blob: 56e41792a78800f6cca063c91d5924e4b8135ab9
1 | /* |
2 | * Copyright (c) 2012 Nicolas George |
3 | * |
4 | * This file is part of FFmpeg. |
5 | * |
6 | * FFmpeg is free software; you can redistribute it and/or |
7 | * modify it under the terms of the GNU Lesser General Public |
8 | * License as published by the Free Software Foundation; either |
9 | * version 2.1 of the License, or (at your option) any later version. |
10 | * |
11 | * FFmpeg is distributed in the hope that it will be useful, |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. |
14 | * See the GNU Lesser General Public License for more details. |
15 | * |
16 | * You should have received a copy of the GNU Lesser General Public License |
17 | * along with FFmpeg; if not, write to the Free Software Foundation, Inc., |
18 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
19 | */ |
20 | |
21 | /** |
22 | * @file |
23 | * concat audio-video filter |
24 | */ |
25 | |
26 | #include "libavutil/avassert.h" |
27 | #include "libavutil/avstring.h" |
28 | #include "libavutil/channel_layout.h" |
29 | #include "libavutil/opt.h" |
30 | #include "avfilter.h" |
31 | #define FF_BUFQUEUE_SIZE 256 |
32 | #include "bufferqueue.h" |
33 | #include "internal.h" |
34 | #include "video.h" |
35 | #include "audio.h" |
36 | |
37 | #define TYPE_ALL 2 |
38 | |
39 | typedef struct { |
40 | const AVClass *class; |
41 | unsigned nb_streams[TYPE_ALL]; /**< number of out streams of each type */ |
42 | unsigned nb_segments; |
43 | unsigned cur_idx; /**< index of the first input of current segment */ |
44 | int64_t delta_ts; /**< timestamp to add to produce output timestamps */ |
45 | unsigned nb_in_active; /**< number of active inputs in current segment */ |
46 | unsigned unsafe; |
47 | struct concat_in { |
48 | int64_t pts; |
49 | int64_t nb_frames; |
50 | unsigned eof; |
51 | struct FFBufQueue queue; |
52 | } *in; |
53 | } ConcatContext; |
54 | |
55 | #define OFFSET(x) offsetof(ConcatContext, x) |
56 | #define A AV_OPT_FLAG_AUDIO_PARAM |
57 | #define F AV_OPT_FLAG_FILTERING_PARAM |
58 | #define V AV_OPT_FLAG_VIDEO_PARAM |
59 | |
60 | static const AVOption concat_options[] = { |
61 | { "n", "specify the number of segments", OFFSET(nb_segments), |
62 | AV_OPT_TYPE_INT, { .i64 = 2 }, 1, INT_MAX, V|A|F}, |
63 | { "v", "specify the number of video streams", |
64 | OFFSET(nb_streams[AVMEDIA_TYPE_VIDEO]), |
65 | AV_OPT_TYPE_INT, { .i64 = 1 }, 0, INT_MAX, V|F }, |
66 | { "a", "specify the number of audio streams", |
67 | OFFSET(nb_streams[AVMEDIA_TYPE_AUDIO]), |
68 | AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, A|F}, |
69 | { "unsafe", "enable unsafe mode", |
70 | OFFSET(unsafe), |
71 | AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, V|A|F}, |
72 | { NULL } |
73 | }; |
74 | |
75 | AVFILTER_DEFINE_CLASS(concat); |
76 | |
77 | static int query_formats(AVFilterContext *ctx) |
78 | { |
79 | ConcatContext *cat = ctx->priv; |
80 | unsigned type, nb_str, idx0 = 0, idx, str, seg; |
81 | AVFilterFormats *formats, *rates = NULL; |
