2 * This file is part of FFmpeg.
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 * Join multiple audio inputs as different channels in
27 #include "libavutil/avassert.h"
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/common.h"
30 #include "libavutil/opt.h"
37 typedef struct ChannelMap {
38 int input; ///< input stream index
39 int in_channel_idx; ///< index of in_channel in the input stream data
40 uint64_t in_channel; ///< layout describing the input channel
41 uint64_t out_channel; ///< layout describing the output channel
44 typedef struct JoinContext {
49 char *channel_layout_str;
50 uint64_t channel_layout;
56 * Temporary storage for input frames, until we get one on each input.
58 AVFrame **input_frames;
61 * Temporary storage for buffer references, for assembling the output frame.
63 AVBufferRef **buffers;
66 #define OFFSET(x) offsetof(JoinContext, x)
67 #define A AV_OPT_FLAG_AUDIO_PARAM
68 #define F AV_OPT_FLAG_FILTERING_PARAM
69 static const AVOption join_options[] = {
70 { "inputs", "Number of input streams.", OFFSET(inputs), AV_OPT_TYPE_INT, { .i64 = 2 }, 1, INT_MAX, A|F },
71 { "channel_layout", "Channel layout of the "
72 "output stream.", OFFSET(channel_layout_str), AV_OPT_TYPE_STRING, {.str = "stereo"}, 0, 0, A|F },
73 { "map", "A comma-separated list of channels maps in the format "
74 "'input_stream.input_channel-output_channel.",
75 OFFSET(map), AV_OPT_TYPE_STRING, .flags = A|F },
79 AVFILTER_DEFINE_CLASS(join);
81 static int filter_frame(AVFilterLink *link, AVFrame *frame)
83 AVFilterContext *ctx = link->dst;
84 JoinContext *s = ctx->priv;
87 for (i = 0; i < ctx->nb_inputs; i++)
88 if (link == ctx->inputs[i])
90 av_assert0(i < ctx->nb_inputs);
91 av_assert0(!s->input_frames[i]);
92 s->input_frames[i] = frame;
97 static int parse_maps(AVFilterContext *ctx)
99 JoinContext *s = ctx->priv;
100 char separator = '|';
103 #if FF_API_OLD_FILTER_OPTS
104 if (cur && strchr(cur, ',')) {
105 av_log(ctx, AV_LOG_WARNING, "This syntax is deprecated, use '|' to "
106 "separate the mappings.\n");
111 while (cur && *cur) {
112 char *sep, *next, *p;
113 uint64_t in_channel = 0, out_channel = 0;
114 int input_idx, out_ch_idx, in_ch_idx;
116 next = strchr(cur, separator);
120 /* split the map into input and output parts */
121 if (!(sep = strchr(cur, '-'))) {
122 av_log(ctx, AV_LOG_ERROR, "Missing separator '-' in channel "
124 return AVERROR(EINVAL);
128 #define PARSE_CHANNEL(str, var, inout) \
129 if (!(var = av_get_channel_layout(str))) { \
130 av_log(ctx, AV_LOG_ERROR, "Invalid " inout " channel: %s.\n", str);\
131 return AVERROR(EINVAL); \
133 if (av_get_channel_layout_nb_channels(var) != 1) { \
134 av_log(ctx, AV_LOG_ERROR, "Channel map describes more than one " \
135 inout " channel.\n"); \
136 return AVERROR(EINVAL); \
139 /* parse output channel */
140 PARSE_CHANNEL(sep, out_channel, "output");
141 if (!(out_channel & s->channel_layout)) {
142 av_log(ctx, AV_LOG_ERROR, "Output channel '%s' is not present in "
143 "requested channel layout.\n", sep);
144 return AVERROR(EINVAL);
147 out_ch_idx = av_get_channel_layout_channel_index(s->channel_layout,
149 if (s->channels[out_ch_idx].input >= 0) {
150 av_log(ctx, AV_LOG_ERROR, "Multiple maps for output channel "
152 return AVERROR(EINVAL);
155 /* parse input channel */
156 input_idx = strtol(cur, &cur, 0);
157 if (input_idx < 0 || input_idx >= s->inputs) {
158 av_log(ctx, AV_LOG_ERROR, "Invalid input stream index: %d.\n",
160 return AVERROR(EINVAL);
166 in_ch_idx = strtol(cur, &p, 0);
168 /* channel specifier is not a number,
169 * try to parse as channel name */
170 PARSE_CHANNEL(cur, in_channel, "input");
173 s->channels[out_ch_idx].input = input_idx;
175 s->channels[out_ch_idx].in_channel = in_channel;
177 s->channels[out_ch_idx].in_channel_idx = in_ch_idx;
184 static av_cold int join_init(AVFilterContext *ctx)
186 JoinContext *s = ctx->priv;
189 if (!(s->channel_layout = av_get_channel_layout(s->channel_layout_str))) {
