muxer.cpp revision afde4e56566af19b36f1fe5e7aa7f226bf1703dd 
1/*
2 * Copyright (C) 2013 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 *      http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17//#define LOG_NDEBUG 0
18#define LOG_TAG "muxer"
19#include <utils/Log.h>
20
21#include <binder/ProcessState.h>
22#include <media/stagefright/foundation/ABuffer.h>
23#include <media/stagefright/foundation/ADebug.h>
24#include <media/stagefright/foundation/ALooper.h>
25#include <media/stagefright/foundation/AMessage.h>
26#include <media/stagefright/foundation/AString.h>
27#include <media/stagefright/DataSource.h>
28#include <media/stagefright/MediaCodec.h>
29#include <media/stagefright/MediaDefs.h>
30#include <media/stagefright/MediaMuxer.h>
31#include <media/stagefright/MetaData.h>
32#include <media/stagefright/NuMediaExtractor.h>
33
34static void usage(const char *me) {
35    fprintf(stderr, "usage: %s [-a] [-v] [-s <trim start time>]"
36                    " [-e <trim end time>] [-o <output file>]"
37                    " <input video file>\n", me);
38    fprintf(stderr, "       -h help\n");
39    fprintf(stderr, "       -a use audio\n");
40    fprintf(stderr, "       -v use video\n");
41    fprintf(stderr, "       -s Time in milli-seconds when the trim should start\n");
42    fprintf(stderr, "       -e Time in milli-seconds when the trim should end\n");
43    fprintf(stderr, "       -o output file name. Default is /sdcard/muxeroutput.mp4\n");
44
45    exit(1);
46}
47
48using namespace android;
49
50static int muxing(
51        const android::sp<android::ALooper> &looper,
52        const char *path,
53        bool useAudio,
54        bool useVideo,
55        const char *outputFileName,
56        bool enableTrim,
57        int trimStartTimeMs,
58        int trimEndTimeMs) {
59    sp<NuMediaExtractor> extractor = new NuMediaExtractor;
60    if (extractor->setDataSource(path) != OK) {
61        fprintf(stderr, "unable to instantiate extractor. %s\n", path);
62        return 1;
63    }
64
65    if (outputFileName == NULL) {
66        outputFileName = "/sdcard/muxeroutput.mp4";
67    }
68
69    ALOGV("input file %s, output file %s", path, outputFileName);
70    ALOGV("useAudio %d, useVideo %d", useAudio, useVideo);
71
72    sp<MediaMuxer> muxer = new MediaMuxer(outputFileName,
73                                          MediaMuxer::OUTPUT_FORMAT_MPEG_4);
74
75    size_t trackCount = extractor->countTracks();
76    // Map the extractor's track index to the muxer's track index.
77    KeyedVector<size_t, ssize_t> trackIndexMap;
78    size_t bufferSize = 1 * 1024 * 1024;  // default buffer size is 1MB.
79
80    bool haveAudio = false;
81    bool haveVideo = false;
82
83    int64_t trimStartTimeUs = trimStartTimeMs * 1000;
84    int64_t trimEndTimeUs = trimEndTimeMs * 1000;
85    bool trimStarted = false;
86    int64_t trimOffsetTimeUs = 0;
87
88    for (size_t i = 0; i < trackCount; ++i) {
89        sp<AMessage> format;
90        status_t err = extractor->getTrackFormat(i, &format);
91        CHECK_EQ(err, (status_t)OK);
92        ALOGV("extractor getTrackFormat: %s", format->debugString().c_str());
93
94        AString mime;
95        CHECK(format->findString("mime", &mime));
96
97        bool isAudio = !strncasecmp(mime.c_str(), "audio/", 6);
98        bool isVideo = !strncasecmp(mime.c_str(), "video/", 6);
99
100        if (useAudio && !haveAudio && isAudio) {
101            haveAudio = true;
102        } else if (useVideo && !haveVideo && isVideo) {
103            haveVideo = true;
104        } else {
105            continue;
106        }
107
108        if (isVideo) {
109            int width , height;
110            CHECK(format->findInt32("width", &width));
111            CHECK(format->findInt32("height", &height));
112            bufferSize = width * height * 4;  // Assuming it is maximally 4BPP
113        }
114
115        int64_t duration;
116        CHECK(format->findInt64("durationUs", &duration));
117
118        // Since we got the duration now, correct the start time.
