blob: 3f540e2dff20fa9183da3ac5e6384ce89d119011
1 | /* |
2 | * Copyright (C) 2001-2010 Krzysztof Foltman, Markus Schmidt, Thor Harald Johansen and others |
3 | * Copyright (c) 2015 Paul B Mahol |
4 | * |
5 | * This file is part of FFmpeg. |
6 | * |
7 | * FFmpeg is free software; you can redistribute it and/or |
8 | * modify it under the terms of the GNU Lesser General Public |
9 | * License as published by the Free Software Foundation; either |
10 | * version 2.1 of the License, or (at your option) any later version. |
11 | * |
12 | * FFmpeg is distributed in the hope that it will be useful, |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
15 | * Lesser General Public License for more details. |
16 | * |
17 | * You should have received a copy of the GNU Lesser General Public |
18 | * License along with FFmpeg; if not, write to the Free Software |
19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
20 | */ |
21 | |
22 | /** |
23 | * @file |
24 | * Audio (Sidechain) Compressor filter |
25 | */ |
26 | |
27 | #include "libavutil/audio_fifo.h" |
28 | #include "libavutil/avassert.h" |
29 | #include "libavutil/channel_layout.h" |
30 | #include "libavutil/common.h" |
31 | #include "libavutil/opt.h" |
32 | |
33 | #include "audio.h" |
34 | #include "avfilter.h" |
35 | #include "formats.h" |
36 | #include "hermite.h" |
37 | #include "internal.h" |
38 | |
39 | typedef struct SidechainCompressContext { |
40 | const AVClass *class; |
41 | |
42 | double level_in; |
43 | double level_sc; |
44 | double attack, attack_coeff; |
45 | double release, release_coeff; |
46 | double lin_slope; |
47 | double ratio; |
48 | double threshold; |
49 | double makeup; |
50 | double mix; |
51 | double thres; |
52 | double knee; |
53 | double knee_start; |
54 | double knee_stop; |
55 | double lin_knee_start; |
56 | double adj_knee_start; |
57 | double compressed_knee_stop; |
58 | int link; |
59 | int detection; |
60 | |
61 | AVAudioFifo *fifo[2]; |
62 | int64_t pts; |
63 | } SidechainCompressContext; |
64 | |
65 | #define OFFSET(x) offsetof(SidechainCompressContext, x) |
66 | #define A AV_OPT_FLAG_AUDIO_PARAM |
67 | #define F AV_OPT_FLAG_FILTERING_PARAM |
68 | |
69 | static const AVOption options[] = { |
70 | { "level_in", "set input gain", OFFSET(level_in), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.015625, 64, A|F }, |
71 | { "threshold", "set threshold", OFFSET(threshold), AV_OPT_TYPE_DOUBLE, {.dbl=0.125}, 0.000976563, 1, A|F }, |
72 | { "ratio", "set ratio", OFFSET(ratio), AV_OPT_TYPE_DOUBLE, {.dbl=2}, 1, 20, A|F }, |
73 | { "attack", "set attack", OFFSET(attack), AV_OPT_TYPE_DOUBLE, {.dbl=20}, 0.01, 2000, A|F }, |
74 | { "release", "set release", OFFSET(release), AV_OPT_TYPE_DOUBLE, {.dbl=250}, 0.01, 9000, A|F }, |
75 | { "makeup", "set make up gain", OFFSET(makeup), AV_OPT_TYPE_DOUBLE, {.dbl=2}, 1, 64, A|F }, |
76 | { "knee", "set knee", OFFSET(knee), AV_OPT_TYPE_DOUBLE, {.dbl=2.82843}, 1, 8, A|F }, |
77 | { "link", "set link type", OFFSET(link), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, A|F, "link" }, |
78 | { "average", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A|F, "link" }, |
