FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
af_pan.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2002 Anders Johansson <ajh@atri.curtin.edu.au>
3  * Copyright (c) 2011 Clément Bœsch <u pkh me>
4  * Copyright (c) 2011 Nicolas George <nicolas.george@normalesup.org>
5  *
6  * This file is part of FFmpeg.
7  *
8  * FFmpeg is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU Lesser General Public
10  * License as published by the Free Software Foundation; either
11  * version 2.1 of the License, or (at your option) any later version.
12  *
13  * FFmpeg is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16  * GNU Lesser General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public
19  * License along with FFmpeg; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  */
22 
23 /**
24  * @file
25  * Audio panning filter (channels mixing)
26  * Original code written by Anders Johansson for MPlayer,
27  * reimplemented for FFmpeg.
28  */
29 
30 #include <stdio.h>
31 #include "libavutil/avstring.h"
33 #include "libavutil/opt.h"
35 #include "audio.h"
36 #include "avfilter.h"
37 #include "formats.h"
38 #include "internal.h"
39 
40 #define MAX_CHANNELS 63
41 
42 typedef struct PanContext {
43  const AVClass *class;
44  char *args;
47  int64_t need_renorm;
50 
52  /* channel mapping specific */
54  struct SwrContext *swr;
55 } PanContext;
56 
57 static void skip_spaces(char **arg)
58 {
59  int len = 0;
60 
61  sscanf(*arg, " %n", &len);
62  *arg += len;
63 }
64 
65 static int parse_channel_name(char **arg, int *rchannel, int *rnamed)
66 {
67  char buf[8];
68  int len, i, channel_id = 0;
69  int64_t layout, layout0;
70 
71  skip_spaces(arg);
72  /* try to parse a channel name, e.g. "FL" */
73  if (sscanf(*arg, "%7[A-Z]%n", buf, &len)) {
74  layout0 = layout = av_get_channel_layout(buf);
75  /* channel_id <- first set bit in layout */
76  for (i = 32; i > 0; i >>= 1) {
77  if (layout >= (int64_t)1 << i) {
78  channel_id += i;
79  layout >>= i;
80  }
81  }
82  /* reject layouts that are not a single channel */
83  if (channel_id >= MAX_CHANNELS || layout0 != (int64_t)1 << channel_id)
84  return AVERROR(EINVAL);
85  *rchannel = channel_id;
86  *rnamed = 1;
87  *arg += len;
88  return 0;
89  }
90  /* try to parse a channel number, e.g. "c2" */
91  if (sscanf(*arg, "c%d%n", &channel_id, &len) &&
92  channel_id >= 0 && channel_id < MAX_CHANNELS) {
93  *rchannel = channel_id;
94  *rnamed = 0;
95  *arg += len;
96  return 0;
97  }
98  return AVERROR(EINVAL);
99 }
100 
101 static av_cold int init(AVFilterContext *ctx)
102 {
103  PanContext *const pan = ctx->priv;
104  char *arg, *arg0, *tokenizer, *args = av_strdup(pan->args);
105  int out_ch_id, in_ch_id, len, named, ret;
106  int nb_in_channels[2] = { 0, 0 }; // number of unnamed and named input channels
107  double gain;
108 
109  if (!pan->args) {
110  av_log(ctx, AV_LOG_ERROR,
111  "pan filter needs a channel layout and a set "
112  "of channels definitions as parameter\n");
113  return AVERROR(EINVAL);
114  }
115  if (!args)
116  return AVERROR(ENOMEM);
117  arg = av_strtok(args, "|", &tokenizer);
119  &pan->nb_output_channels, arg, ctx);
120  if (ret < 0)
121  goto fail;
122 
123  /* parse channel specifications */
124  while ((arg = arg0 = av_strtok(NULL, "|", &tokenizer))) {
125  /* channel name */
126  if (parse_channel_name(&arg, &out_ch_id, &named)) {
127  av_log(ctx, AV_LOG_ERROR,
128  "Expected out channel name, got \"%.8s\"\n", arg);
129  ret = AVERROR(EINVAL);
130  goto fail;
131  }
132  if (named) {
133  if (!((pan->out_channel_layout >> out_ch_id) & 1)) {
134  av_log(ctx, AV_LOG_ERROR,
135  "Channel \"%.8s\" does not exist in the chosen layout\n", arg0);
136  ret = AVERROR(EINVAL);
137  goto fail;
138  }
139  /* get the channel number in the output channel layout:
140  * out_channel_layout & ((1 << out_ch_id) - 1) are all the
141  * channels that come before out_ch_id,
142  * so their count is the index of out_ch_id */
