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/avstring.h"
27 #include "libavutil/channel_layout.h"
28 #include "libavutil/opt.h"
38 typedef struct ConcatContext {
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 */
53 #define OFFSET(x) offsetof(ConcatContext, x)
54 #define A AV_OPT_FLAG_AUDIO_PARAM
55 #define F AV_OPT_FLAG_FILTERING_PARAM
56 #define V AV_OPT_FLAG_VIDEO_PARAM
58 static const AVOption concat_options[] = {
59 { "n", "specify the number of segments", OFFSET(nb_segments),
60 AV_OPT_TYPE_INT, { .i64 = 2 }, 1, INT_MAX, V|A|F},
61 { "v", "specify the number of video streams",
62 OFFSET(nb_streams[AVMEDIA_TYPE_VIDEO]),
63 AV_OPT_TYPE_INT, { .i64 = 1 }, 0, INT_MAX, V|F },
64 { "a", "specify the number of audio streams",
65 OFFSET(nb_streams[AVMEDIA_TYPE_AUDIO]),
66 AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, A|F},
67 { "unsafe", "enable unsafe mode",
69 AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, V|A|F},
73 AVFILTER_DEFINE_CLASS(concat);
75 static int query_formats(AVFilterContext *ctx)
77 ConcatContext *cat = ctx->priv;
78 unsigned type, nb_str, idx0 = 0, idx, str, seg;
79 AVFilterFormats *formats, *rates = NULL;
80 AVFilterChannelLayouts *layouts = NULL;
83 for (type = 0; type < TYPE_ALL; type++) {
84 nb_str = cat->nb_streams[type];
85 for (str = 0; str < nb_str; str++) {
88 /* Set the output formats */
89 formats = ff_all_formats(type);
90 if ((ret = ff_formats_ref(formats, &ctx->outputs[idx]->incfg.formats)) < 0)
93 if (type == AVMEDIA_TYPE_AUDIO) {
94 rates = ff_all_samplerates();
95 if ((ret = ff_formats_ref(rates, &ctx->outputs[idx]->incfg.samplerates)) < 0)
97 layouts = ff_all_channel_layouts();
98 if ((ret = ff_channel_layouts_ref(layouts, &ctx->outputs[idx]->incfg.channel_layouts)) < 0)
102 /* Set the same formats for each corresponding input */
103 for (seg = 0; seg < cat->nb_segments; seg++) {
104 if ((ret = ff_formats_ref(formats, &ctx->inputs[idx]->outcfg.formats)) < 0)
106 if (type == AVMEDIA_TYPE_AUDIO) {
107 if ((ret = ff_formats_ref(rates, &ctx->inputs[idx]->outcfg.samplerates)) < 0 ||
108 (ret = ff_channel_layouts_ref(layouts, &ctx->inputs[idx]->outcfg.channel_layouts)) < 0)
111 idx += ctx->nb_outputs;
120 static int config_output(AVFilterLink *outlink)
122 AVFilterContext *ctx = outlink->src;
123 ConcatContext *cat = ctx->priv;
124 unsigned out_no = FF_OUTLINK_IDX(outlink);
125 unsigned in_no = out_no, seg;
126 AVFilterLink *inlink = ctx->inputs[in_no];
128 /* enhancement: find a common one */
129 outlink->time_base = AV_TIME_BASE_Q;
130 outlink->w = inlink->w;
131 outlink->h = inlink->h;
132 outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
133 outlink->format = inlink->format;
134 outlink->frame_rate = inlink->frame_rate;
136 for (seg = 1; seg < cat->nb_segments; seg++) {
137 inlink = ctx->inputs[in_no + seg * ctx->nb_outputs];
138 if (outlink->frame_rate.num != inlink->frame_rate.num ||
139 outlink->frame_rate.den != inlink->frame_rate.den) {
140 av_log(ctx, AV_LOG_VERBOSE,
141 "Video inputs have different frame rates, output will be VFR\n");
142 outlink->frame_rate = av_make_q(1, 0);
147 for (seg = 1; seg < cat->nb_segments; seg++) {
148 inlink = ctx->inputs[in_no + seg * ctx->nb_outputs];
149 if (!outlink->sample_aspect_ratio.num)
150 outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
151 /* possible enhancement: unsafe mode, do not check */
152 if (outlink->w != inlink->w ||
153 outlink->h != inlink->h ||
154 outlink->sample_aspect_ratio.num != inlink->sample_aspect_ratio.num &&
155 inlink->sample_aspect_ratio.num ||
156 outlink->sample_aspect_ratio.den != inlink->sample_aspect_ratio.den) {
157 av_log(ctx, AV_LOG_ERROR, "Input link %s parameters "