79 | { "maximum", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A|F, "link" }, |
80 | { "detection", "set detection", OFFSET(detection), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, A|F, "detection" }, |
81 | { "peak", 0, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, A|F, "detection" }, |
82 | { "rms", 0, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, A|F, "detection" }, |
83 | { "level_sc", "set sidechain gain", OFFSET(level_sc), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.015625, 64, A|F }, |
84 | { "mix", "set mix", OFFSET(mix), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0, 1, A|F }, |
85 | { NULL } |
86 | }; |
87 | |
88 | #define sidechaincompress_options options |
89 | AVFILTER_DEFINE_CLASS(sidechaincompress); |
90 | |
91 | // A fake infinity value (because real infinity may break some hosts) |
92 | #define FAKE_INFINITY (65536.0 * 65536.0) |
93 | |
94 | // Check for infinity (with appropriate-ish tolerance) |
95 | #define IS_FAKE_INFINITY(value) (fabs(value-FAKE_INFINITY) < 1.0) |
96 | |
97 | static double output_gain(double lin_slope, double ratio, double thres, |
98 | double knee, double knee_start, double knee_stop, |
99 | double compressed_knee_stop, int detection) |
100 | { |
101 | double slope = log(lin_slope); |
102 | double gain = 0.0; |
103 | double delta = 0.0; |
104 | |
105 | if (detection) |
106 | slope *= 0.5; |
107 | |
108 | if (IS_FAKE_INFINITY(ratio)) { |
109 | gain = thres; |
110 | delta = 0.0; |
111 | } else { |
112 | gain = (slope - thres) / ratio + thres; |
113 | delta = 1.0 / ratio; |
114 | } |
115 | |
116 | if (knee > 1.0 && slope < knee_stop) |
117 | gain = hermite_interpolation(slope, knee_start, knee_stop, |
118 | knee_start, compressed_knee_stop, |
119 | 1.0, delta); |
120 | |
121 | return exp(gain - slope); |
122 | } |
123 | |
124 | static int compressor_config_output(AVFilterLink *outlink) |
125 | { |
126 | AVFilterContext *ctx = outlink->src; |
127 | SidechainCompressContext *s = ctx->priv; |
128 | |
129 | s->thres = log(s->threshold); |
130 | s->lin_knee_start = s->threshold / sqrt(s->knee); |
131 | s->adj_knee_start = s->lin_knee_start * s->lin_knee_start; |
132 | s->knee_start = log(s->lin_knee_start); |
133 | s->knee_stop = log(s->threshold * sqrt(s->knee)); |
134 | s->compressed_knee_stop = (s->knee_stop - s->thres) / s->ratio + s->thres; |
135 | |
136 | s->attack_coeff = FFMIN(1., 1. / (s->attack * outlink->sample_rate / 4000.)); |
137 | s->release_coeff = FFMIN(1., 1. / (s->release * outlink->sample_rate / 4000.)); |
138 | |
139 | return 0; |
140 | } |
141 | |
142 | static void compressor(SidechainCompressContext *s, |
143 | const double *src, double *dst, const double *scsrc, int nb_samples, |
144 | double level_in, double level_sc, |
145 | AVFilterLink *inlink, AVFilterLink *sclink) |
146 | { |
147 | const double makeup = s->makeup; |
148 | const double mix = s->mix; |
149 | int i, c; |
150 | |
151 | for (i = 0; i < nb_samples; i++) { |
152 | double abs_sample, gain = 1.0; |
153 | |
154 | abs_sample = fabs(scsrc[0] * level_sc); |
155 | |
156 | if (s->link == 1) { |
157 | for (c = 1; c < sclink->channels; c++) |