190 av_log(ctx, AV_LOG_ERROR, "Error parsing channel layout '%s'.\n",
191 s->channel_layout_str);
192 return AVERROR(EINVAL);
195 s->nb_channels = av_get_channel_layout_nb_channels(s->channel_layout);
196 s->channels = av_mallocz_array(s->nb_channels, sizeof(*s->channels));
197 s->buffers = av_mallocz_array(s->nb_channels, sizeof(*s->buffers));
198 s->input_frames = av_mallocz_array(s->inputs, sizeof(*s->input_frames));
199 if (!s->channels || !s->buffers|| !s->input_frames)
200 return AVERROR(ENOMEM);
202 for (i = 0; i < s->nb_channels; i++) {
203 s->channels[i].out_channel = av_channel_layout_extract_channel(s->channel_layout, i);
204 s->channels[i].input = -1;
207 if ((ret = parse_maps(ctx)) < 0)
210 for (i = 0; i < s->inputs; i++) {
212 AVFilterPad pad = { 0 };
214 snprintf(name, sizeof(name), "input%d", i);
215 pad.type = AVMEDIA_TYPE_AUDIO;
216 pad.name = av_strdup(name);
218 return AVERROR(ENOMEM);
219 pad.filter_frame = filter_frame;
223 ff_insert_inpad(ctx, i, &pad);
229 static av_cold void join_uninit(AVFilterContext *ctx)
231 JoinContext *s = ctx->priv;
234 for (i = 0; i < ctx->nb_inputs; i++) {
235 av_freep(&ctx->input_pads[i].name);
236 av_frame_free(&s->input_frames[i]);
239 av_freep(&s->channels);
240 av_freep(&s->buffers);
241 av_freep(&s->input_frames);
244 static int join_query_formats(AVFilterContext *ctx)
246 JoinContext *s = ctx->priv;
247 AVFilterChannelLayouts *layouts = NULL;
250 if ((ret = ff_add_channel_layout(&layouts, s->channel_layout)) < 0 ||
251 (ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts)) < 0)
254 for (i = 0; i < ctx->nb_inputs; i++) {
255 layouts = ff_all_channel_layouts();
256 if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts)) < 0)
260 if ((ret = ff_set_common_formats(ctx, ff_planar_sample_fmts())) < 0 ||
261 (ret = ff_set_common_samplerates(ctx, ff_all_samplerates())) < 0)
267 static void guess_map_matching(AVFilterContext *ctx, ChannelMap *ch,
272 for (i = 0; i < ctx->nb_inputs; i++) {
273 AVFilterLink *link = ctx->inputs[i];
275 if (ch->out_channel & link->channel_layout &&
276 !(ch->out_channel & inputs[i])) {
278 ch->in_channel = ch->out_channel;
279 inputs[i] |= ch->out_channel;
285 static void guess_map_any(AVFilterContext *ctx, ChannelMap *ch,
290 for (i = 0; i < ctx->nb_inputs; i++) {
291 AVFilterLink *link = ctx->inputs[i];
293 if ((inputs[i] & link->channel_layout) != link->channel_layout) {
294 uint64_t unused = link->channel_layout & ~inputs[i];
297 ch->in_channel = av_channel_layout_extract_channel(unused, 0);
298 inputs[i] |= ch->in_channel;
304 static int join_config_output(AVFilterLink *outlink)
306 AVFilterContext *ctx = outlink->src;
307 JoinContext *s = ctx->priv;
308 uint64_t *inputs; // nth element tracks which channels are used from nth input
311 /* initialize inputs to user-specified mappings */
312 if (!(inputs = av_mallocz_array(ctx->nb_inputs, sizeof(*inputs))))
313 return AVERROR(ENOMEM);
314 for (i = 0; i < s->nb_channels; i++) {
315 ChannelMap *ch = &s->channels[i];
316 AVFilterLink *inlink;
321 inlink = ctx->inputs[ch->input];
324 ch->in_channel = av_channel_layout_extract_channel(inlink->channel_layout,
327 if (!(ch->in_channel & inlink->channel_layout)) {
328 av_log(ctx, AV_LOG_ERROR, "Requested channel %s is not present in "
329 "input stream #%d.\n", av_get_channel_name(ch->in_channel),
331 ret = AVERROR(EINVAL);
335 inputs[ch->input] |= ch->in_channel;
338 /* guess channel maps when not explicitly defined */
339 /* first try unused matching channels */
340 for (i = 0; i < s->nb_channels; i++) {
341 ChannelMap *ch = &s->channels[i];
344 guess_map_matching(ctx, ch, inputs);
347 /* if the above failed, try to find _any_ unused input channel */
348 for (i = 0; i < s->nb_channels; i++) {
349 ChannelMap *ch = &s->channels[i];
352 guess_map_any(ctx, ch, inputs);
355 av_log(ctx, AV_LOG_ERROR, "Could not find input channel for "
356 "output channel '%s'.\n",