143  out_ch_id = av_get_channel_layout_nb_channels(pan->out_channel_layout & (((int64_t)1 << out_ch_id) - 1));
144  }
145  if (out_ch_id < 0 || out_ch_id >= pan->nb_output_channels) {
146  av_log(ctx, AV_LOG_ERROR,
147  "Invalid out channel name \"%.8s\"\n", arg0);
148  ret = AVERROR(EINVAL);
149  goto fail;
150  }
151  skip_spaces(&arg);
152  if (*arg == '=') {
153  arg++;
154  } else if (*arg == '<') {
155  pan->need_renorm |= (int64_t)1 << out_ch_id;
156  arg++;
157  } else {
158  av_log(ctx, AV_LOG_ERROR,
159  "Syntax error after channel name in \"%.8s\"\n", arg0);
160  ret = AVERROR(EINVAL);
161  goto fail;
162  }
163  /* gains */
164  while (1) {
165  gain = 1;
166  if (sscanf(arg, "%lf%n *%n", &gain, &len, &len))
167  arg += len;
168  if (parse_channel_name(&arg, &in_ch_id, &named)){
169  av_log(ctx, AV_LOG_ERROR,
170  "Expected in channel name, got \"%.8s\"\n", arg);
171  ret = AVERROR(EINVAL);
172  goto fail;
173  }
174  nb_in_channels[named]++;
175  if (nb_in_channels[!named]) {
176  av_log(ctx, AV_LOG_ERROR,
177  "Can not mix named and numbered channels\n");
178  ret = AVERROR(EINVAL);
179  goto fail;
180  }
181  pan->gain[out_ch_id][in_ch_id] = gain;
182  skip_spaces(&arg);
183  if (!*arg)
184  break;
185  if (*arg != '+') {
186  av_log(ctx, AV_LOG_ERROR, "Syntax error near \"%.8s\"\n", arg);
187  ret = AVERROR(EINVAL);
188  goto fail;
189  }
190  arg++;
191  }
192  }
193  pan->need_renumber = !!nb_in_channels[1];
194 
195  ret = 0;
196 fail:
197  av_free(args);
198  return ret;
199 }
200 
201 static int are_gains_pure(const PanContext *pan)
202 {
203  int i, j;
204 
205  for (i = 0; i < MAX_CHANNELS; i++) {
206  int nb_gain = 0;
207 
208  for (j = 0; j < MAX_CHANNELS; j++) {
209  double gain = pan->gain[i][j];
210 
211  /* channel mapping is effective only if 0% or 100% of a channel is
212  * selected... */
213  if (gain != 0. && gain != 1.)
214  return 0;
215  /* ...and if the output channel is only composed of one input */
216  if (gain && nb_gain++)
217  return 0;
218  }
219  }
220  return 1;
221 }
222 
224 {
225  PanContext *pan = ctx->priv;
226  AVFilterLink *inlink = ctx->inputs[0];
227  AVFilterLink *outlink = ctx->outputs[0];
228  AVFilterFormats *formats = NULL;
230 
231  pan->pure_gains = are_gains_pure(pan);
232  /* libswr supports any sample and packing formats */
234 
235  formats = ff_all_samplerates();
236  if (!formats)
237  return AVERROR(ENOMEM);
238  ff_set_common_samplerates(ctx, formats);
239 
240  // inlink supports any channel layout
241  layouts = ff_all_channel_counts();
242  ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts);
243 
244  // outlink supports only requested output channel layout
245  layouts = NULL;
246  ff_add_channel_layout(&layouts,
249  ff_channel_layouts_ref(layouts, &outlink->in_channel_layouts);
250  return 0;
251 }
252 
253 static int config_props(AVFilterLink *link)
254 {
255  AVFilterContext *ctx = link->dst;
256  PanContext *pan = ctx->priv;
257  char buf[1024], *cur;
258  int i, j, k, r;
259  double t;
260 
261  if (pan->need_renumber) {
262  // input channels were given by their name: renumber them
263  for (i = j = 0; i < MAX_CHANNELS; i++) {
264  if ((link->channel_layout >> i) & 1) {
265  for (k = 0; k < pan->nb_output_channels; k++)
266  pan->gain[k][j] = pan->gain[k][i];
267  j++;
268  }
269  }
270  }
271 
272  // sanity check; can't be done in query_formats since the inlink
273  // channel layout is unknown at that time
274  if (link->channels > SWR_CH_MAX ||
276  av_log(ctx, AV_LOG_ERROR,
277  "libswresample support a maximum of %d channels. "
278  "Feel free to ask for a higher limit.\n", SWR_CH_MAX);
279  return AVERROR_PATCHWELCOME;
280  }
281 
282  // init libswresample context
283  pan->swr = swr_alloc_set_opts(pan->swr,
284  pan->out_channel_layout, link->format, link->sample_rate,
285  link->channel_layout, link->format, link->sample_rate,
286  0, ctx);
287  if (!pan->swr)
288  return AVERROR(ENOMEM);
289  if (!link->channel_layout)