82 | AVFilterChannelLayouts *layouts = NULL; |
83 | int ret; |
84 | |
85 | for (type = 0; type < TYPE_ALL; type++) { |
86 | nb_str = cat->nb_streams[type]; |
87 | for (str = 0; str < nb_str; str++) { |
88 | idx = idx0; |
89 | |
90 | /* Set the output formats */ |
91 | formats = ff_all_formats(type); |
92 | if ((ret = ff_formats_ref(formats, &ctx->outputs[idx]->in_formats)) < 0) |
93 | return ret; |
94 | |
95 | if (type == AVMEDIA_TYPE_AUDIO) { |
96 | rates = ff_all_samplerates(); |
97 | if ((ret = ff_formats_ref(rates, &ctx->outputs[idx]->in_samplerates)) < 0) |
98 | return ret; |
99 | layouts = ff_all_channel_layouts(); |
100 | if ((ret = ff_channel_layouts_ref(layouts, &ctx->outputs[idx]->in_channel_layouts)) < 0) |
101 | return ret; |
102 | } |
103 | |
104 | /* Set the same formats for each corresponding input */ |
105 | for (seg = 0; seg < cat->nb_segments; seg++) { |
106 | if ((ret = ff_formats_ref(formats, &ctx->inputs[idx]->out_formats)) < 0) |
107 | return ret; |
108 | if (type == AVMEDIA_TYPE_AUDIO) { |
109 | if ((ret = ff_formats_ref(rates, &ctx->inputs[idx]->out_samplerates)) < 0 || |
110 | (ret = ff_channel_layouts_ref(layouts, &ctx->inputs[idx]->out_channel_layouts)) < 0) |
111 | return ret; |
112 | } |
113 | idx += ctx->nb_outputs; |
114 | } |
115 | |
116 | idx0++; |
117 | } |
118 | } |
119 | return 0; |
120 | } |
121 | |
122 | static int config_output(AVFilterLink *outlink) |
123 | { |
124 | AVFilterContext *ctx = outlink->src; |
125 | ConcatContext *cat = ctx->priv; |
126 | unsigned out_no = FF_OUTLINK_IDX(outlink); |
127 | unsigned in_no = out_no, seg; |
128 | AVFilterLink *inlink = ctx->inputs[in_no]; |
129 | |
130 | /* enhancement: find a common one */ |
131 | outlink->time_base = AV_TIME_BASE_Q; |
132 | outlink->w = inlink->w; |
133 | outlink->h = inlink->h; |
134 | outlink->sample_aspect_ratio = inlink->sample_aspect_ratio; |
135 | outlink->format = inlink->format; |
136 | for (seg = 1; seg < cat->nb_segments; seg++) { |
137 | inlink = ctx->inputs[in_no += ctx->nb_outputs]; |
138 | if (!outlink->sample_aspect_ratio.num) |
139 | outlink->sample_aspect_ratio = inlink->sample_aspect_ratio; |
140 | /* possible enhancement: unsafe mode, do not check */ |
141 | if (outlink->w != inlink->w || |
142 | outlink->h != inlink->h || |
143 | outlink->sample_aspect_ratio.num != inlink->sample_aspect_ratio.num && |
144 | inlink->sample_aspect_ratio.num || |
145 | outlink->sample_aspect_ratio.den != inlink->sample_aspect_ratio.den) { |
146 | av_log(ctx, AV_LOG_ERROR, "Input link %s parameters " |
147 | "(size %dx%d, SAR %d:%d) do not match the corresponding " |
148 | "output link %s parameters (%dx%d, SAR %d:%d)\n", |
149 | ctx->input_pads[in_no].name, inlink->w, inlink->h, |
150 | inlink->sample_aspect_ratio.num, |
151 | inlink->sample_aspect_ratio.den, |
152 | ctx->input_pads[out_no].name, outlink->w, outlink->h, |
153 | outlink->sample_aspect_ratio.num, |
154 | outlink->sample_aspect_ratio.den); |
155 | if (!cat->unsafe) |
156 | return AVERROR(EINVAL); |
157 | } |
158 | } |
