2 * Copyright (c) 2012 Nicolas George
4 * This file is part of FFmpeg.
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.
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.
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
23 * concat audio-video filter
26 #include "libavutil/audioconvert.h"
27 #include "libavutil/avassert.h"
28 #include "libavutil/opt.h"
30 #define FF_BUFQUEUE_SIZE 256
31 #include "bufferqueue.h"
40 unsigned nb_streams[TYPE_ALL]; /**< number of out streams of each type */
42 unsigned cur_idx; /**< index of the first input of current segment */
43 int64_t delta_ts; /**< timestamp to add to produce output timestamps */
44 unsigned nb_in_active; /**< number of active inputs in current segment */
49 struct FFBufQueue queue;
53 #define OFFSET(x) offsetof(ConcatContext, x)
55 static const AVOption concat_options[] = {
56 { "n", "specify the number of segments", OFFSET(nb_segments),
57 AV_OPT_TYPE_INT, { .dbl = 2 }, 2, INT_MAX },
58 { "v", "specify the number of video streams",
59 OFFSET(nb_streams[AVMEDIA_TYPE_VIDEO]),
60 AV_OPT_TYPE_INT, { .dbl = 1 }, 0, INT_MAX },
61 { "a", "specify the number of audio streams",
62 OFFSET(nb_streams[AVMEDIA_TYPE_AUDIO]),
63 AV_OPT_TYPE_INT, { .dbl = 0 }, 0, INT_MAX },
67 AVFILTER_DEFINE_CLASS(concat);
69 static int query_formats(AVFilterContext *ctx)
71 ConcatContext *cat = ctx->priv;
72 unsigned type, nb_str, idx0 = 0, idx, str, seg;
73 AVFilterFormats *formats, *rates;
74 AVFilterChannelLayouts *layouts;
76 for (type = 0; type < TYPE_ALL; type++) {
77 nb_str = cat->nb_streams[type];
78 for (str = 0; str < nb_str; str++) {
81 /* Set the output formats */
82 formats = ff_all_formats(type);
84 return AVERROR(ENOMEM);
85 ff_formats_ref(formats, &ctx->outputs[idx]->in_formats);
86 if (type == AVMEDIA_TYPE_AUDIO) {
87 rates = ff_all_samplerates();
89 return AVERROR(ENOMEM);
90 ff_formats_ref(rates, &ctx->outputs[idx]->in_samplerates);
91 layouts = ff_all_channel_layouts();
93 return AVERROR(ENOMEM);
94 ff_channel_layouts_ref(layouts, &ctx->outputs[idx]->in_channel_layouts);
97 /* Set the same formats for each corresponding input */
98 for (seg = 0; seg < cat->nb_segments; seg++) {
99 ff_formats_ref(formats, &ctx->inputs[idx]->out_formats);
100 if (type == AVMEDIA_TYPE_AUDIO) {
101 ff_formats_ref(rates, &ctx->inputs[idx]->out_samplerates);
102 ff_channel_layouts_ref(layouts, &ctx->inputs[idx]->out_channel_layouts);
104 idx += ctx->nb_outputs;
113 static int config_output(AVFilterLink *outlink)
115 AVFilterContext *ctx = outlink->src;
116 ConcatContext *cat = ctx->priv;
117 unsigned out_no = FF_OUTLINK_IDX(outlink);
118 unsigned in_no = out_no, seg;
119 AVFilterLink *inlink = ctx->inputs[in_no];
121 /* enhancement: find a common one */
122 outlink->time_base = AV_TIME_BASE_Q;
123 outlink->w = inlink->w;
124 outlink->h = inlink->h;
125 outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
126 outlink->format = inlink->format;
127 for (seg = 1; seg < cat->nb_segments; seg++) {
128 inlink = ctx->inputs[in_no += ctx->nb_outputs];
129 /* possible enhancement: unsafe mode, do not check */
130 if (outlink->w != inlink->w ||
131 outlink->h != inlink->h ||
132 outlink->sample_aspect_ratio.num != inlink->sample_aspect_ratio.num ||
133 outlink->sample_aspect_ratio.den != inlink->sample_aspect_ratio.den) {
134 av_log(ctx, AV_LOG_ERROR, "Input link %s parameters "
135 "(size %dx%d, SAR %d:%d) do not match the corresponding "
136 "output link %s parameters (%dx%d, SAR %d:%d)\n",
137 ctx->input_pads[in_no].name, inlink->w, inlink->h,
138 inlink->sample_aspect_ratio.num,
139 inlink->sample_aspect_ratio.den,
140 ctx->input_pads[out_no].name, outlink->w, outlink->h,
141 outlink->sample_aspect_ratio.num,
142 outlink->sample_aspect_ratio.den);
143 return AVERROR(EINVAL);
150 static void push_frame(AVFilterContext *ctx, unsigned in_no,
