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 * Lookahead limiter filter
25 */
26
27 #include "libavutil/channel_layout.h"
28 #include "libavutil/common.h"
29 #include "libavutil/fifo.h"
30 #include "libavutil/opt.h"
31
32 #include "audio.h"
33 #include "avfilter.h"
34 #include "formats.h"
35 #include "internal.h"
36
37 typedef struct MetaItem {
38 int64_t pts;
39 int nb_samples;
40 } MetaItem;
41
42 typedef struct AudioLimiterContext {
43 const AVClass *class;
44
45 double limit;
46 double attack;
47 double release;
48 double att;
49 double level_in;
50 double level_out;
51 int auto_release;
52 int auto_level;
53 double asc;
54 int asc_c;
55 int asc_pos;
56 double asc_coeff;
57
58 double *buffer;
59 int buffer_size;
60 int pos;
61 int *nextpos;
62 double *nextdelta;
63
64 int in_trim;
65 int out_pad;
66 int64_t next_in_pts;
67 int64_t next_out_pts;
68 int latency;
69
70 AVFifo *fifo;
71
72 double delta;
73 int nextiter;
74 int nextlen;
75 int asc_changed;
76 } AudioLimiterContext;
77
78 #define OFFSET(x) offsetof(AudioLimiterContext, x)
79 #define AF AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_FILTERING_PARAM | AV_OPT_FLAG_RUNTIME_PARAM
80
81 static const AVOption alimiter_options[] = {
82 { "level_in", "set input level", OFFSET(level_in), AV_OPT_TYPE_DOUBLE, {.dbl=1},.015625, 64, AF },
83 { "level_out", "set output level", OFFSET(level_out), AV_OPT_TYPE_DOUBLE, {.dbl=1},.015625, 64, AF },
84 { "limit", "set limit", OFFSET(limit), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.0625, 1, AF },
85 { "attack", "set attack", OFFSET(attack), AV_OPT_TYPE_DOUBLE, {.dbl=5}, 0.1, 80, AF },
86 { "release", "set release", OFFSET(release), AV_OPT_TYPE_DOUBLE, {.dbl=50}, 1, 8000, AF },
87 { "asc", "enable asc", OFFSET(auto_release), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, AF },
88 { "asc_level", "set asc level", OFFSET(asc_coeff), AV_OPT_TYPE_DOUBLE, {.dbl=0.5}, 0, 1, AF },
89 { "level", "auto level", OFFSET(auto_level), AV_OPT_TYPE_BOOL, {.i64=1}, 0, 1, AF },
90 { "latency", "compensate delay", OFFSET(latency), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, AF },
91 { NULL }
92 };
93
94 AVFILTER_DEFINE_CLASS(alimiter);
95
init(AVFilterContext * ctx)96 static av_cold int init(AVFilterContext *ctx)
97 {
98 AudioLimiterContext *s = ctx->priv;
99
100 s->attack /= 1000.;
101 s->release /= 1000.;
102 s->att = 1.;
103 s->asc_pos = -1;
104 s->asc_coeff = pow(0.5, s->asc_coeff - 0.5) * 2 * -1;
105
106 return 0;
107 }
108
get_rdelta(AudioLimiterContext * s,double release,int sample_rate,double peak,double limit,double patt,int asc)109 static double get_rdelta(AudioLimiterContext *s, double release, int sample_rate,
110 double peak, double limit, double patt, int asc)
111 {
112 double rdelta = (1.0 - patt) / (sample_rate * release);
113
114 if (asc && s->auto_release && s->asc_c > 0) {
115 double a_att = limit / (s->asc_coeff * s->asc) * (double)s->asc_c;
116
117 if (a_att > patt) {
118 double delta = FFMAX((a_att - patt) / (sample_rate * release), rdelta / 10);
119
120 if (delta < rdelta)
121 rdelta = delta;
122 }
123 }
124
125 return rdelta;
126 }
127
filter_frame(AVFilterLink * inlink,AVFrame * in)128 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
129 {
130 AVFilterContext *ctx = inlink->dst;