159 | |
160 | return 0; |
161 | } |
162 | |
163 | static int push_frame(AVFilterContext *ctx, unsigned in_no, AVFrame *buf) |
164 | { |
165 | ConcatContext *cat = ctx->priv; |
166 | unsigned out_no = in_no % ctx->nb_outputs; |
167 | AVFilterLink * inlink = ctx-> inputs[ in_no]; |
168 | AVFilterLink *outlink = ctx->outputs[out_no]; |
169 | struct concat_in *in = &cat->in[in_no]; |
170 | |
171 | buf->pts = av_rescale_q(buf->pts, inlink->time_base, outlink->time_base); |
172 | in->pts = buf->pts; |
173 | in->nb_frames++; |
174 | /* add duration to input PTS */ |
175 | if (inlink->sample_rate) |
176 | /* use number of audio samples */ |
177 | in->pts += av_rescale_q(buf->nb_samples, |
178 | av_make_q(1, inlink->sample_rate), |
179 | outlink->time_base); |
180 | else if (in->nb_frames >= 2) |
181 | /* use mean duration */ |
182 | in->pts = av_rescale(in->pts, in->nb_frames, in->nb_frames - 1); |
183 | |
184 | buf->pts += cat->delta_ts; |
185 | return ff_filter_frame(outlink, buf); |
186 | } |
187 | |
188 | static int process_frame(AVFilterLink *inlink, AVFrame *buf) |
189 | { |
190 | AVFilterContext *ctx = inlink->dst; |
191 | ConcatContext *cat = ctx->priv; |
192 | unsigned in_no = FF_INLINK_IDX(inlink); |
193 | |
194 | if (in_no < cat->cur_idx) { |
195 | av_log(ctx, AV_LOG_ERROR, "Frame after EOF on input %s\n", |
196 | ctx->input_pads[in_no].name); |
197 | av_frame_free(&buf); |
198 | } else if (in_no >= cat->cur_idx + ctx->nb_outputs) { |
199 | ff_bufqueue_add(ctx, &cat->in[in_no].queue, buf); |
200 | } else { |
201 | return push_frame(ctx, in_no, buf); |
202 | } |
203 | return 0; |
204 | } |
205 | |
206 | static AVFrame *get_video_buffer(AVFilterLink *inlink, int w, int h) |
207 | { |
208 | AVFilterContext *ctx = inlink->dst; |
209 | unsigned in_no = FF_INLINK_IDX(inlink); |
210 | AVFilterLink *outlink = ctx->outputs[in_no % ctx->nb_outputs]; |
211 | |
212 | return ff_get_video_buffer(outlink, w, h); |
213 | } |
214 | |
215 | static AVFrame *get_audio_buffer(AVFilterLink *inlink, int nb_samples) |
216 | { |
217 | AVFilterContext *ctx = inlink->dst; |
218 | unsigned in_no = FF_INLINK_IDX(inlink); |
219 | AVFilterLink *outlink = ctx->outputs[in_no % ctx->nb_outputs]; |
220 | |
221 | return ff_get_audio_buffer(outlink, nb_samples); |
222 | } |
223 | |
224 | static int filter_frame(AVFilterLink *inlink, AVFrame *buf) |
225 | { |
226 | return process_frame(inlink, buf); |
227 | } |
228 | |
229 | static void close_input(AVFilterContext *ctx, unsigned in_no) |
230 | { |
231 | ConcatContext *cat = ctx->priv; |
232 | |
233 | cat->in[in_no].eof = 1; |
234 | cat->nb_in_active--; |
235 | av_log(ctx, AV_LOG_VERBOSE, "EOF on %s, %d streams left in segment.\n", |
236 | ctx->input_pads[in_no].name, cat->nb_in_active); |
237 | } |
238 | |
239 | static void find_next_delta_ts(AVFilterContext *ctx, int64_t *seg_delta) |
240 | { |
241 | ConcatContext *cat = ctx->priv; |
242 | unsigned i = cat->cur_idx; |
243 | unsigned imax = i + ctx->nb_outputs; |
244 | int64_t pts; |
245 | |
246 | pts = cat->in[i++].pts; |
