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 <ubitux@gmail.com>
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;
51 
53  /* channel mapping specific */
55  struct SwrContext *swr;
56 } PanContext;
57 
58 static void skip_spaces(char **arg)
59 {
60  int len = 0;
61 
62  sscanf(*arg, " %n", &len);
63  *arg += len;
64 }
65 
66 static int parse_channel_name(char **arg, int *rchannel, int *rnamed)
67 {
68  char buf[8];
69  int len, i, channel_id = 0;
70  int64_t layout, layout0;
71 
72  skip_spaces(arg);
73  /* try to parse a channel name, e.g. "FL" */
74  if (sscanf(*arg, "%7[A-Z]%n", buf, &len)) {
75  layout0 = layout = av_get_channel_layout(buf);
76  /* channel_id <- first set bit in layout */
77  for (i = 32; i > 0; i >>= 1) {
78  if (layout >= (int64_t)1 << i) {
79  channel_id += i;
80  layout >>= i;
81  }
82  }
83  /* reject layouts that are not a single channel */
84  if (channel_id >= MAX_CHANNELS || layout0 != (int64_t)1 << channel_id)
85  return AVERROR(EINVAL);
86  *rchannel = channel_id;
87  *rnamed = 1;
88  *arg += len;
89  return 0;
90  }
91  /* try to parse a channel number, e.g. "c2" */
92  if (sscanf(*arg, "c%d%n", &channel_id, &len) &&
93  channel_id >= 0 && channel_id < MAX_CHANNELS) {
94  *rchannel = channel_id;
95  *rnamed = 0;
96  *arg += len;
97  return 0;
98  }
99  return AVERROR(EINVAL);
100 }
101 
102 static av_cold int init(AVFilterContext *ctx)
103 {
104  PanContext *const pan = ctx->priv;
105  char *arg, *arg0, *tokenizer, *args = av_strdup(pan->args);
106  int out_ch_id, in_ch_id, len, named, ret;
107  int nb_in_channels[2] = { 0, 0 }; // number of unnamed and named input channels
108  double gain;
109 
110  if (!pan->args) {
111  av_log(ctx, AV_LOG_ERROR,
112  "pan filter needs a channel layout and a set "
113  "of channels definitions as parameter\n");
114  return AVERROR(EINVAL);
115  }
116  if (!args)
117  return AVERROR(ENOMEM);
118  arg = av_strtok(args, "|", &tokenizer);
119  ret = ff_parse_channel_layout(&pan->out_channel_layout, arg, ctx);
120  if (ret < 0)
121  goto fail;
123 
124  /* parse channel specifications */
125  while ((arg = arg0 = av_strtok(NULL, "|", &tokenizer))) {
126  /* channel name */
127  if (parse_channel_name(&arg, &out_ch_id, &named)) {
128  av_log(ctx, AV_LOG_ERROR,
129  "Expected out channel name, got \"%.8s\"\n", arg);
130  ret = AVERROR(EINVAL);
131  goto fail;
132  }
133  if (named) {
134  if (!((pan->out_channel_layout >> out_ch_id) & 1)) {
135  av_log(ctx, AV_LOG_ERROR,
136  "Channel \"%.8s\" does not exist in the chosen layout\n", arg0);
137  ret = AVERROR(EINVAL);
138  goto fail;
139  }
140  /* get the channel number in the output channel layout:
141  * out_channel_layout & ((1 << out_ch_id) - 1) are all the
142  * channels that come before out_ch_id,
143  * so their count is the index of out_ch_id */
144  out_ch_id = av_get_channel_layout_nb_channels(pan->out_channel_layout & (((int64_t)1 << out_ch_id) - 1));
145  }
146  if (out_ch_id < 0 || out_ch_id >= pan->nb_output_channels) {
147  av_log(ctx, AV_LOG_ERROR,
148  "Invalid out channel name \"%.8s\"\n", arg0);
149  ret = AVERROR(EINVAL);
150  goto fail;
151  }
152  skip_spaces(&arg);
153  if (*arg == '=') {
154  arg++;
155  } else if (*arg == '<') {
156  pan->need_renorm |= (int64_t)1 << out_ch_id;