158 | abs_sample = FFMAX(fabs(scsrc[c] * level_sc), abs_sample); |
159 | } else { |
160 | for (c = 1; c < sclink->channels; c++) |
161 | abs_sample += fabs(scsrc[c] * level_sc); |
162 | |
163 | abs_sample /= sclink->channels; |
164 | } |
165 | |
166 | if (s->detection) |
167 | abs_sample *= abs_sample; |
168 | |
169 | s->lin_slope += (abs_sample - s->lin_slope) * (abs_sample > s->lin_slope ? s->attack_coeff : s->release_coeff); |
170 | |
171 | if (s->lin_slope > 0.0 && s->lin_slope > (s->detection ? s->adj_knee_start : s->lin_knee_start)) |
172 | gain = output_gain(s->lin_slope, s->ratio, s->thres, s->knee, |
173 | s->knee_start, s->knee_stop, |
174 | s->compressed_knee_stop, s->detection); |
175 | |
176 | for (c = 0; c < inlink->channels; c++) |
177 | dst[c] = src[c] * level_in * (gain * makeup * mix + (1. - mix)); |
178 | |
179 | src += inlink->channels; |
180 | dst += inlink->channels; |
181 | scsrc += sclink->channels; |
182 | } |
183 | } |
184 | |
185 | #if CONFIG_SIDECHAINCOMPRESS_FILTER |
186 | static int filter_frame(AVFilterLink *link, AVFrame *frame) |
187 | { |
188 | AVFilterContext *ctx = link->dst; |
189 | SidechainCompressContext *s = ctx->priv; |
190 | AVFilterLink *outlink = ctx->outputs[0]; |
191 | AVFrame *out = NULL, *in[2] = { NULL }; |
192 | double *dst; |
193 | int nb_samples; |
194 | int i; |
195 | |
196 | for (i = 0; i < 2; i++) |
197 | if (link == ctx->inputs[i]) |
198 | break; |
199 | av_assert0(i < 2); |
200 | av_audio_fifo_write(s->fifo[i], (void **)frame->extended_data, |
201 | frame->nb_samples); |
202 | av_frame_free(&frame); |
203 | |
204 | nb_samples = FFMIN(av_audio_fifo_size(s->fifo[0]), av_audio_fifo_size(s->fifo[1])); |
205 | if (!nb_samples) |
206 | return 0; |
207 | |
208 | out = ff_get_audio_buffer(outlink, nb_samples); |
209 | if (!out) |
210 | return AVERROR(ENOMEM); |
211 | for (i = 0; i < 2; i++) { |
212 | in[i] = ff_get_audio_buffer(ctx->inputs[i], nb_samples); |
213 | if (!in[i]) { |
214 | av_frame_free(&in[0]); |
215 | av_frame_free(&in[1]); |
216 | av_frame_free(&out); |
217 | return AVERROR(ENOMEM); |
218 | } |
219 | av_audio_fifo_read(s->fifo[i], (void **)in[i]->data, nb_samples); |
220 | } |
221 | |
222 | dst = (double *)out->data[0]; |
223 | out->pts = s->pts; |
224 | s->pts += nb_samples; |
225 | |
226 | compressor(s, (double *)in[0]->data[0], dst, |
227 | (double *)in[1]->data[0], nb_samples, |
228 | s->level_in, s->level_sc, |
229 | ctx->inputs[0], ctx->inputs[1]); |
230 | |
231 | av_frame_free(&in[0]); |
232 | av_frame_free(&in[1]); |
233 | |
234 | return ff_filter_frame(outlink, out); |
235 | } |
236 | |
237 | static int request_frame(AVFilterLink *outlink) |
238 | { |
239 | AVFilterContext *ctx = outlink->src; |
240 | SidechainCompressContext *s = ctx->priv; |
241 | int i; |
242 | |
243 | /* get a frame on each input */ |
244 | for (i = 0; i < 2; i++) { |
245 | AVFilterLink *inlink = ctx->inputs[i]; |
246 | if (!av_audio_fifo_size(s->fifo[i])) |
247 | return ff_request_frame(inlink); |
248 | } |
249 | |
250 | return 0; |
251 | } |
252 | |
253 | static int query_formats(AVFilterContext *ctx) |
254 | { |
255 | AVFilterFormats *formats; |