131 AudioLimiterContext *s = ctx->priv;
132 AVFilterLink *outlink = ctx->outputs[0];
133 const double *src = (const double *)in->data[0];
134 const int channels = inlink->ch_layout.nb_channels;
135 const int buffer_size = s->buffer_size;
136 double *dst, *buffer = s->buffer;
137 const double release = s->release;
138 const double limit = s->limit;
139 double *nextdelta = s->nextdelta;
140 double level = s->auto_level ? 1 / limit : 1;
141 const double level_out = s->level_out;
142 const double level_in = s->level_in;
143 int *nextpos = s->nextpos;
144 AVFrame *out;
145 double *buf;
146 int n, c, i;
147 int new_out_samples;
148 int64_t out_duration;
149 int64_t in_duration;
150 int64_t in_pts;
151 MetaItem meta;
152
153 if (av_frame_is_writable(in)) {
154 out = in;
155 } else {
156 out = ff_get_audio_buffer(outlink, in->nb_samples);
157 if (!out) {
158 av_frame_free(&in);
159 return AVERROR(ENOMEM);
160 }
161 av_frame_copy_props(out, in);
162 }
163 dst = (double *)out->data[0];
164
165 for (n = 0; n < in->nb_samples; n++) {
166 double peak = 0;
167
168 for (c = 0; c < channels; c++) {
169 double sample = src[c] * level_in;
170
171 buffer[s->pos + c] = sample;
172 peak = FFMAX(peak, fabs(sample));
173 }
174
175 if (s->auto_release && peak > limit) {
176 s->asc += peak;
177 s->asc_c++;
178 }
179
180 if (peak > limit) {
181 double patt = FFMIN(limit / peak, 1.);
182 double rdelta = get_rdelta(s, release, inlink->sample_rate,
183 peak, limit, patt, 0);
184 double delta = (limit / peak - s->att) / buffer_size * channels;
185 int found = 0;
186
187 if (delta < s->delta) {
188 s->delta = delta;
189 nextpos[0] = s->pos;
190 nextpos[1] = -1;
191 nextdelta[0] = rdelta;
192 s->nextlen = 1;
193 s->nextiter= 0;
194 } else {
195 for (i = s->nextiter; i < s->nextiter + s->nextlen; i++) {
196 int j = i % buffer_size;
197 double ppeak, pdelta;
198
199 ppeak = fabs(buffer[nextpos[j]]) > fabs(buffer[nextpos[j] + 1]) ?
200 fabs(buffer[nextpos[j]]) : fabs(buffer[nextpos[j] + 1]);
201 pdelta = (limit / peak - limit / ppeak) / (((buffer_size - nextpos[j] + s->pos) % buffer_size) / channels);
202 if (pdelta < nextdelta[j]) {
203 nextdelta[j] = pdelta;
204 found = 1;
205 break;
206 }
207 }
208 if (found) {
209 s->nextlen = i - s->nextiter + 1;
210 nextpos[(s->nextiter + s->nextlen) % buffer_size] = s->pos;
211 nextdelta[(s->nextiter + s->nextlen) % buffer_size] = rdelta;
212 nextpos[(s->nextiter + s->nextlen + 1) % buffer_size] = -1;
213 s->nextlen++;
214 }
215 }
216 }
217
218 buf = &s->buffer[(s->pos + channels) % buffer_size];
219 peak = 0;
220 for (c = 0; c < channels; c++) {
221 double sample = buf[c];
222
223 peak = FFMAX(peak, fabs(sample));
224 }
225
226 if (s->pos == s->asc_pos && !s->asc_changed)
227 s->asc_pos = -1;
228
229 if (s->auto_release && s->asc_pos == -1 && peak > limit) {
230 s->asc -= peak;
231 s->asc_c--;
232 }
233
234 s->att += s->delta;
235
236 for (c = 0; c < channels; c++)
237 dst[c] = buf[c] * s->att;
238
239 if ((s->pos + channels) % buffer_size == nextpos[s->nextiter]) {
240 if (s->auto_release) {
241 s->delta = get_rdelta(s, release, inlink->sample_rate,
242 peak, limit, s->att, 1);
243 if (s->nextlen > 1) {
244 int pnextpos = nextpos[(s->nextiter + 1) % buffer_size];
245 double ppeak = fabs(buffer[pnextpos]) > fabs(buffer[pnextpos + 1]) ?
