blob: a477bde460b001968d2a592b7c2aba42108f21ce
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" |
32 | #include "libavutil/channel_layout.h" |
33 | #include "libavutil/opt.h" |
34 | #include "libswresample/swresample.h" |
35 | #include "audio.h" |
36 | #include "avfilter.h" |
37 | #include "formats.h" |
38 | #include "internal.h" |
39 | |
40 | #define MAX_CHANNELS 64 |
41 | |
42 | typedef struct PanContext { |
43 | const AVClass *class; |
44 | char *args; |
45 | int64_t out_channel_layout; |
46 | double gain[MAX_CHANNELS][MAX_CHANNELS]; |
47 | int64_t need_renorm; |
48 | int need_renumber; |
49 | int nb_output_channels; |
50 | |
51 | int pure_gains; |
52 | /* channel mapping specific */ |
53 | int channel_map[MAX_CHANNELS]; |
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, sign = 1; |
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 channel definitions as parameter\n"); |
113 | return AVERROR(EINVAL); |
114 | } |
115 | if (!args) |
116 | return AVERROR(ENOMEM); |
117 | arg = av_strtok(args, "|", &tokenizer); |
118 | if (!arg) { |
119 | av_log(ctx, AV_LOG_ERROR, "Channel layout not specified\n"); |
120 | ret = AVERROR(EINVAL); |
121 | goto fail; |
122 | } |
123 | ret = ff_parse_channel_layout(&pan->out_channel_layout, |
124 | &pan->nb_output_channels, arg, ctx); |
125 | if (ret < 0) |
126 | goto fail; |
127 | |
128 | /* parse channel specifications */ |
129 | while ((arg = arg0 = av_strtok(NULL, "|", &tokenizer))) { |
130 | /* channel name */ |
131 | if (parse_channel_name(&arg, &out_ch_id, &named)) { |
132 | av_log(ctx, AV_LOG_ERROR, |
133 | "Expected out channel name, got \"%.8s\"\n", arg); |
134 | ret = AVERROR(EINVAL); |
135 | goto fail; |
136 | } |
137 | if (named) { |
138 | if (!((pan->out_channel_layout >> out_ch_id) & 1)) { |
139 | av_log(ctx, AV_LOG_ERROR, |
140 | "Channel \"%.8s\" does not exist in the chosen layout\n", arg0); |
141 | ret = AVERROR(EINVAL); |
142 | goto fail; |
143 | } |
144 | /* get the channel number in the output channel layout: |
145 | * out_channel_layout & ((1 << out_ch_id) - 1) are all the |
146 | * channels that come before out_ch_id, |
147 | * so their count is the index of out_ch_id */ |
148 | out_ch_id = av_get_channel_layout_nb_channels(pan->out_channel_layout & (((int64_t)1 << out_ch_id) - 1)); |
149 | } |
150 | if (out_ch_id < 0 || out_ch_id >= pan->nb_output_channels) { |
151 | av_log(ctx, AV_LOG_ERROR, |
152 | "Invalid out channel name \"%.8s\"\n", arg0); |
153 | ret = AVERROR(EINVAL); |
154 | goto fail; |
155 | } |
156 | skip_spaces(&arg); |
157 | if (*arg == '=') { |
158 | arg++; |
159 | } else if (*arg == '<') { |
160 | pan->need_renorm |= (int64_t)1 << out_ch_id; |
161 | arg++; |
162 | } else { |
163 | av_log(ctx, AV_LOG_ERROR, |
164 | "Syntax error after channel name in \"%.8s\"\n", arg0); |
165 | ret = AVERROR(EINVAL); |
166 | goto fail; |
167 | } |
168 | /* gains */ |
169 | while (1) { |
170 | gain = 1; |
171 | if (sscanf(arg, "%lf%n *%n", &gain, &len, &len)) |
172 | arg += len; |
173 | if (parse_channel_name(&arg, &in_ch_id, &named)){ |
174 | av_log(ctx, AV_LOG_ERROR, |
175 | "Expected in channel name, got \"%.8s\"\n", arg); |
176 | ret = AVERROR(EINVAL); |
177 | goto fail; |
178 | } |
179 | nb_in_channels[named]++; |
180 | if (nb_in_channels[!named]) { |
181 | av_log(ctx, AV_LOG_ERROR, |
182 | "Can not mix named and numbered channels\n"); |
183 | ret = AVERROR(EINVAL); |
184 | goto fail; |
185 | } |
186 | pan->gain[out_ch_id][in_ch_id] = sign * gain; |
187 | skip_spaces(&arg); |
188 | if (!*arg) |
189 | break; |
190 | if (*arg == '-') { |
191 | sign = -1; |
192 | } else if (*arg != '+') { |
193 | av_log(ctx, AV_LOG_ERROR, "Syntax error near \"%.8s\"\n", arg); |
194 | ret = AVERROR(EINVAL); |
195 | goto fail; |
196 | } else { |
197 | sign = 1; |
198 | } |
199 | arg++; |
200 | } |
201 | } |
202 | pan->need_renumber = !!nb_in_channels[1]; |
203 | |
204 | ret = 0; |
205 | fail: |
206 | av_free(args); |
