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 /** The min size of an XMV header. */
36 #define XMV_MIN_HEADER_SIZE 36
38 /** Audio flag: ADPCM'd 5.1 stream, front left / right channels */
39 #define XMV_AUDIO_ADPCM51_FRONTLEFTRIGHT 1
40 /** Audio flag: ADPCM'd 5.1 stream, front center / low frequency channels */
41 #define XMV_AUDIO_ADPCM51_FRONTCENTERLOW 2
42 /** Audio flag: ADPCM'd 5.1 stream, rear left / right channels */
43 #define XMV_AUDIO_ADPCM51_REARLEFTRIGHT 4
45 /** Audio flag: Any of the ADPCM'd 5.1 stream flags. */
46 #define XMV_AUDIO_ADPCM51 (XMV_AUDIO_ADPCM51_FRONTLEFTRIGHT | \
47 XMV_AUDIO_ADPCM51_FRONTCENTERLOW | \
48 XMV_AUDIO_ADPCM51_REARLEFTRIGHT)
50 /** A video packet with an XMV file. */
51 typedef struct XMVVideoPacket {
52 int stream_index; ///< The decoder stream index for this video packet.
54 uint32_t data_size; ///< The size of the remaining video data.
55 uint64_t data_offset; ///< The offset of the video data within the file.
57 uint32_t current_frame; ///< The current frame within this video packet.
58 uint32_t frame_count; ///< The amount of frames within this video packet.
60 int has_extradata; ///< Does the video packet contain extra data?
61 uint8_t extradata[4]; ///< The extra data
63 int64_t last_pts; ///< PTS of the last video frame.
64 int64_t pts; ///< PTS of the most current video frame.
67 /** An audio packet with an XMV file. */
68 typedef struct XMVAudioPacket {
69 int stream_index; ///< The decoder stream index for this audio packet.
71 /* Stream format properties. */
72 uint16_t compression; ///< The type of compression.
73 uint16_t channels; ///< Number of channels.
74 uint32_t sample_rate; ///< Sampling rate.
75 uint16_t bits_per_sample; ///< Bits per compressed sample.
76 uint32_t bit_rate; ///< Bits of compressed data per second.
77 uint16_t flags; ///< Flags
78 uint16_t block_align; ///< Bytes per compressed block.
79 uint16_t block_samples; ///< Decompressed samples per compressed block.
81 enum CodecID codec_id; ///< The codec ID of the compression scheme.
83 uint32_t data_size; ///< The size of the remaining audio data.
84 uint64_t data_offset; ///< The offset of the audio data within the file.
86 uint32_t frame_size; ///< Number of bytes to put into an audio frame.
88 uint64_t block_count; ///< Running counter of decompressed audio block.
91 /** Context for demuxing an XMV file. */
92 typedef struct XMVDemuxContext {
93 uint16_t audio_track_count; ///< Number of audio track in this file.
95 uint32_t this_packet_size; ///< Size of the current packet.
96 uint32_t next_packet_size; ///< Size of the next packet.
98 uint64_t this_packet_offset; ///< Offset of the current packet.
99 uint64_t next_packet_offset; ///< Offset of the next packet.
101 uint16_t current_stream; ///< The index of the stream currently handling.
102 uint16_t stream_count; ///< The number of streams in this file.
104 XMVVideoPacket video; ///< The video packet contained in each packet.
105 XMVAudioPacket *audio; ///< The audio packets contained in each packet.
108 static int xmv_probe(AVProbeData *p)
110 uint32_t file_version;
112 if (p->buf_size < XMV_MIN_HEADER_SIZE)
115 file_version = AV_RL32(p->buf + 16);
116 if ((file_version == 0) || (file_version > 4))
119 if (!memcmp(p->buf + 12, "xobX", 4))
120 return AVPROBE_SCORE_MAX;
125 static int xmv_read_header(AVFormatContext *s,
126 AVFormatParameters *ap)
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 av_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 packet->bit_rate = packet->bits_per_sample *
187 packet->sample_rate *
189 packet->block_align = 36 * packet->channels;
190 packet->block_samples = 64;
191 packet->codec_id = ff_wav_codec_get_id(packet->compression,
192 packet->bits_per_sample);
194 packet->stream_index = -1;
196 packet->frame_size = 0;
197 packet->block_count = 0;
199 /* TODO: ADPCM'd 5.1 sound is encoded in three separate streams.