151 AVFilterBufferRef *buf)
153 ConcatContext *cat = ctx->priv;
154 unsigned out_no = in_no % ctx->nb_outputs;
155 AVFilterLink * inlink = ctx-> inputs[ in_no];
156 AVFilterLink *outlink = ctx->outputs[out_no];
157 struct concat_in *in = &cat->in[in_no];
159 buf->pts = av_rescale_q(buf->pts, inlink->time_base, outlink->time_base);
162 /* add duration to input PTS */
163 if (inlink->sample_rate)
164 /* use number of audio samples */
165 in->pts += av_rescale_q(buf->audio->nb_samples,
166 (AVRational){ 1, inlink->sample_rate },
168 else if (in->nb_frames >= 2)
169 /* use mean duration */
170 in->pts = av_rescale(in->pts, in->nb_frames, in->nb_frames - 1);
172 buf->pts += cat->delta_ts;
174 case AVMEDIA_TYPE_VIDEO:
175 ff_start_frame(outlink, buf);
176 ff_draw_slice(outlink, 0, outlink->h, 1);
177 ff_end_frame(outlink);
179 case AVMEDIA_TYPE_AUDIO:
180 ff_filter_samples(outlink, buf);
185 static void process_frame(AVFilterLink *inlink, AVFilterBufferRef *buf)
187 AVFilterContext *ctx = inlink->dst;
188 ConcatContext *cat = ctx->priv;
189 unsigned in_no = FF_INLINK_IDX(inlink);
191 if (in_no < cat->cur_idx) {
192 av_log(ctx, AV_LOG_ERROR, "Frame after EOF on input %s\n",
193 ctx->input_pads[in_no].name);
194 avfilter_unref_buffer(buf);
195 } if (in_no >= cat->cur_idx + ctx->nb_outputs) {
196 ff_bufqueue_add(ctx, &cat->in[in_no].queue, buf);
198 push_frame(ctx, in_no, buf);
202 static AVFilterBufferRef *get_video_buffer(AVFilterLink *inlink, int perms,
205 AVFilterContext *ctx = inlink->dst;
206 unsigned in_no = FF_INLINK_IDX(inlink);
207 AVFilterLink *outlink = ctx->outputs[in_no % ctx->nb_outputs];
209 return ff_get_video_buffer(outlink, perms, w, h);
212 static AVFilterBufferRef *get_audio_buffer(AVFilterLink *inlink, int perms,
215 AVFilterContext *ctx = inlink->dst;
216 unsigned in_no = FF_INLINK_IDX(inlink);
217 AVFilterLink *outlink = ctx->outputs[in_no % ctx->nb_outputs];
219 return ff_get_audio_buffer(outlink, perms, nb_samples);
222 static int start_frame(AVFilterLink *inlink, AVFilterBufferRef *buf)
227 static int draw_slice(AVFilterLink *inlink, int y, int h, int dir)
232 static int end_frame(AVFilterLink *inlink)
234 process_frame(inlink, inlink->cur_buf);
235 inlink->cur_buf = NULL;
239 static int filter_samples(AVFilterLink *inlink, AVFilterBufferRef *buf)
241 process_frame(inlink, buf);
242 return 0; /* enhancement: handle error return */
245 static void close_input(AVFilterContext *ctx, unsigned in_no)
247 ConcatContext *cat = ctx->priv;
249 cat->in[in_no].eof = 1;
251 av_log(ctx, AV_LOG_VERBOSE, "EOF on %s, %d streams left in segment.\n",
252 ctx->input_pads[in_no].name, cat->nb_in_active);
255 static void find_next_delta_ts(AVFilterContext *ctx)
257 ConcatContext *cat = ctx->priv;
258 unsigned i = cat->cur_idx;
259 unsigned imax = i + ctx->nb_outputs;
262 pts = cat->in[i++].pts;
263 for (; i < imax; i++)
264 pts = FFMAX(pts, cat->in[i].pts);
265 cat->delta_ts += pts;
268 static void send_silence(AVFilterContext *ctx, unsigned in_no, unsigned out_no)
270 ConcatContext *cat = ctx->priv;
271 AVFilterLink *outlink = ctx->outputs[out_no];
272 int64_t base_pts = cat->in[in_no].pts + cat->delta_ts;
273 int64_t nb_samples, sent = 0;
274 int frame_nb_samples;
275 AVRational rate_tb = { 1, ctx->inputs[in_no]->sample_rate };
276 AVFilterBufferRef *buf;
277 int nb_channels = av_get_channel_layout_nb_channels(outlink->channel_layout);
281 nb_samples = av_rescale_q(cat->delta_ts - base_pts,
282 outlink->time_base, rate_tb);
283 frame_nb_samples = FFMAX(9600, rate_tb.den / 5); /* arbitrary */
285 frame_nb_samples = FFMIN(frame_nb_samples, nb_samples);
286 buf = ff_get_audio_buffer(outlink, AV_PERM_WRITE, frame_nb_samples);
289 av_samples_set_silence(buf->extended_data, 0, frame_nb_samples,
290 nb_channels, outlink->format);
291 buf->pts = base_pts + av_rescale_q(sent, rate_tb, outlink->time_base);