246 fabs(buffer[pnextpos]) :
247 fabs(buffer[pnextpos + 1]);
248 double pdelta = (limit / ppeak - s->att) /
249 (((buffer_size + pnextpos -
250 ((s->pos + channels) % buffer_size)) %
251 buffer_size) / channels);
252 if (pdelta < s->delta)
253 s->delta = pdelta;
254 }
255 } else {
256 s->delta = nextdelta[s->nextiter];
257 s->att = limit / peak;
258 }
259
260 s->nextlen -= 1;
261 nextpos[s->nextiter] = -1;
262 s->nextiter = (s->nextiter + 1) % buffer_size;
263 }
264
265 if (s->att > 1.) {
266 s->att = 1.;
267 s->delta = 0.;
268 s->nextiter = 0;
269 s->nextlen = 0;
270 nextpos[0] = -1;
271 }
272
273 if (s->att <= 0.) {
274 s->att = 0.0000000000001;
275 s->delta = (1.0 - s->att) / (inlink->sample_rate * release);
276 }
277
278 if (s->att != 1. && (1. - s->att) < 0.0000000000001)
279 s->att = 1.;
280
281 if (s->delta != 0. && fabs(s->delta) < 0.00000000000001)
282 s->delta = 0.;
283
284 for (c = 0; c < channels; c++)
285 dst[c] = av_clipd(dst[c], -limit, limit) * level * level_out;
286
287 s->pos = (s->pos + channels) % buffer_size;
288 src += channels;
289 dst += channels;
290 }
291
292 in_duration = av_rescale_q(in->nb_samples, inlink->time_base, av_make_q(1, in->sample_rate));
293 in_pts = in->pts;
294 meta = (MetaItem){ in->pts, in->nb_samples };
295 av_fifo_write(s->fifo, &meta, 1);
296 if (in != out)
297 av_frame_free(&in);
298
299 new_out_samples = out->nb_samples;
300 if (s->in_trim > 0) {
301 int trim = FFMIN(new_out_samples, s->in_trim);
302 new_out_samples -= trim;
303 s->in_trim -= trim;
304 }
305
306 if (new_out_samples <= 0) {
307 av_frame_free(&out);
308 return 0;
309 } else if (new_out_samples < out->nb_samples) {
310 int offset = out->nb_samples - new_out_samples;
311 memmove(out->extended_data[0], out->extended_data[0] + sizeof(double) * offset * out->ch_layout.nb_channels,
312 sizeof(double) * new_out_samples * out->ch_layout.nb_channels);
313 out->nb_samples = new_out_samples;
314 s->in_trim = 0;
315 }
316
317 av_fifo_read(s->fifo, &meta, 1);
318
319 out_duration = av_rescale_q(out->nb_samples, inlink->time_base, av_make_q(1, out->sample_rate));
320 in_duration = av_rescale_q(meta.nb_samples, inlink->time_base, av_make_q(1, out->sample_rate));
321 in_pts = meta.pts;
322
323 if (s->next_out_pts != AV_NOPTS_VALUE && out->pts != s->next_out_pts &&
324 s->next_in_pts != AV_NOPTS_VALUE && in_pts == s->next_in_pts) {
325 out->pts = s->next_out_pts;
326 } else {
327 out->pts = in_pts;
328 }
329 s->next_in_pts = in_pts + in_duration;
330 s->next_out_pts = out->pts + out_duration;
331
332 return ff_filter_frame(outlink, out);
333 }
334
request_frame(AVFilterLink * outlink)335 static int request_frame(AVFilterLink* outlink)
336 {
337 AVFilterContext *ctx = outlink->src;
338 AudioLimiterContext *s = (AudioLimiterContext*)ctx->priv;
339 int ret;
340
