
565 xmv->current_stream = 0;
566 xmv->video.current_frame += 1;
572 AVInputFormat ff_xmv_demuxer = {
574 .long_name = NULL_IF_CONFIG_SMALL("Microsoft XMV"),
575 .priv_data_size = sizeof(XMVDemuxContext),
576 .read_probe = xmv_probe,
577 .read_header = xmv_read_header,
578 .read_packet = xmv_read_packet,
579 .read_close = xmv_read_close,