207 | return ret; |
208 | } |
209 | |
210 | static int are_gains_pure(const PanContext *pan) |
211 | { |
212 | int i, j; |
213 | |
214 | for (i = 0; i < MAX_CHANNELS; i++) { |
215 | int nb_gain = 0; |
216 | |
217 | for (j = 0; j < MAX_CHANNELS; j++) { |
218 | double gain = pan->gain[i][j]; |
219 | |
220 | /* channel mapping is effective only if 0% or 100% of a channel is |
221 | * selected... */ |
222 | if (gain != 0. && gain != 1.) |
223 | return 0; |
224 | /* ...and if the output channel is only composed of one input */ |
225 | if (gain && nb_gain++) |
226 | return 0; |
227 | } |
228 | } |
229 | return 1; |
230 | } |
231 | |
232 | static int query_formats(AVFilterContext *ctx) |
233 | { |
234 | PanContext *pan = ctx->priv; |
235 | AVFilterLink *inlink = ctx->inputs[0]; |
236 | AVFilterLink *outlink = ctx->outputs[0]; |
237 | AVFilterFormats *formats = NULL; |
238 | AVFilterChannelLayouts *layouts; |
239 | int ret; |
240 | |
241 | pan->pure_gains = are_gains_pure(pan); |
242 | /* libswr supports any sample and packing formats */ |
243 | if ((ret = ff_set_common_formats(ctx, ff_all_formats(AVMEDIA_TYPE_AUDIO))) < 0) |
244 | return ret; |
245 | |
246 | formats = ff_all_samplerates(); |
247 | if ((ret = ff_set_common_samplerates(ctx, formats)) < 0) |
248 | return ret; |
249 | |
250 | // inlink supports any channel layout |
251 | layouts = ff_all_channel_counts(); |
252 | if ((ret = ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts)) < 0) |
253 | return ret; |
254 | |
255 | // outlink supports only requested output channel layout |
256 | layouts = NULL; |
257 | if ((ret = ff_add_channel_layout(&layouts, |
258 | pan->out_channel_layout ? pan->out_channel_layout : |
259 | FF_COUNT2LAYOUT(pan->nb_output_channels))) < 0) |
260 | return ret; |
261 | return ff_channel_layouts_ref(layouts, &outlink->in_channel_layouts); |
262 | } |
263 | |
264 | static int config_props(AVFilterLink *link) |
265 | { |
266 | AVFilterContext *ctx = link->dst; |
267 | PanContext *pan = ctx->priv; |
268 | char buf[1024], *cur; |
269 | int i, j, k, r; |
270 | double t; |
271 | |
272 | if (pan->need_renumber) { |
273 | // input channels were given by their name: renumber them |
274 | for (i = j = 0; i < MAX_CHANNELS; i++) { |
275 | if ((link->channel_layout >> i) & 1) { |
276 | for (k = 0; k < pan->nb_output_channels; k++) |
277 | pan->gain[k][j] = pan->gain[k][i]; |
278 | j++; |
279 | } |
280 | } |
281 | } |
282 | |
283 | // sanity check; can't be done in query_formats since the inlink |
284 | // channel layout is unknown at that time |
285 | if (link->channels > MAX_CHANNELS || |
286 | pan->nb_output_channels > MAX_CHANNELS) { |
287 | av_log(ctx, AV_LOG_ERROR, |
288 | "af_pan supports a maximum of %d channels. " |
289 | "Feel free to ask for a higher limit.\n", MAX_CHANNELS); |
290 | return AVERROR_PATCHWELCOME; |
291 | } |
292 | |
293 | // init libswresample context |
294 | pan->swr = swr_alloc_set_opts(pan->swr, |
295 | pan->out_channel_layout, link->format, link->sample_rate, |
296 | link->channel_layout, link->format, link->sample_rate, |
297 | 0, ctx); |
298 | if (!pan->swr) |
299 | return AVERROR(ENOMEM); |
300 | if (!link->channel_layout) { |
301 | if (av_opt_set_int(pan->swr, "ich", link->channels, 0) < 0) |
302 | return AVERROR(EINVAL); |
303 | } |
304 | if (!pan->out_channel_layout) { |
305 | if (av_opt_set_int(pan->swr, "och", pan->nb_output_channels, 0) < 0) |
306 | return AVERROR(EINVAL); |
307 | } |
308 | |
309 | // gains are pure, init the channel mapping |
310 | if (pan->pure_gains) { |
311 | |
312 | // get channel map from the pure gains |
313 | for (i = 0; i < pan->nb_output_channels; i++) { |
314 | int ch_id = -1; |
315 | for (j = 0; j < link->channels; j++) { |
316 | if (pan->gain[i][j]) { |
317 | ch_id = j; |
318 | break; |
319 | } |
320 | } |
321 | pan->channel_map[i] = ch_id; |
322 | } |
323 | |
324 | av_opt_set_int(pan->swr, "icl", pan->out_channel_layout, 0); |