341 ret = ff_request_frame(ctx->inputs[0]);
342
343 if (ret == AVERROR_EOF && s->out_pad > 0) {
344 AVFrame *frame = ff_get_audio_buffer(outlink, FFMIN(1024, s->out_pad));
345 if (!frame)
346 return AVERROR(ENOMEM);
347
348 s->out_pad -= frame->nb_samples;
349 frame->pts = s->next_in_pts;
350 return filter_frame(ctx->inputs[0], frame);
351 }
352 return ret;
353 }
354
config_input(AVFilterLink * inlink)355 static int config_input(AVFilterLink *inlink)
356 {
357 AVFilterContext *ctx = inlink->dst;
358 AudioLimiterContext *s = ctx->priv;
359 int obuffer_size;
360
361 obuffer_size = inlink->sample_rate * inlink->ch_layout.nb_channels * 100 / 1000. + inlink->ch_layout.nb_channels;
362 if (obuffer_size < inlink->ch_layout.nb_channels)
363 return AVERROR(EINVAL);
364
365 s->buffer = av_calloc(obuffer_size, sizeof(*s->buffer));
366 s->nextdelta = av_calloc(obuffer_size, sizeof(*s->nextdelta));
367 s->nextpos = av_malloc_array(obuffer_size, sizeof(*s->nextpos));
368 if (!s->buffer || !s->nextdelta || !s->nextpos)
369 return AVERROR(ENOMEM);
370
371 memset(s->nextpos, -1, obuffer_size * sizeof(*s->nextpos));
372 s->buffer_size = inlink->sample_rate * s->attack * inlink->ch_layout.nb_channels;
373 s->buffer_size -= s->buffer_size % inlink->ch_layout.nb_channels;
374 if (s->latency)
375 s->in_trim = s->out_pad = s->buffer_size / inlink->ch_layout.nb_channels - 1;
376 s->next_out_pts = AV_NOPTS_VALUE;
377 s->next_in_pts = AV_NOPTS_VALUE;
378
379 s->fifo = av_fifo_alloc2(8, sizeof(MetaItem), AV_FIFO_FLAG_AUTO_GROW);
380 if (!s->fifo) {
381 return AVERROR(ENOMEM);
382 }
383
384 if (s->buffer_size <= 0) {
385 av_log(ctx, AV_LOG_ERROR, "Attack is too small.\n");
386 return AVERROR(EINVAL);
387 }
388
389 return 0;
390 }
391
uninit(AVFilterContext * ctx)392 static av_cold void uninit(AVFilterContext *ctx)
393 {
394 AudioLimiterContext *s = ctx->priv;
395
396 av_freep(&s->buffer);
397 av_freep(&s->nextdelta);
398 av_freep(&s->nextpos);
399
400 av_fifo_freep2(&s->fifo);
401 }
402
403 static const AVFilterPad alimiter_inputs[] = {
404 {
405 .name = "main",
406 .type = AVMEDIA_TYPE_AUDIO,
407 .filter_frame = filter_frame,
408 .config_props = config_input,
409 },
410 };
411
412 static const AVFilterPad alimiter_outputs[] = {
413 {
414 .name = "default",
415 .type = AVMEDIA_TYPE_AUDIO,
416 .request_frame = request_frame,
417 },
418 };
419
420 const AVFilter ff_af_alimiter = {
421 .name = "alimiter",
422 .description = NULL_IF_CONFIG_SMALL("Audio lookahead limiter."),
423 .priv_size = sizeof(AudioLimiterContext),
424 .priv_class = &alimiter_class,
425 .init = init,
426 .uninit = uninit,
427 FILTER_INPUTS(alimiter_inputs),
428 FILTER_OUTPUTS(alimiter_outputs),
429 FILTER_SINGLE_SAMPLEFMT(AV_SAMPLE_FMT_DBL),
430 .process_command = ff_filter_process_command,
431 .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
432 };
433