157  arg++;
158  } else {
159  av_log(ctx, AV_LOG_ERROR,
160  "Syntax error after channel name in \"%.8s\"\n", arg0);
161  ret = AVERROR(EINVAL);
162  goto fail;
163  }
164  /* gains */
165  while (1) {
166  gain = 1;
167  if (sscanf(arg, "%lf%n *%n", &gain, &len, &len))
168  arg += len;
169  if (parse_channel_name(&arg, &in_ch_id, &named)){
170  av_log(ctx, AV_LOG_ERROR,
171  "Expected in channel name, got \"%.8s\"\n", arg);
172  ret = AVERROR(EINVAL);
173  goto fail;
174  }
175  nb_in_channels[named]++;
176  if (nb_in_channels[!named]) {
177  av_log(ctx, AV_LOG_ERROR,
178  "Can not mix named and numbered channels\n");
179  ret = AVERROR(EINVAL);
180  goto fail;
181  }
182  pan->gain[out_ch_id][in_ch_id] = gain;
183  skip_spaces(&arg);
184  if (!*arg)
185  break;
186  if (*arg != '+') {
187  av_log(ctx, AV_LOG_ERROR, "Syntax error near \"%.8s\"\n", arg);
188  ret = AVERROR(EINVAL);
189  goto fail;
190  }
191  arg++;
192  }
193  }
194  pan->need_renumber = !!nb_in_channels[1];
195 
196  ret = 0;
197 fail:
198  av_free(args);
199  return ret;
200 }
201 
202 static int are_gains_pure(const PanContext *pan)
203 {
204  int i, j;
205 
206  for (i = 0; i < MAX_CHANNELS; i++) {
207  int nb_gain = 0;
208 
209  for (j = 0; j < MAX_CHANNELS; j++) {
210  double gain = pan->gain[i][j];
211 
212  /* channel mapping is effective only if 0% or 100% of a channel is
213  * selected... */
214  if (gain != 0. && gain != 1.)
215  return 0;
216  /* ...and if the output channel is only composed of one input */
217  if (gain && nb_gain++)
218  return 0;
219  }
220  }
221  return 1;
222 }
223 
225 {
226  PanContext *pan = ctx->priv;
227  AVFilterLink *inlink = ctx->inputs[0];
228  AVFilterLink *outlink = ctx->outputs[0];
229  AVFilterFormats *formats = NULL;
231 
232  pan->pure_gains = are_gains_pure(pan);
233  /* libswr supports any sample and packing formats */
235 
236  formats = ff_all_samplerates();
237  if (!formats)
238  return AVERROR(ENOMEM);
239  ff_set_common_samplerates(ctx, formats);
240 
241  // inlink supports any channel layout
242  layouts = ff_all_channel_layouts();
243  ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts);
244 
245  // outlink supports only requested output channel layout
246  layouts = NULL;
248  ff_channel_layouts_ref(layouts, &outlink->in_channel_layouts);
249  return 0;
250 }
251 
252 static int config_props(AVFilterLink *link)
253 {
254  AVFilterContext *ctx = link->dst;
255  PanContext *pan = ctx->priv;
256  char buf[1024], *cur;
257  int i, j, k, r;
258  double t;
259 
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 (pan->nb_input_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 
290  // gains are pure, init the channel mapping
291  if (pan->pure_gains) {
292 
293  // get channel map from the pure gains
294  for (i = 0; i < pan->nb_output_channels; i++) {
295  int ch_id = -1;
296  for (j = 0; j < pan->nb_input_channels; j++) {
297  if (pan->gain[i][j]) {
298  ch_id = j;
299  break;
300  }
301  }
302  pan->channel_map[i] = ch_id;
303  }
304 
305  av_opt_set_int(pan->swr, "icl", pan->out_channel_layout, 0);
306  av_opt_set_int(pan->swr, "uch", pan->nb_output_channels, 0);
308  } else {
309  // renormalize
310  for (i = 0; i < pan->nb_output_channels; i++) {