119        if (enableTrim) {
120            if (trimStartTimeUs > duration) {
121                fprintf(stderr, "Warning: trimStartTimeUs > duration,"
122                                " reset to 0\n");
123                trimStartTimeUs = 0;
124            }
125        }
126
127        ALOGV("selecting track %d", i);
128
129        err = extractor->selectTrack(i);
130        CHECK_EQ(err, (status_t)OK);
131
132        ssize_t newTrackIndex = muxer->addTrack(format);
133        CHECK_GE(newTrackIndex, 0);
134        trackIndexMap.add(i, newTrackIndex);
135    }
136
137    int64_t muxerStartTimeUs = ALooper::GetNowUs();
138
139    bool sawInputEOS = false;
140
141    size_t trackIndex = -1;
142    sp<ABuffer> newBuffer = new ABuffer(bufferSize);
143
144    muxer->start();
145
146    while (!sawInputEOS) {
147        status_t err = extractor->getSampleTrackIndex(&trackIndex);
148        if (err != OK) {
149            ALOGV("saw input eos, err %d", err);
150            sawInputEOS = true;
151            break;
152        } else {
153            err = extractor->readSampleData(newBuffer);
154            CHECK_EQ(err, (status_t)OK);
155
156            int64_t timeUs;
157            err = extractor->getSampleTime(&timeUs);
158            CHECK_EQ(err, (status_t)OK);
159
160            sp<MetaData> meta;
161            err = extractor->getSampleMeta(&meta);
162            CHECK_EQ(err, (status_t)OK);
163
164            uint32_t sampleFlags = 0;
165            int32_t val;
166            if (meta->findInt32(kKeyIsSyncFrame, &val) && val != 0) {
167                // We only support BUFFER_FLAG_SYNCFRAME in the flag for now.
168                sampleFlags |= MediaCodec::BUFFER_FLAG_SYNCFRAME;
169
170                // We turn on trimming at the sync frame.
171                if (enableTrim && timeUs > trimStartTimeUs &&
172                    timeUs <= trimEndTimeUs) {
173                    if (trimStarted == false) {
174                        trimOffsetTimeUs = timeUs;
175                    }
176                    trimStarted = true;
177                }
178            }
179            // Trim can end at any non-sync frame.
180            if (enableTrim && timeUs > trimEndTimeUs) {
181                trimStarted = false;
182            }
183
184            if (!enableTrim || (enableTrim && trimStarted)) {
185                err = muxer->writeSampleData(newBuffer,
186                                             trackIndexMap.valueFor(trackIndex),
187                                             timeUs - trimOffsetTimeUs, sampleFlags);
188            }
189
190            extractor->advance();
191        }
192    }
193
194    muxer->stop();
195    newBuffer.clear();
196    trackIndexMap.clear();
197
198    int64_t elapsedTimeUs = ALooper::GetNowUs() - muxerStartTimeUs;
199    fprintf(stderr, "SUCCESS: muxer generate the video in %lld ms\n",
200            elapsedTimeUs / 1000);
201
202    return 0;
203}
204
205int main(int argc, char **argv) {
206    const char *me = argv[0];
207
208    bool useAudio = false;
209    bool useVideo = false;
210    char *outputFileName = NULL;
211    int trimStartTimeMs = -1;
212    int trimEndTimeMs = -1;
213    // When trimStartTimeMs and trimEndTimeMs seems valid, we turn this switch
214    // to true.
215    bool enableTrim = false;
216
217    int res;
218    while ((res = getopt(argc, argv, "h?avo:s:e:")) >= 0) {
219        switch (res) {
220            case 'a':
221            {
222                useAudio = true;
223                break;
224            }
225
226            case 'v':
227            {
228                useVideo = true;
229                break;
230            }
231
232            case 'o':
233            {
234                outputFileName = optarg;
235                break;
236            }
237
238            case 's':
239            {
240                trimStartTimeMs = atoi(optarg);
241                break;
242            }
243
244            case 'e':
245            {
246                trimEndTimeMs = atoi(optarg);
247                break;
248            }
249
250            case '?':
251            case 'h':
252            default:
253            {
254                usage(me);
255            }
256        }
257    }
258
259    argc -= optind;
260    argv += optind;
261
262    if (argc != 1) {
263        usage(me);
264    }
265
266    if (trimStartTimeMs < 0 || trimEndTimeMs < 0) {
267        // If no input on either 's' or 'e', or they are obviously wrong input,
268        // then turn off trimming.
269        ALOGV("Trimming is disabled, copying the whole length video.");
270        enableTrim = false;
271    } else if (trimStartTimeMs > trimEndTimeMs) {
272        fprintf(stderr, "ERROR: start time is bigger\n");
273        return 1;
274    } else {
275        enableTrim = true;
276    }
277
278    if (!useAudio && !useVideo) {
279        fprintf(stderr, "ERROR: Missing both -a and -v, no track to mux then.\n");
280        return 1;
281    }
282    ProcessState::self()->startThreadPool();
283
284    // Make sure setDataSource() works.
285    DataSource::RegisterDefaultSniffers();
286
287    sp<ALooper> looper = new ALooper;
288    looper->start();
289
290    int result = muxing(looper, argv[0], useAudio, useVideo, outputFileName,
291                        enableTrim, trimStartTimeMs, trimEndTimeMs);
292
293    looper->stop();
294
295    return result;
296}
297