256 | AVFilterChannelLayouts *layouts = NULL; |
257 | static const enum AVSampleFormat sample_fmts[] = { |
258 | AV_SAMPLE_FMT_DBL, |
259 | AV_SAMPLE_FMT_NONE |
260 | }; |
261 | int ret, i; |
262 | |
263 | if (!ctx->inputs[0]->in_channel_layouts || |
264 | !ctx->inputs[0]->in_channel_layouts->nb_channel_layouts) { |
265 | av_log(ctx, AV_LOG_WARNING, |
266 | "No channel layout for input 1\n"); |
267 | return AVERROR(EAGAIN); |
268 | } |
269 | |
270 | if ((ret = ff_add_channel_layout(&layouts, ctx->inputs[0]->in_channel_layouts->channel_layouts[0])) < 0 || |
271 | (ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts)) < 0) |
272 | return ret; |
273 | |
274 | for (i = 0; i < 2; i++) { |
275 | layouts = ff_all_channel_counts(); |
276 | if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts)) < 0) |
277 | return ret; |
278 | } |
279 | |
280 | formats = ff_make_format_list(sample_fmts); |
281 | if ((ret = ff_set_common_formats(ctx, formats)) < 0) |
282 | return ret; |
283 | |
284 | formats = ff_all_samplerates(); |
285 | return ff_set_common_samplerates(ctx, formats); |
286 | } |
287 | |
288 | static int config_output(AVFilterLink *outlink) |
289 | { |
290 | AVFilterContext *ctx = outlink->src; |
291 | SidechainCompressContext *s = ctx->priv; |
292 | |
293 | if (ctx->inputs[0]->sample_rate != ctx->inputs[1]->sample_rate) { |
294 | av_log(ctx, AV_LOG_ERROR, |
295 | "Inputs must have the same sample rate " |
296 | "%d for in0 vs %d for in1\n", |
297 | ctx->inputs[0]->sample_rate, ctx->inputs[1]->sample_rate); |
298 | return AVERROR(EINVAL); |
299 | } |
300 | |
301 | outlink->sample_rate = ctx->inputs[0]->sample_rate; |
302 | outlink->time_base = ctx->inputs[0]->time_base; |
303 | outlink->channel_layout = ctx->inputs[0]->channel_layout; |
304 | outlink->channels = ctx->inputs[0]->channels; |
305 | |
306 | s->fifo[0] = av_audio_fifo_alloc(ctx->inputs[0]->format, ctx->inputs[0]->channels, 1024); |
307 | s->fifo[1] = av_audio_fifo_alloc(ctx->inputs[1]->format, ctx->inputs[1]->channels, 1024); |
308 | if (!s->fifo[0] || !s->fifo[1]) |
309 | return AVERROR(ENOMEM); |
310 | |
311 | compressor_config_output(outlink); |
312 | |
313 | return 0; |
314 | } |
315 | |
316 | static av_cold void uninit(AVFilterContext *ctx) |
317 | { |
318 | SidechainCompressContext *s = ctx->priv; |
319 | |
320 | av_audio_fifo_free(s->fifo[0]); |
321 | av_audio_fifo_free(s->fifo[1]); |
322 | } |
323 | |
324 | static const AVFilterPad sidechaincompress_inputs[] = { |
325 | { |
326 | .name = "main", |
327 | .type = AVMEDIA_TYPE_AUDIO, |
328 | .filter_frame = filter_frame, |
329 | },{ |
330 | .name = "sidechain", |
331 | .type = AVMEDIA_TYPE_AUDIO, |
332 | .filter_frame = filter_frame, |
333 | }, |
334 | { NULL } |
335 | }; |
336 | |
337 | static const AVFilterPad sidechaincompress_outputs[] = { |
338 | { |
339 | .name = "default", |
340 | .type = AVMEDIA_TYPE_AUDIO, |
341 | .config_props = config_output, |
342 | .request_frame = request_frame, |
343 | }, |
344 | { NULL } |
345 | }; |
346 | |
347 | AVFilter ff_af_sidechaincompress = { |
348 | .name = "sidechaincompress", |
349 | .description = NULL_IF_CONFIG_SMALL("Sidechain compressor."), |