158 "(size %dx%d, SAR %d:%d) do not match the corresponding "
159 "output link %s parameters (%dx%d, SAR %d:%d)\n",
160 ctx->input_pads[in_no].name, inlink->w, inlink->h,
161 inlink->sample_aspect_ratio.num,
162 inlink->sample_aspect_ratio.den,
163 ctx->input_pads[out_no].name, outlink->w, outlink->h,
164 outlink->sample_aspect_ratio.num,
165 outlink->sample_aspect_ratio.den);
167 return AVERROR(EINVAL);
174 static int push_frame(AVFilterContext *ctx, unsigned in_no, AVFrame *buf)
176 ConcatContext *cat = ctx->priv;
177 unsigned out_no = in_no % ctx->nb_outputs;
178 AVFilterLink * inlink = ctx-> inputs[ in_no];
179 AVFilterLink *outlink = ctx->outputs[out_no];
180 struct concat_in *in = &cat->in[in_no];
182 buf->pts = av_rescale_q(buf->pts, inlink->time_base, outlink->time_base);
183 buf->duration = av_rescale_q(buf->duration, inlink->time_base, outlink->time_base);
186 /* add duration to input PTS */
187 if (inlink->sample_rate)
188 /* use number of audio samples */
189 in->pts += av_rescale_q(buf->nb_samples,
190 av_make_q(1, inlink->sample_rate),
192 else if (in->nb_frames >= 2)
193 /* use mean duration */
194 in->pts = av_rescale(in->pts, in->nb_frames, in->nb_frames - 1);
196 buf->pts += cat->delta_ts;
197 return ff_filter_frame(outlink, buf);
200 static AVFrame *get_video_buffer(AVFilterLink *inlink, int w, int h)
202 AVFilterContext *ctx = inlink->dst;
203 unsigned in_no = FF_INLINK_IDX(inlink);
204 AVFilterLink *outlink = ctx->outputs[in_no % ctx->nb_outputs];
206 return ff_get_video_buffer(outlink, w, h);
209 static AVFrame *get_audio_buffer(AVFilterLink *inlink, int nb_samples)
211 AVFilterContext *ctx = inlink->dst;
212 unsigned in_no = FF_INLINK_IDX(inlink);
213 AVFilterLink *outlink = ctx->outputs[in_no % ctx->nb_outputs];
215 return ff_get_audio_buffer(outlink, nb_samples);
218 static void close_input(AVFilterContext *ctx, unsigned in_no)
220 ConcatContext *cat = ctx->priv;
222 cat->in[in_no].eof = 1;
224 av_log(ctx, AV_LOG_VERBOSE, "EOF on %s, %d streams left in segment.\n",
225 ctx->input_pads[in_no].name, cat->nb_in_active);
228 static void find_next_delta_ts(AVFilterContext *ctx, int64_t *seg_delta)
230 ConcatContext *cat = ctx->priv;
231 unsigned i = cat->cur_idx;
232 unsigned imax = i + ctx->nb_outputs;
235 pts = cat->in[i++].pts;
236 for (; i < imax; i++)
237 pts = FFMAX(pts, cat->in[i].pts);
238 cat->delta_ts += pts;
242 static int send_silence(AVFilterContext *ctx, unsigned in_no, unsigned out_no,
245 ConcatContext *cat = ctx->priv;
246 AVFilterLink *outlink = ctx->outputs[out_no];
247 int64_t base_pts = cat->in[in_no].pts + cat->delta_ts - seg_delta;
248 int64_t nb_samples, sent = 0;
249 int frame_nb_samples, ret;
250 AVRational rate_tb = { 1, ctx->inputs[in_no]->sample_rate };
255 if (cat->in[in_no].pts < INT64_MIN + seg_delta)
256 return AVERROR_INVALIDDATA;
257 if (seg_delta < cat->in[in_no].pts)
258 return AVERROR_INVALIDDATA;
259 nb_samples = av_rescale_q(seg_delta - cat->in[in_no].pts,
260 outlink->time_base, rate_tb);
261 frame_nb_samples = FFMAX(9600, rate_tb.den / 5); /* arbitrary */
263 frame_nb_samples = FFMIN(frame_nb_samples, nb_samples);
264 buf = ff_get_audio_buffer(outlink, frame_nb_samples);
266 return AVERROR(ENOMEM);
267 av_samples_set_silence(buf->extended_data, 0, frame_nb_samples,
268 outlink->ch_layout.nb_channels, outlink->format);
269 buf->pts = base_pts + av_rescale_q(sent, rate_tb, outlink->time_base);
270 ret = ff_filter_frame(outlink, buf);
273 sent += frame_nb_samples;
274 nb_samples -= frame_nb_samples;
279 static int flush_segment(AVFilterContext *ctx)
282 ConcatContext *cat = ctx->priv;
283 unsigned str, str_max;
286 find_next_delta_ts(ctx, &seg_delta);