357 av_get_channel_name(ch->out_channel));
361 ch->in_channel_idx = av_get_channel_layout_channel_index(ctx->inputs[ch->input]->channel_layout,
366 av_log(ctx, AV_LOG_VERBOSE, "mappings: ");
367 for (i = 0; i < s->nb_channels; i++) {
368 ChannelMap *ch = &s->channels[i];
369 av_log(ctx, AV_LOG_VERBOSE, "%d.%s => %s ", ch->input,
370 av_get_channel_name(ch->in_channel),
371 av_get_channel_name(ch->out_channel));
373 av_log(ctx, AV_LOG_VERBOSE, "\n");
375 for (i = 0; i < ctx->nb_inputs; i++) {
377 av_log(ctx, AV_LOG_WARNING, "No channels are used from input "
386 static int join_request_frame(AVFilterLink *outlink)
388 AVFilterContext *ctx = outlink->src;
389 JoinContext *s = ctx->priv;
391 int linesize = INT_MAX;
396 /* get a frame on each input */
397 for (i = 0; i < ctx->nb_inputs; i++) {
398 AVFilterLink *inlink = ctx->inputs[i];
400 if (!s->input_frames[i] &&
401 (ret = ff_request_frame(inlink)) < 0)
404 /* request the same number of samples on all inputs */
406 nb_samples = s->input_frames[0]->nb_samples;
408 for (j = 1; !i && j < ctx->nb_inputs; j++)
409 ctx->inputs[j]->request_samples = nb_samples;
413 /* setup the output frame */
414 frame = av_frame_alloc();
416 return AVERROR(ENOMEM);
417 if (s->nb_channels > FF_ARRAY_ELEMS(frame->data)) {
418 frame->extended_data = av_mallocz_array(s->nb_channels,
419 sizeof(*frame->extended_data));
420 if (!frame->extended_data) {
421 ret = AVERROR(ENOMEM);
426 /* copy the data pointers */
427 for (i = 0; i < s->nb_channels; i++) {
428 ChannelMap *ch = &s->channels[i];
429 AVFrame *cur = s->input_frames[ch->input];
432 frame->extended_data[i] = cur->extended_data[ch->in_channel_idx];
433 linesize = FFMIN(linesize, cur->linesize[0]);
435 /* add the buffer where this plan is stored to the list if it's
436 * not already there */
437 buf = av_frame_get_plane_buffer(cur, ch->in_channel_idx);
439 ret = AVERROR(EINVAL);
442 for (j = 0; j < nb_buffers; j++)
443 if (s->buffers[j]->buffer == buf->buffer)
446 s->buffers[nb_buffers++] = buf;
449 /* create references to the buffers we copied to output */
450 if (nb_buffers > FF_ARRAY_ELEMS(frame->buf)) {
451 frame->nb_extended_buf = nb_buffers - FF_ARRAY_ELEMS(frame->buf);
452 frame->extended_buf = av_mallocz_array(frame->nb_extended_buf,
453 sizeof(*frame->extended_buf));
454 if (!frame->extended_buf) {
455 frame->nb_extended_buf = 0;
456 ret = AVERROR(ENOMEM);
460 for (i = 0; i < FFMIN(FF_ARRAY_ELEMS(frame->buf), nb_buffers); i++) {
461 frame->buf[i] = av_buffer_ref(s->buffers[i]);
462 if (!frame->buf[i]) {
463 ret = AVERROR(ENOMEM);
467 for (i = 0; i < frame->nb_extended_buf; i++) {
468 frame->extended_buf[i] = av_buffer_ref(s->buffers[i +
469 FF_ARRAY_ELEMS(frame->buf)]);
470 if (!frame->extended_buf[i]) {
471 ret = AVERROR(ENOMEM);
476 frame->nb_samples = nb_samples;
477 frame->channel_layout = outlink->channel_layout;
478 av_frame_set_channels(frame, outlink->channels);
479 frame->sample_rate = outlink->sample_rate;
480 frame->format = outlink->format;
481 frame->pts = s->input_frames[0]->pts;
482 frame->linesize[0] = linesize;
483 if (frame->data != frame->extended_data) {
484 memcpy(frame->data, frame->extended_data, sizeof(*frame->data) *
485 FFMIN(FF_ARRAY_ELEMS(frame->data), s->nb_channels));
488 ret = ff_filter_frame(outlink, frame);
490 for (i = 0; i < ctx->nb_inputs; i++)
491 av_frame_free(&s->input_frames[i]);
496 av_frame_free(&frame);
500 static const AVFilterPad avfilter_af_join_outputs[] = {
503 .type = AVMEDIA_TYPE_AUDIO,
504 .config_props = join_config_output,
505 .request_frame = join_request_frame,
510 AVFilter ff_af_join = {
512 .description = NULL_IF_CONFIG_SMALL("Join multiple audio streams into "
513 "multi-channel output."),
514 .priv_size = sizeof(JoinContext),
515 .priv_class = &join_class,
517 .uninit = join_uninit,
518 .query_formats = join_query_formats,
520 .outputs = avfilter_af_join_outputs,
521 .flags = AVFILTER_FLAG_DYNAMIC_INPUTS,