200 * Those need to be interleaved to a proper 5.1 stream. */
201 if (packet->flags & XMV_AUDIO_ADPCM51)
202 av_log(s, AV_LOG_WARNING, "Unsupported 5.1 ADPCM audio stream "
203 "(0x%04X)\n", packet->flags);
205 ast = avformat_new_stream(s, NULL);
207 return AVERROR(ENOMEM);
209 ast->codec->codec_type = AVMEDIA_TYPE_AUDIO;
210 ast->codec->codec_id = packet->codec_id;
211 ast->codec->codec_tag = packet->compression;
212 ast->codec->channels = packet->channels;
213 ast->codec->sample_rate = packet->sample_rate;
214 ast->codec->bits_per_coded_sample = packet->bits_per_sample;
215 ast->codec->bit_rate = packet->bit_rate;
216 ast->codec->block_align = 36 * packet->channels;
218 av_set_pts_info(ast, 32, packet->block_samples, packet->sample_rate);
220 packet->stream_index = ast->index;
222 ast->duration = vst->duration;
226 /* Initialize the packet context */
228 xmv->next_packet_offset = avio_tell(pb);
229 xmv->next_packet_size = this_packet_size - xmv->next_packet_offset;
230 xmv->stream_count = xmv->audio_track_count + 1;
235 static void xmv_read_extradata(uint8_t *extradata, AVIOContext *pb)
237 /* Read the XMV extradata */
239 uint32_t data = avio_rl32(pb);
241 int mspel_bit = !!(data & 0x01);
242 int loop_filter = !!(data & 0x02);
243 int abt_flag = !!(data & 0x04);
244 int j_type_bit = !!(data & 0x08);
245 int top_left_mv_flag = !!(data & 0x10);
246 int per_mb_rl_bit = !!(data & 0x20);
247 int slice_count = (data >> 6) & 7;
249 /* Write it back as standard WMV2 extradata */
253 data |= mspel_bit << 15;
254 data |= loop_filter << 14;
255 data |= abt_flag << 13;
256 data |= j_type_bit << 12;
257 data |= top_left_mv_flag << 11;
258 data |= per_mb_rl_bit << 10;
259 data |= slice_count << 7;
261 AV_WB32(extradata, data);
264 static int xmv_process_packet_header(AVFormatContext *s)
266 XMVDemuxContext *xmv = s->priv_data;
267 AVIOContext *pb = s->pb;
270 uint16_t audio_track;
271 uint64_t data_offset;
273 /* Next packet size */
274 xmv->next_packet_size = avio_rl32(pb);
276 /* Packet video header */
278 if (avio_read(pb, data, 8) != 8)
281 xmv->video.data_size = AV_RL32(data) & 0x007FFFFF;
283 xmv->video.current_frame = 0;
284 xmv->video.frame_count = (AV_RL32(data) >> 23) & 0xFF;
286 xmv->video.has_extradata = (data[3] & 0x80) != 0;
288 /* Adding the audio data sizes and the video data size keeps you 4 bytes
289 * short for every audio track. But as playing around with XMV files with
290 * ADPCM audio showed, taking the extra 4 bytes from the audio data gives
291 * you either completely distorted audio or click (when skipping the
292 * remaining 68 bytes of the ADPCM block). Substracting 4 bytes for every
293 * audio track from the video data works at least for the audio. Probably
294 * some alignment thing?
295 * The video data has (always?) lots of padding, so it should work out...
297 xmv->video.data_size -= xmv->audio_track_count * 4;
299 xmv->current_stream = 0;
300 if (!xmv->video.frame_count) {
301 xmv->video.frame_count = 1;
302 xmv->current_stream = 1;
305 /* Packet audio header */
307 for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
308 XMVAudioPacket *packet = &xmv->audio[audio_track];
310 if (avio_read(pb, data, 4) != 4)
313 packet->data_size = AV_RL32(data) & 0x007FFFFF;
314 if ((packet->data_size == 0) && (audio_track != 0))
315 /* This happens when I create an XMV with several identical audio
316 * streams. From the size calculations, duplicating the previous
317 * stream's size works out, but the track data itself is silent.
318 * Maybe this should also redirect the offset to the previous track?