287 cat->cur_idx += ctx->nb_outputs;
288 cat->nb_in_active = ctx->nb_outputs;
289 av_log(ctx, AV_LOG_VERBOSE, "Segment finished at pts=%"PRId64"\n",
292 if (cat->cur_idx < ctx->nb_inputs) {
293 /* pad audio streams with silence */
294 str = cat->nb_streams[AVMEDIA_TYPE_VIDEO];
295 str_max = str + cat->nb_streams[AVMEDIA_TYPE_AUDIO];
296 for (; str < str_max; str++) {
297 ret = send_silence(ctx, cat->cur_idx - ctx->nb_outputs + str, str,
306 static av_cold int init(AVFilterContext *ctx)
308 ConcatContext *cat = ctx->priv;
309 unsigned seg, type, str;
312 /* create input pads */
313 for (seg = 0; seg < cat->nb_segments; seg++) {
314 for (type = 0; type < TYPE_ALL; type++) {
315 for (str = 0; str < cat->nb_streams[type]; str++) {
319 if (type == AVMEDIA_TYPE_VIDEO)
320 pad.get_buffer.video = get_video_buffer;
322 pad.get_buffer.audio = get_audio_buffer;
323 pad.name = av_asprintf("in%d:%c%d", seg, "va"[type], str);
324 if ((ret = ff_append_inpad_free_name(ctx, &pad)) < 0)
329 /* create output pads */
330 for (type = 0; type < TYPE_ALL; type++) {
331 for (str = 0; str < cat->nb_streams[type]; str++) {
334 .config_props = config_output,
336 pad.name = av_asprintf("out:%c%d", "va"[type], str);
337 if ((ret = ff_append_outpad_free_name(ctx, &pad)) < 0)
342 cat->in = av_calloc(ctx->nb_inputs, sizeof(*cat->in));
344 return AVERROR(ENOMEM);
345 cat->nb_in_active = ctx->nb_outputs;
349 static av_cold void uninit(AVFilterContext *ctx)
351 ConcatContext *cat = ctx->priv;
356 static int activate(AVFilterContext *ctx)
358 ConcatContext *cat = ctx->priv;
364 /* Forward status back */
365 for (i = 0; i < ctx->nb_outputs; i++) {
366 status = ff_outlink_get_status(ctx->outputs[i]);
369 for (j = i; j < ctx->nb_inputs; j += ctx->nb_outputs) {
370 if (!cat->in[j].eof) {
372 ff_inlink_set_status(ctx->inputs[j], status);
379 /* Forward available frames */
380 if (cat->cur_idx < ctx->nb_inputs) {
381 for (i = 0; i < ctx->nb_outputs; i++) {
382 ret = ff_inlink_consume_frame(ctx->inputs[cat->cur_idx + i], &frame);
386 ff_filter_set_ready(ctx, 10);
387 return push_frame(ctx, cat->cur_idx + i, frame);
392 /* Forward status change */
393 if (cat->cur_idx < ctx->nb_inputs) {
394 for (i = 0; i < ctx->nb_outputs; i++) {
395 AVFilterLink *inlink = ctx->inputs[cat->cur_idx + i];
397 ret = ff_inlink_acknowledge_status(inlink, &status, &pts);
400 close_input(ctx, cat->cur_idx + i);
401 if (cat->cur_idx + ctx->nb_outputs >= ctx->nb_inputs) {
402 int64_t eof_pts = cat->delta_ts;
404 eof_pts += av_rescale_q(pts, inlink->time_base, ctx->outputs[i]->time_base);
405 ff_outlink_set_status(ctx->outputs[i], status, eof_pts);
407 if (!cat->nb_in_active) {
408 ret = flush_segment(ctx);
412 ff_filter_set_ready(ctx, 10);
418 ret = FFERROR_NOT_READY;
419 for (i = 0; i < ctx->nb_outputs; i++) {
420 if (ff_outlink_frame_wanted(ctx->outputs[i])) {
421 if (cat->in[cat->cur_idx + i].eof) {
422 for (j = 0; j < ctx->nb_outputs; j++)
423 if (!cat->in[cat->cur_idx + j].eof)
424 ff_inlink_request_frame(ctx->inputs[cat->cur_idx + j]);
427 ff_inlink_request_frame(ctx->inputs[cat->cur_idx + i]);
436 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
437 char *res, int res_len, int flags)
439 int ret = AVERROR(ENOSYS);
441 if (!strcmp(cmd, "next")) {
442 av_log(ctx, AV_LOG_VERBOSE, "Command received: next\n");
443 return flush_segment(ctx);
449 const AVFilter ff_avf_concat = {
451 .description = NULL_IF_CONFIG_SMALL("Concatenate audio and video streams."),
454 .activate = activate,
455 .priv_size = sizeof(ConcatContext),
458 .priv_class = &concat_class,
459 .flags = AVFILTER_FLAG_DYNAMIC_INPUTS | AVFILTER_FLAG_DYNAMIC_OUTPUTS,
460 FILTER_QUERY_FUNC(query_formats),
461 .process_command = process_command,