292 ff_filter_samples(outlink, buf);
293 sent += frame_nb_samples;
294 nb_samples -= frame_nb_samples;
298 static void flush_segment(AVFilterContext *ctx)
300 ConcatContext *cat = ctx->priv;
301 unsigned str, str_max;
303 find_next_delta_ts(ctx);
304 cat->cur_idx += ctx->nb_outputs;
305 cat->nb_in_active = ctx->nb_outputs;
306 av_log(ctx, AV_LOG_VERBOSE, "Segment finished at pts=%"PRId64"\n",
309 if (cat->cur_idx < ctx->nb_inputs) {
310 /* pad audio streams with silence */
311 str = cat->nb_streams[AVMEDIA_TYPE_VIDEO];
312 str_max = str + cat->nb_streams[AVMEDIA_TYPE_AUDIO];
313 for (; str < str_max; str++)
314 send_silence(ctx, cat->cur_idx - ctx->nb_outputs + str, str);
315 /* flush queued buffers */
316 /* possible enhancement: flush in PTS order */
317 str_max = cat->cur_idx + ctx->nb_outputs;
318 for (str = cat->cur_idx; str < str_max; str++)
319 while (cat->in[str].queue.available)
320 push_frame(ctx, str, ff_bufqueue_get(&cat->in[str].queue));
324 static int request_frame(AVFilterLink *outlink)
326 AVFilterContext *ctx = outlink->src;
327 ConcatContext *cat = ctx->priv;
328 unsigned out_no = FF_OUTLINK_IDX(outlink);
329 unsigned in_no = out_no + cat->cur_idx;
330 unsigned str, str_max;
334 if (in_no >= ctx->nb_inputs)
336 if (!cat->in[in_no].eof) {
337 ret = ff_request_frame(ctx->inputs[in_no]);
338 if (ret != AVERROR_EOF)
340 close_input(ctx, in_no);
342 /* cycle on all inputs to finish the segment */
343 /* possible enhancement: request in PTS order */
344 str_max = cat->cur_idx + ctx->nb_outputs - 1;
345 for (str = cat->cur_idx; cat->nb_in_active;
346 str = str == str_max ? cat->cur_idx : str + 1) {
347 if (cat->in[str].eof)
349 ret = ff_request_frame(ctx->inputs[str]);
350 if (ret == AVERROR_EOF)
351 close_input(ctx, str);
356 in_no += ctx->nb_outputs;
360 static av_cold int init(AVFilterContext *ctx, const char *args)
362 ConcatContext *cat = ctx->priv;
364 unsigned seg, type, str;
367 cat->class = &concat_class;
368 av_opt_set_defaults(cat);
369 ret = av_set_options_string(cat, args, "=", ":");
371 av_log(ctx, AV_LOG_ERROR, "Error parsing options: '%s'\n", args);
375 /* create input pads */
376 for (seg = 0; seg < cat->nb_segments; seg++) {
377 for (type = 0; type < TYPE_ALL; type++) {
378 for (str = 0; str < cat->nb_streams[type]; str++) {
381 .min_perms = AV_PERM_READ,
382 .rej_perms = AV_PERM_REUSE2,
383 .get_video_buffer = get_video_buffer,
384 .get_audio_buffer = get_audio_buffer,
386 snprintf(name, sizeof(name), "in%d:%c%d", seg, "va"[type], str);
387 pad.name = av_strdup(name);
388 if (type == AVMEDIA_TYPE_VIDEO) {
389 pad.start_frame = start_frame;
390 pad.draw_slice = draw_slice;
391 pad.end_frame = end_frame;
393 pad.filter_samples = filter_samples;
395 ff_insert_inpad(ctx, ctx->nb_inputs, &pad);
399 /* create output pads */
400 for (type = 0; type < TYPE_ALL; type++) {
401 for (str = 0; str < cat->nb_streams[type]; str++) {
404 .config_props = config_output,
405 .request_frame = request_frame,
407 snprintf(name, sizeof(name), "out:%c%d", "va"[type], str);
408 pad.name = av_strdup(name);
409 ff_insert_outpad(ctx, ctx->nb_outputs, &pad);
413 cat->in = av_calloc(ctx->nb_inputs, sizeof(*cat->in));
415 return AVERROR(ENOMEM);
416 cat->nb_in_active = ctx->nb_outputs;
420 static av_cold void uninit(AVFilterContext *ctx)
422 ConcatContext *cat = ctx->priv;
425 for (i = 0; i < ctx->nb_inputs; i++) {
426 av_freep(&ctx->input_pads[i].name);
427 ff_bufqueue_discard_all(&cat->in[i].queue);
429 for (i = 0; i < ctx->nb_outputs; i++)
430 av_freep(&ctx->output_pads[i].name);
434 AVFilter avfilter_avf_concat = {
436 .description = NULL_IF_CONFIG_SMALL("Concatenate audio and video streams."),
439 .query_formats = query_formats,
440 .priv_size = sizeof(ConcatContext),
441 .inputs = (const AVFilterPad[]) { { .name = NULL } },
442 .outputs = (const AVFilterPad[]) { { .name = NULL } },