290  av_opt_set_int(pan->swr, "ich", link->channels, 0);
291  if (!pan->out_channel_layout)
292  av_opt_set_int(pan->swr, "och", pan->nb_output_channels, 0);
293 
294  // gains are pure, init the channel mapping
295  if (pan->pure_gains) {
296 
297  // get channel map from the pure gains
298  for (i = 0; i < pan->nb_output_channels; i++) {
299  int ch_id = -1;
300  for (j = 0; j < link->channels; j++) {
301  if (pan->gain[i][j]) {
302  ch_id = j;
303  break;
304  }
305  }
306  pan->channel_map[i] = ch_id;
307  }
308 
309  av_opt_set_int(pan->swr, "icl", pan->out_channel_layout, 0);
310  av_opt_set_int(pan->swr, "uch", pan->nb_output_channels, 0);
312  } else {
313  // renormalize
314  for (i = 0; i < pan->nb_output_channels; i++) {
315  if (!((pan->need_renorm >> i) & 1))
316  continue;
317  t = 0;
318  for (j = 0; j < link->channels; j++)
319  t += pan->gain[i][j];
320  if (t > -1E-5 && t < 1E-5) {
321  // t is almost 0 but not exactly, this is probably a mistake
322  if (t)
323  av_log(ctx, AV_LOG_WARNING,
324  "Degenerate coefficients while renormalizing\n");
325  continue;
326  }
327  for (j = 0; j < link->channels; j++)
328  pan->gain[i][j] /= t;
329  }
330  av_opt_set_int(pan->swr, "icl", link->channel_layout, 0);
331  av_opt_set_int(pan->swr, "ocl", pan->out_channel_layout, 0);
332  swr_set_matrix(pan->swr, pan->gain[0], pan->gain[1] - pan->gain[0]);
333  }
334 
335  r = swr_init(pan->swr);
336  if (r < 0)
337  return r;
338 
339  // summary
340  for (i = 0; i < pan->nb_output_channels; i++) {
341  cur = buf;
342  for (j = 0; j < link->channels; j++) {
343  r = snprintf(cur, buf + sizeof(buf) - cur, "%s%.3g i%d",
344  j ? " + " : "", pan->gain[i][j], j);
345  cur += FFMIN(buf + sizeof(buf) - cur, r);
346  }
347  av_log(ctx, AV_LOG_VERBOSE, "o%d = %s\n", i, buf);
348  }
349  // add channel mapping summary if possible
350  if (pan->pure_gains) {
351  av_log(ctx, AV_LOG_INFO, "Pure channel mapping detected:");
352  for (i = 0; i < pan->nb_output_channels; i++)
353  if (pan->channel_map[i] < 0)
354  av_log(ctx, AV_LOG_INFO, " M");
355  else
356  av_log(ctx, AV_LOG_INFO, " %d", pan->channel_map[i]);
357  av_log(ctx, AV_LOG_INFO, "\n");
358  return 0;
359  }
360  return 0;
361 }
362 
363 static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
364 {
365  int ret;
366  int n = insamples->nb_samples;
367  AVFilterLink *const outlink = inlink->dst->outputs[0];
368  AVFrame *outsamples = ff_get_audio_buffer(outlink, n);
369  PanContext *pan = inlink->dst->priv;
370 
371  if (!outsamples)
372  return AVERROR(ENOMEM);
373  swr_convert(pan->swr, outsamples->extended_data, n,
374  (void *)insamples->extended_data, n);
375  av_frame_copy_props(outsamples, insamples);
376  outsamples->channel_layout = outlink->channel_layout;
377  av_frame_set_channels(outsamples, outlink->channels);
378 
379  ret = ff_filter_frame(outlink, outsamples);
380  av_frame_free(&insamples);
381  return ret;
382 }
383 
384 static av_cold void uninit(AVFilterContext *ctx)
385 {
386  PanContext *pan = ctx->priv;
387  swr_free(&pan->swr);
388 }
389 
390 #define OFFSET(x) offsetof(PanContext, x)
391 
392 static const AVOption pan_options[] = {
393  { "args", NULL, OFFSET(args), AV_OPT_TYPE_STRING, { .str = NULL }, CHAR_MIN, CHAR_MAX, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_FILTERING_PARAM },
394  { NULL }
395 };
396 
398 
399 static const AVFilterPad pan_inputs[] = {
400  {
401  .name = "default",
402  .type = AVMEDIA_TYPE_AUDIO,
403  .config_props = config_props,
404  .filter_frame = filter_frame,
405  },
406  { NULL }
407 };
408 
409 static const AVFilterPad pan_outputs[] = {
410  {
411  .name = "default",
412  .type = AVMEDIA_TYPE_AUDIO,
413  },
414  { NULL }
415 };
416 
418  .name = "pan",
419  .description = NULL_IF_CONFIG_SMALL("Remix channels with coefficients (panning)."),
420  .priv_size = sizeof(PanContext),
421  .priv_class = &pan_class,
422  .init = init,
423  .uninit = uninit,
425  .inputs = pan_inputs,
426  .outputs = pan_outputs,
427 };