320 packet->data_size = xmv->audio[audio_track - 1].data_size;
322 /* Carve up the audio data in frame_count slices */
323 packet->frame_size = packet->data_size / xmv->video.frame_count;
324 packet->frame_size -= packet->frame_size % packet->block_align;
327 /* Packet data offsets */
329 data_offset = avio_tell(pb);
331 xmv->video.data_offset = data_offset;
332 data_offset += xmv->video.data_size;
334 for (audio_track = 0; audio_track < xmv->audio_track_count; audio_track++) {
335 xmv->audio[audio_track].data_offset = data_offset;
336 data_offset += xmv->audio[audio_track].data_size;
339 /* Video frames header */
341 /* Read new video extra data */
342 if (xmv->video.data_size > 0) {
343 if (xmv->video.has_extradata) {
344 xmv_read_extradata(xmv->video.extradata, pb);
346 xmv->video.data_size -= 4;
347 xmv->video.data_offset += 4;
349 if (xmv->video.stream_index >= 0) {
350 AVStream *vst = s->streams[xmv->video.stream_index];
352 assert(xmv->video.stream_index < s->nb_streams);
354 if (vst->codec->extradata_size < 4) {
355 av_free(vst->codec->extradata);
357 vst->codec->extradata =
358 av_malloc(4 + FF_INPUT_BUFFER_PADDING_SIZE);
359 vst->codec->extradata_size = 4;
362 memcpy(vst->codec->extradata, xmv->video.extradata, 4);
370 static int xmv_fetch_new_packet(AVFormatContext *s)
372 XMVDemuxContext *xmv = s->priv_data;
373 AVIOContext *pb = s->pb;
377 xmv->this_packet_offset = xmv->next_packet_offset;
378 if (avio_seek(pb, xmv->this_packet_offset, SEEK_SET) != xmv->this_packet_offset)
381 /* Update the size */
382 xmv->this_packet_size = xmv->next_packet_size;
383 if (xmv->this_packet_size < (12 + xmv->audio_track_count * 4))
386 /* Process the header */
387 result = xmv_process_packet_header(s);
391 /* Update the offset */
392 xmv->next_packet_offset = xmv->this_packet_offset + xmv->this_packet_size;
397 static int xmv_fetch_audio_packet(AVFormatContext *s,
398 AVPacket *pkt, uint32_t stream)
400 XMVDemuxContext *xmv = s->priv_data;
401 AVIOContext *pb = s->pb;
402 XMVAudioPacket *audio = &xmv->audio[stream];
405 uint32_t block_count;
409 if (avio_seek(pb, audio->data_offset, SEEK_SET) != audio->data_offset)
412 if ((xmv->video.current_frame + 1) < xmv->video.frame_count)
413 /* Not the last frame, get at most frame_size bytes. */
414 data_size = FFMIN(audio->frame_size, audio->data_size);
416 /* Last frame, get the rest. */
417 data_size = audio->data_size;
419 /* Read the packet */
420 result = av_get_packet(pb, pkt, data_size);
424 pkt->stream_index = audio->stream_index;
426 /* Calculate the PTS */
428 block_count = data_size / audio->block_align;
430 pkt->duration = block_count;
431 pkt->pts = audio->block_count;
432 pkt->dts = AV_NOPTS_VALUE;
434 audio->block_count += block_count;
437 audio->data_size -= data_size;
438 audio->data_offset += data_size;
443 static int xmv_fetch_video_packet(AVFormatContext *s,
446 XMVDemuxContext *xmv = s->priv_data;
447 AVIOContext *pb = s->pb;
448 XMVVideoPacket *video = &xmv->video;
451 uint32_t frame_header;
452 uint32_t frame_size, frame_timestamp;
456 if (avio_seek(pb, video->data_offset, SEEK_SET) != video->data_offset)
459 /* Read the frame header */
460 frame_header = avio_rl32(pb);
462 frame_size = (frame_header & 0x1FFFF) * 4 + 4;
463 frame_timestamp = (frame_header >> 17);
465 if ((frame_size + 4) > video->data_size)
468 /* Get the packet data */
469 result = av_get_packet(pb, pkt, frame_size);
470 if (result != frame_size)
473 /* Contrary to normal WMV2 video, the bit stream in XMV's
474 * WMV2 is little-endian.
475 * TODO: This manual swap is of course suboptimal.
477 for (data = pkt->data, end = pkt->data + frame_size; data < end; data += 4)
478 AV_WB32(data, AV_RL32(data));
480 pkt->stream_index = video->stream_index;
482 /* Calculate the PTS */
484 video->last_pts = frame_timestamp + video->pts;
487 pkt->pts = video->last_pts;
488 pkt->dts = AV_NOPTS_VALUE;
490 video->pts += frame_timestamp;
493 pkt->flags = (pkt->data[0] & 0x80) ? 0 : AV_PKT_FLAG_KEY;
496 video->data_size -= frame_size + 4;
497 video->data_offset += frame_size + 4;
502 static int xmv_read_packet(AVFormatContext *s,
505 XMVDemuxContext *xmv = s->priv_data;
508 if (xmv->video.current_frame == xmv->video.frame_count) {
509 /* No frames left in this packet, so we fetch a new one */
511 result = xmv_fetch_new_packet(s);
516 if (xmv->current_stream == 0) {
517 /* Fetch a video frame */
519 result = xmv_fetch_video_packet(s, pkt);
524 /* Fetch an audio frame */
526 result = xmv_fetch_audio_packet(s, pkt, xmv->current_stream - 1);
531 /* Increase our counters */
532 if (++xmv->current_stream >= xmv->stream_count) {
533 xmv->current_stream = 0;
534 xmv->video.current_frame += 1;
540 static int xmv_read_close(AVFormatContext *s)
542 XMVDemuxContext *xmv = s->priv_data;
549 AVInputFormat ff_xmv_demuxer = {
551 .long_name = NULL_IF_CONFIG_SMALL("Microsoft XMV"),
552 .priv_data_size = sizeof(XMVDemuxContext),
553 .read_probe = xmv_probe,
554 .read_header = xmv_read_header,
555 .read_packet = xmv_read_packet,
556 .read_close = xmv_read_close,