350 | .priv_size = sizeof(SidechainCompressContext), |
351 | .priv_class = &sidechaincompress_class, |
352 | .query_formats = query_formats, |
353 | .uninit = uninit, |
354 | .inputs = sidechaincompress_inputs, |
355 | .outputs = sidechaincompress_outputs, |
356 | }; |
357 | #endif /* CONFIG_SIDECHAINCOMPRESS_FILTER */ |
358 | |
359 | #if CONFIG_ACOMPRESSOR_FILTER |
360 | static int acompressor_filter_frame(AVFilterLink *inlink, AVFrame *in) |
361 | { |
362 | const double *src = (const double *)in->data[0]; |
363 | AVFilterContext *ctx = inlink->dst; |
364 | SidechainCompressContext *s = ctx->priv; |
365 | AVFilterLink *outlink = ctx->outputs[0]; |
366 | AVFrame *out; |
367 | double *dst; |
368 | |
369 | if (av_frame_is_writable(in)) { |
370 | out = in; |
371 | } else { |
372 | out = ff_get_audio_buffer(inlink, in->nb_samples); |
373 | if (!out) { |
374 | av_frame_free(&in); |
375 | return AVERROR(ENOMEM); |
376 | } |
377 | av_frame_copy_props(out, in); |
378 | } |
379 | dst = (double *)out->data[0]; |
380 | |
381 | compressor(s, src, dst, src, in->nb_samples, |
382 | s->level_in, s->level_in, |
383 | inlink, inlink); |
384 | |
385 | if (out != in) |
386 | av_frame_free(&in); |
387 | return ff_filter_frame(outlink, out); |
388 | } |
389 | |
390 | static int acompressor_query_formats(AVFilterContext *ctx) |
391 | { |
392 | AVFilterFormats *formats; |
393 | AVFilterChannelLayouts *layouts; |
394 | static const enum AVSampleFormat sample_fmts[] = { |
395 | AV_SAMPLE_FMT_DBL, |
396 | AV_SAMPLE_FMT_NONE |
397 | }; |
398 | int ret; |
399 | |
400 | layouts = ff_all_channel_counts(); |
401 | if (!layouts) |
402 | return AVERROR(ENOMEM); |
403 | ret = ff_set_common_channel_layouts(ctx, layouts); |
404 | if (ret < 0) |
405 | return ret; |
406 | |
407 | formats = ff_make_format_list(sample_fmts); |
408 | if (!formats) |
409 | return AVERROR(ENOMEM); |
410 | ret = ff_set_common_formats(ctx, formats); |
411 | if (ret < 0) |
412 | return ret; |
413 | |
414 | formats = ff_all_samplerates(); |
415 | if (!formats) |
416 | return AVERROR(ENOMEM); |
417 | return ff_set_common_samplerates(ctx, formats); |
418 | } |
419 | |
420 | #define acompressor_options options |
421 | AVFILTER_DEFINE_CLASS(acompressor); |
422 | |
423 | static const AVFilterPad acompressor_inputs[] = { |
424 | { |
425 | .name = "default", |
426 | .type = AVMEDIA_TYPE_AUDIO, |
427 | .filter_frame = acompressor_filter_frame, |
428 | }, |
429 | { NULL } |
430 | }; |
431 | |
432 | static const AVFilterPad acompressor_outputs[] = { |
433 | { |
434 | .name = "default", |
435 | .type = AVMEDIA_TYPE_AUDIO, |
436 | .config_props = compressor_config_output, |
437 | }, |
438 | { NULL } |
439 | }; |
440 | |
441 | AVFilter ff_af_acompressor = { |
442 | .name = "acompressor", |
443 | .description = NULL_IF_CONFIG_SMALL("Audio compressor."), |
444 | .priv_size = sizeof(SidechainCompressContext), |
445 | .priv_class = &acompressor_class, |
446 | .query_formats = acompressor_query_formats, |
447 | .inputs = acompressor_inputs, |
448 | .outputs = acompressor_outputs, |
449 | }; |
450 | #endif /* CONFIG_ACOMPRESSOR_FILTER */ |
451 |