311  if (!((pan->need_renorm >> i) & 1))
312  continue;
313  t = 0;
314  for (j = 0; j < pan->nb_input_channels; j++)
315  t += pan->gain[i][j];
316  if (t > -1E-5 && t < 1E-5) {
317  // t is almost 0 but not exactly, this is probably a mistake
318  if (t)
319  av_log(ctx, AV_LOG_WARNING,
320  "Degenerate coefficients while renormalizing\n");
321  continue;
322  }
323  for (j = 0; j < pan->nb_input_channels; j++)
324  pan->gain[i][j] /= t;
325  }
326  av_opt_set_int(pan->swr, "icl", link->channel_layout, 0);
327  av_opt_set_int(pan->swr, "ocl", pan->out_channel_layout, 0);
328  swr_set_matrix(pan->swr, pan->gain[0], pan->gain[1] - pan->gain[0]);
329  }
330 
331  r = swr_init(pan->swr);
332  if (r < 0)
333  return r;
334 
335  // summary
336  for (i = 0; i < pan->nb_output_channels; i++) {
337  cur = buf;
338  for (j = 0; j < pan->nb_input_channels; j++) {
339  r = snprintf(cur, buf + sizeof(buf) - cur, "%s%.3g i%d",
340  j ? " + " : "", pan->gain[i][j], j);
341  cur += FFMIN(buf + sizeof(buf) - cur, r);
342  }
343  av_log(ctx, AV_LOG_VERBOSE, "o%d = %s\n", i, buf);
344  }
345  // add channel mapping summary if possible
346  if (pan->pure_gains) {
347  av_log(ctx, AV_LOG_INFO, "Pure channel mapping detected:");
348  for (i = 0; i < pan->nb_output_channels; i++)
349  if (pan->channel_map[i] < 0)
350  av_log(ctx, AV_LOG_INFO, " M");
351  else
352  av_log(ctx, AV_LOG_INFO, " %d", pan->channel_map[i]);
353  av_log(ctx, AV_LOG_INFO, "\n");
354  return 0;
355  }
356  return 0;
357 }
358 
359 static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
360 {
361  int ret;
362  int n = insamples->nb_samples;
363  AVFilterLink *const outlink = inlink->dst->outputs[0];
364  AVFrame *outsamples = ff_get_audio_buffer(outlink, n);
365  PanContext *pan = inlink->dst->priv;
366 
367  if (!outsamples)
368  return AVERROR(ENOMEM);
369  swr_convert(pan->swr, outsamples->data, n, (void *)insamples->data, n);
370  av_frame_copy_props(outsamples, insamples);
371  outsamples->channel_layout = outlink->channel_layout;
372  av_frame_set_channels(outsamples, outlink->channels);
373 
374  ret = ff_filter_frame(outlink, outsamples);
375  av_frame_free(&insamples);
376  return ret;
377 }
378 
379 static av_cold void uninit(AVFilterContext *ctx)
380 {
381  PanContext *pan = ctx->priv;
382  swr_free(&pan->swr);
383 }
384 
385 #define OFFSET(x) offsetof(PanContext, x)
386 
387 static const AVOption pan_options[] = {
388  { "args", NULL, OFFSET(args), AV_OPT_TYPE_STRING, { .str = NULL }, CHAR_MIN, CHAR_MAX, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_FILTERING_PARAM },
389  { NULL }
390 };
391 
393 
394 
395 static const AVFilterPad pan_inputs[] = {
396  {
397  .name = "default",
398  .type = AVMEDIA_TYPE_AUDIO,
399  .config_props = config_props,
400  .filter_frame = filter_frame,
401  },
402  { NULL }
403 };
404 
405 static const AVFilterPad pan_outputs[] = {
406  {
407  .name = "default",
408  .type = AVMEDIA_TYPE_AUDIO,
409  },
410  { NULL }
411 };
412 
414  .name = "pan",
415  .description = NULL_IF_CONFIG_SMALL("Remix channels with coefficients (panning)."),
416  .priv_size = sizeof(PanContext),
417  .priv_class = &pan_class,
418  .init = init,
419  .uninit = uninit,
421  .inputs = pan_inputs,
422  .outputs = pan_outputs,
423 };