Browse code

avfilter: add acrusher filter

Paul B Mahol authored on 2016/08/10 23:11:37
Showing 6 changed files
... ...
@@ -14,6 +14,7 @@ version <next>:
14 14
 - MediaCodec hwaccel
15 15
 - True Audio (TTA) muxer
16 16
 - crystalizer audio filter
17
+- acrusher audio filter
17 18
 
18 19
 
19 20
 version 3.1:
... ...
@@ -441,6 +441,64 @@ ffmpeg -i first.flac -i second.flac -filter_complex acrossfade=d=10:o=0:c1=exp:c
441 441
 @end example
442 442
 @end itemize
443 443
 
444
+@section acrusher
445
+
446
+Reduce audio bit resolution.
447
+
448
+This filter is bit crusher with enhanced funcionality. A bit crusher
449
+is used to audibly reduce number of bits an audio signal is sampled
450
+with. This doesn't change the bit depth at all, it just produces the
451
+effect. Material reduced in bit depth sounds more harsh and "digital".
452
+This filter is able to even round to continous values instead of discrete
453
+bit depths.
454
+Additionally it has a D/C offset which results in different crushing of
455
+the lower and the upper half of the signal.
456
+An Anti-Aliasing setting is able to produce "softer" crushing sounds.
457
+
458
+Another feature of this filter is the logarithmic mode.
459
+This setting switches from linear distances between bits to logarithmic ones.
460
+The result is a much more "natural" sounding crusher which doesn't gate low
461
+signals for example. The human ear has a logarithmic perception, too
462
+so this kind of crushing is much more pleasant.
463
+Logarithmic crushing is also able to get anti-aliased.
464
+
465
+The filter accepts the following options:
466
+
467
+@table @option
468
+@item level_in
469
+Set level in.
470
+
471
+@item level_out
472
+Set level out.
473
+
474
+@item bits
475
+Set bit reduction.
476
+
477
+@item mix
478
+Set mixing ammount.
479
+
480
+@item mode
481
+Can be linear: @code{lin} or logarithmic: @code{log}.
482
+
483
+@item dc
484
+Set DC.
485
+
486
+@item aa
487
+Set anti-aliasing.
488
+
489
+@item samples
490
+Set sample reduction.
491
+
492
+@item lfo
493
+Enable LFO. By default disabled.
494
+
495
+@item lforange
496
+Set LFO range.
497
+
498
+@item lforate
499
+Set LFO rate.
500
+@end table
501
+
444 502
 @section adelay
445 503
 