247 | for (; i < imax; i++) |
248 | pts = FFMAX(pts, cat->in[i].pts); |
249 | cat->delta_ts += pts; |
250 | *seg_delta = pts; |
251 | } |
252 | |
253 | static int send_silence(AVFilterContext *ctx, unsigned in_no, unsigned out_no, |
254 | int64_t seg_delta) |
255 | { |
256 | ConcatContext *cat = ctx->priv; |
257 | AVFilterLink *outlink = ctx->outputs[out_no]; |
258 | int64_t base_pts = cat->in[in_no].pts + cat->delta_ts - seg_delta; |
259 | int64_t nb_samples, sent = 0; |
260 | int frame_nb_samples, ret; |
261 | AVRational rate_tb = { 1, ctx->inputs[in_no]->sample_rate }; |
262 | AVFrame *buf; |
263 | |
264 | if (!rate_tb.den) |
265 | return AVERROR_BUG; |
266 | nb_samples = av_rescale_q(seg_delta - cat->in[in_no].pts, |
267 | outlink->time_base, rate_tb); |
268 | frame_nb_samples = FFMAX(9600, rate_tb.den / 5); /* arbitrary */ |
269 | while (nb_samples) { |
270 | frame_nb_samples = FFMIN(frame_nb_samples, nb_samples); |
271 | buf = ff_get_audio_buffer(outlink, frame_nb_samples); |
272 | if (!buf) |
273 | return AVERROR(ENOMEM); |
274 | av_samples_set_silence(buf->extended_data, 0, frame_nb_samples, |
275 | outlink->channels, outlink->format); |
276 | buf->pts = base_pts + av_rescale_q(sent, rate_tb, outlink->time_base); |
277 | ret = ff_filter_frame(outlink, buf); |
278 | if (ret < 0) |
279 | return ret; |
280 | sent += frame_nb_samples; |
281 | nb_samples -= frame_nb_samples; |
282 | } |
283 | return 0; |
284 | } |
285 | |
286 | static int flush_segment(AVFilterContext *ctx) |
287 | { |
288 | int ret; |
289 | ConcatContext *cat = ctx->priv; |
290 | unsigned str, str_max; |
291 | int64_t seg_delta; |
292 | |
293 | find_next_delta_ts(ctx, &seg_delta); |
294 | cat->cur_idx += ctx->nb_outputs; |
295 | cat->nb_in_active = ctx->nb_outputs; |
296 | av_log(ctx, AV_LOG_VERBOSE, "Segment finished at pts=%"PRId64"\n", |
297 | cat->delta_ts); |
298 | |
299 | if (cat->cur_idx < ctx->nb_inputs) { |
300 | /* pad audio streams with silence */ |
301 | str = cat->nb_streams[AVMEDIA_TYPE_VIDEO]; |
302 | str_max = str + cat->nb_streams[AVMEDIA_TYPE_AUDIO]; |
303 | for (; str < str_max; str++) { |
304 | ret = send_silence(ctx, cat->cur_idx - ctx->nb_outputs + str, str, |
305 | seg_delta); |
306 | if (ret < 0) |
307 | return ret; |
308 | } |
309 | /* flush queued buffers */ |
310 | /* possible enhancement: flush in PTS order */ |
311 | str_max = cat->cur_idx + ctx->nb_outputs; |
312 | for (str = cat->cur_idx; str < str_max; str++) { |
313 | while (cat->in[str].queue.available) { |
314 | ret = push_frame(ctx, str, ff_bufqueue_get(&cat->in[str].queue)); |
315 | if (ret < 0) |
316 | return ret; |
317 | } |
318 | } |
319 | } |
320 | return 0; |
321 | } |
322 | |
323 | static int request_frame(AVFilterLink *outlink) |
324 | { |
325 | AVFilterContext *ctx = outlink->src; |
326 | ConcatContext *cat = ctx->priv; |
327 | unsigned out_no = FF_OUTLINK_IDX(outlink); |
328 | unsigned in_no = out_no + cat->cur_idx; |
329 | unsigned str, str_max; |
330 | int ret; |
331 | |
332 | while (1) { |