325 | av_opt_set_int(pan->swr, "uch", pan->nb_output_channels, 0); |
326 | swr_set_channel_mapping(pan->swr, pan->channel_map); |
327 | } else { |
328 | // renormalize |
329 | for (i = 0; i < pan->nb_output_channels; i++) { |
330 | if (!((pan->need_renorm >> i) & 1)) |
331 | continue; |
332 | t = 0; |
333 | for (j = 0; j < link->channels; j++) |
334 | t += fabs(pan->gain[i][j]); |
335 | if (t > -1E-5 && t < 1E-5) { |
336 | // t is almost 0 but not exactly, this is probably a mistake |
337 | if (t) |
338 | av_log(ctx, AV_LOG_WARNING, |
339 | "Degenerate coefficients while renormalizing\n"); |
340 | continue; |
341 | } |
342 | for (j = 0; j < link->channels; j++) |
343 | pan->gain[i][j] /= t; |
344 | } |
345 | av_opt_set_int(pan->swr, "icl", link->channel_layout, 0); |
346 | av_opt_set_int(pan->swr, "ocl", pan->out_channel_layout, 0); |
347 | swr_set_matrix(pan->swr, pan->gain[0], pan->gain[1] - pan->gain[0]); |
348 | } |
349 | |
350 | r = swr_init(pan->swr); |
351 | if (r < 0) |
352 | return r; |
353 | |
354 | // summary |
355 | for (i = 0; i < pan->nb_output_channels; i++) { |
356 | cur = buf; |
357 | for (j = 0; j < link->channels; j++) { |
358 | r = snprintf(cur, buf + sizeof(buf) - cur, "%s%.3g i%d", |
359 | j ? " + " : "", pan->gain[i][j], j); |
360 | cur += FFMIN(buf + sizeof(buf) - cur, r); |
361 | } |
362 | av_log(ctx, AV_LOG_VERBOSE, "o%d = %s\n", i, buf); |
363 | } |
364 | // add channel mapping summary if possible |
365 | if (pan->pure_gains) { |
366 | av_log(ctx, AV_LOG_INFO, "Pure channel mapping detected:"); |
367 | for (i = 0; i < pan->nb_output_channels; i++) |
368 | if (pan->channel_map[i] < 0) |
369 | av_log(ctx, AV_LOG_INFO, " M"); |
370 | else |
371 | av_log(ctx, AV_LOG_INFO, " %d", pan->channel_map[i]); |
372 | av_log(ctx, AV_LOG_INFO, "\n"); |
373 | return 0; |
374 | } |
375 | return 0; |
376 | } |
377 | |
378 | static int filter_frame(AVFilterLink *inlink, AVFrame *insamples) |
379 | { |
380 | int ret; |
381 | int n = insamples->nb_samples; |
382 | AVFilterLink *const outlink = inlink->dst->outputs[0]; |
383 | AVFrame *outsamples = ff_get_audio_buffer(outlink, n); |
384 | PanContext *pan = inlink->dst->priv; |
385 | |
386 | if (!outsamples) |
387 | return AVERROR(ENOMEM); |
388 | swr_convert(pan->swr, outsamples->extended_data, n, |
389 | (void *)insamples->extended_data, n); |
390 | av_frame_copy_props(outsamples, insamples); |
391 | outsamples->channel_layout = outlink->channel_layout; |
392 | av_frame_set_channels(outsamples, outlink->channels); |
393 | |
394 | ret = ff_filter_frame(outlink, outsamples); |
395 | av_frame_free(&insamples); |
396 | return ret; |
397 | } |
398 | |
399 | static av_cold void uninit(AVFilterContext *ctx) |
400 | { |
401 | PanContext *pan = ctx->priv; |
402 | swr_free(&pan->swr); |
403 | } |
404 | |
405 | #define OFFSET(x) offsetof(PanContext, x) |
406 | |
407 | static const AVOption pan_options[] = { |
408 | { "args", NULL, OFFSET(args), AV_OPT_TYPE_STRING, { .str = NULL }, CHAR_MIN, CHAR_MAX, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_FILTERING_PARAM }, |
409 | { NULL } |
410 | }; |
411 | |
412 | AVFILTER_DEFINE_CLASS(pan); |
413 | |
414 | static const AVFilterPad pan_inputs[] = { |
415 | { |
416 | .name = "default", |
417 | .type = AVMEDIA_TYPE_AUDIO, |
418 | .config_props = config_props, |
419 | .filter_frame = filter_frame, |
420 | }, |
421 | { NULL } |
422 | }; |
423 | |
424 | static const AVFilterPad pan_outputs[] = { |
425 | { |
426 | .name = "default", |
427 | .type = AVMEDIA_TYPE_AUDIO, |
428 | }, |
429 | { NULL } |
430 | }; |
431 | |
432 | AVFilter ff_af_pan = { |
433 | .name = "pan", |
434 | .description = NULL_IF_CONFIG_SMALL("Remix channels with coefficients (panning)."), |
435 | .priv_size = sizeof(PanContext), |
436 | .priv_class = &pan_class, |
437 | .init = init, |
438 | .uninit = uninit, |
439 | .query_formats = query_formats, |
440 | .inputs = pan_inputs, |
441 | .outputs = pan_outputs, |
442 | }; |
443 |