333 | if (in_no >= ctx->nb_inputs) |
334 | return AVERROR_EOF; |
335 | if (!cat->in[in_no].eof) { |
336 | ret = ff_request_frame(ctx->inputs[in_no]); |
337 | if (ret != AVERROR_EOF) |
338 | return ret; |
339 | close_input(ctx, in_no); |
340 | } |
341 | /* cycle on all inputs to finish the segment */ |
342 | /* possible enhancement: request in PTS order */ |
343 | str_max = cat->cur_idx + ctx->nb_outputs - 1; |
344 | for (str = cat->cur_idx; cat->nb_in_active; |
345 | str = str == str_max ? cat->cur_idx : str + 1) { |
346 | if (cat->in[str].eof) |
347 | continue; |
348 | ret = ff_request_frame(ctx->inputs[str]); |
349 | if (ret != AVERROR_EOF) |
350 | return ret; |
351 | close_input(ctx, str); |
352 | } |
353 | ret = flush_segment(ctx); |
354 | if (ret < 0) |
355 | return ret; |
356 | in_no += ctx->nb_outputs; |
357 | } |
358 | } |
359 | |
360 | static av_cold int init(AVFilterContext *ctx) |
361 | { |
362 | ConcatContext *cat = ctx->priv; |
363 | unsigned seg, type, str; |
364 | |
365 | /* create input pads */ |
366 | for (seg = 0; seg < cat->nb_segments; seg++) { |
367 | for (type = 0; type < TYPE_ALL; type++) { |
368 | for (str = 0; str < cat->nb_streams[type]; str++) { |
369 | AVFilterPad pad = { |
370 | .type = type, |
371 | .get_video_buffer = get_video_buffer, |
372 | .get_audio_buffer = get_audio_buffer, |
373 | .filter_frame = filter_frame, |
374 | }; |
375 | pad.name = av_asprintf("in%d:%c%d", seg, "va"[type], str); |
376 | ff_insert_inpad(ctx, ctx->nb_inputs, &pad); |
377 | } |
378 | } |
379 | } |
380 | /* create output pads */ |
381 | for (type = 0; type < TYPE_ALL; type++) { |
382 | for (str = 0; str < cat->nb_streams[type]; str++) { |
383 | AVFilterPad pad = { |
384 | .type = type, |
385 | .config_props = config_output, |
386 | .request_frame = request_frame, |
387 | }; |
388 | pad.name = av_asprintf("out:%c%d", "va"[type], str); |
389 | ff_insert_outpad(ctx, ctx->nb_outputs, &pad); |
390 | } |
391 | } |
392 | |
393 | cat->in = av_calloc(ctx->nb_inputs, sizeof(*cat->in)); |
394 | if (!cat->in) |
395 | return AVERROR(ENOMEM); |
396 | cat->nb_in_active = ctx->nb_outputs; |
397 | return 0; |
398 | } |
399 | |
400 | static av_cold void uninit(AVFilterContext *ctx) |
401 | { |
402 | ConcatContext *cat = ctx->priv; |
403 | unsigned i; |
404 | |
405 | for (i = 0; i < ctx->nb_inputs; i++) { |
406 | av_freep(&ctx->input_pads[i].name); |
407 | ff_bufqueue_discard_all(&cat->in[i].queue); |
408 | } |
409 | for (i = 0; i < ctx->nb_outputs; i++) |
410 | av_freep(&ctx->output_pads[i].name); |
411 | av_freep(&cat->in); |
412 | } |
413 | |
414 | AVFilter ff_avf_concat = { |
415 | .name = "concat", |
416 | .description = NULL_IF_CONFIG_SMALL("Concatenate audio and video streams."), |
417 | .init = init, |
418 | .uninit = uninit, |
419 | .query_formats = query_formats, |
420 | .priv_size = sizeof(ConcatContext), |
421 | .inputs = NULL, |
422 | .outputs = NULL, |
423 | .priv_class = &concat_class, |
424 | .flags = AVFILTER_FLAG_DYNAMIC_INPUTS | AVFILTER_FLAG_DYNAMIC_OUTPUTS, |
425 | }; |
426 |