446 504
 Delay one or more audio channels.
... ...
@@ -30,6 +30,7 @@ OBJS-$(HAVE_THREADS)                         += pthread.o
30 30
 OBJS-$(CONFIG_ABENCH_FILTER)                 += f_bench.o
31 31
 OBJS-$(CONFIG_ACOMPRESSOR_FILTER)            += af_sidechaincompress.o
32 32
 OBJS-$(CONFIG_ACROSSFADE_FILTER)             += af_afade.o
33
+OBJS-$(CONFIG_ACRUSHER_FILTER)               += af_acrusher.o
33 34
 OBJS-$(CONFIG_ADELAY_FILTER)                 += af_adelay.o
34 35
 OBJS-$(CONFIG_AECHO_FILTER)                  += af_aecho.o
35 36
 OBJS-$(CONFIG_AEMPHASIS_FILTER)              += af_aemphasis.o
36 37
new file mode 100644
... ...
@@ -0,0 +1,362 @@
0
+/*
1
+ * Copyright (c) Markus Schmidt and Christian Holschuh
2
+ *
3
+ * This file is part of FFmpeg.
4
+ *
5
+ * FFmpeg is free software; you can redistribute it and/or
6
+ * modify it under the terms of the GNU Lesser General Public
7
+ * License as published by the Free Software Foundation; either
8
+ * version 2.1 of the License, or (at your option) any later version.
9
+ *
10
+ * FFmpeg is distributed in the hope that it will be useful,
11
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
12
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13
+ * Lesser General Public License for more details.
14
+ *
15
+ * You should have received a copy of the GNU Lesser General Public
16
+ * License along with FFmpeg; if not, write to the Free Software
17
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18
+ */
19
+
20
+#include "libavutil/opt.h"
21
+#include "avfilter.h"
22
+#include "internal.h"
23
+#include "audio.h"
24
+
25
+typedef struct LFOContext {
26
+    double freq;
27
+    double offset;
28
+    int srate;
29
+    double amount;
30
+    double pwidth;
31
+    double phase;
32
+} LFOContext;
33
+
34
+typedef struct SRContext {
35
+    double target;
36
+    double real;
37
+    double samples;
38
+    double last;
39
+} SRContext;
40
+
41
+typedef struct ACrusherContext {
42
+    const AVClass *class;
43
+
44
+    double level_in;
45
+    double level_out;
46
+    double bits;
47
+    double mix;
48
+    int mode;
49
+    double dc;
50
+    double idc;
51
+    double aa;
52
+    double samples;
53
+    int is_lfo;
54
+    double lforange;
55
+    double lforate;
56
+
57
+    double sqr;
58
+    double aa1;
59
+    double coeff;
60
+    int    round;
61
+    double sov;
62
+    double smin;
63
+    double sdiff;
64
+
65
+    LFOContext lfo;
66
+    SRContext *sr;
67
+} ACrusherContext;
68
+
69
+#define OFFSET(x) offsetof(ACrusherContext, x)
70
+#define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
71
+
72
+static const AVOption acrusher_options[] = {
73
+    { "level_in", "set level in",         OFFSET(level_in),  AV_OPT_TYPE_DOUBLE, {.dbl=1},    0.015625, 64, A },
74
+    { "level_out","set level out",        OFFSET(level_out), AV_OPT_TYPE_DOUBLE, {.dbl=1},    0.015625, 64, A },
75
+    { "bits",     "set bit reduction",    OFFSET(bits),      AV_OPT_TYPE_DOUBLE, {.dbl=8},    1,        64, A },
76
+    { "mix",      "set mix",              OFFSET(mix),       AV_OPT_TYPE_DOUBLE, {.dbl=.5},   0,         1, A },
77
+    { "mode",     "set mode",             OFFSET(mode),      AV_OPT_TYPE_INT,    {.i64=0},    0,         1, A, "mode" },
78
+    {   "lin",    "linear",               0,                 AV_OPT_TYPE_CONST,  {.i64=0},    0,         0, A, "mode" },
79
+    {   "log",    "logarithmic",          0,                 AV_OPT_TYPE_CONST,  {.i64=1},    0,         0, A, "mode" },
80
+    { "dc",       "set DC",               OFFSET(dc),        AV_OPT_TYPE_DOUBLE, {.dbl=1},  .25,         4, A },
81
+    { "aa",       "set anti-aliasing",    OFFSET(aa),        AV_OPT_TYPE_DOUBLE, {.dbl=.5},   0,         1, A },
82
+    { "samples",  "set sample reduction", OFFSET(samples),   AV_OPT_TYPE_DOUBLE, {.dbl=1},    1,       250, A },
83
+    { "lfo",      "enable LFO",           OFFSET(is_lfo),    AV_OPT_TYPE_BOOL,   {.i64=0},    0,         1, A },
84
+    { "lforange", "set LFO depth",        OFFSET(lforange),  AV_OPT_TYPE_DOUBLE, {.dbl=20},   1,       250, A },
85
+    { "lforate",  "set LFO rate",         OFFSET(lforate),   AV_OPT_TYPE_DOUBLE, {.dbl=.3}, .01,       200, A },
86
+    { NULL }
87
+};
88
+
89
+AVFILTER_DEFINE_CLASS(acrusher);
90
+
91
+static double samplereduction(ACrusherContext *s, SRContext *sr, double in)
92
+{
93
+    sr->samples++;
94
+    if (sr->samples >= s->round) {
95
+        sr->target += s->samples;
96
+        sr->real += s->round;
97
+        if (sr->target + s->samples >= sr->real + 1) {
98
+            sr->last = in;
99
+            sr->target = 0;
100
+            sr->real   = 0;
101
+        }
102
+        sr->samples = 0;
103
+    }
104
+    return sr->last;
105
+}
106
+
107
+static double add_dc(double s, double dc, double idc)
108
+{
109
+    return s > 0 ? s * dc : s * idc;
110
+}
111
+
112
+static double remove_dc(double s, double dc, double idc)
113
+{
114
+    return s > 0 ? s * idc : s * dc;
115
+}
116
+
117
+static inline double factor(double y, double k, double aa1, double aa)
118
+{
119
+    return 0.5 * (sin(M_PI * (fabs(y - k) - aa1) / aa - M_PI_2) + 1);
120
+}
121
+
122
+static double bitreduction(ACrusherContext *s, double in)
123
+{
124
+    const double sqr = s->sqr;
125
+    const double coeff = s->coeff;
126
+    const double aa = s->aa;
127
+    const double aa1 = s->aa1;
128
+    double y, k;
129
+
130
+    // add dc
131
+    in = add_dc(in, s->dc, s->idc);
132
+
133
+    // main rounding calculation depending on mode
134
+
135
+    // the idea for anti-aliasing:
136
+    // you need a function f which brings you to the scale, where
137
+    // you want to round and the function f_b (with f(f_b)=id) which
138
+    // brings you back to your original scale.
139
+    //
140
+    // then you can use the logic below in the following way:
141
+    // y = f(in) and k = roundf(y)
142
+    // if (y > k + aa1)
143
+    //      k = f_b(k) + ( f_b(k+1) - f_b(k) ) * 0.5 * (sin(x - PI/2) + 1)
144
+    // if (y < k + aa1)
145
+    //      k = f_b(k) - ( f_b(k+1) - f_b(k) ) * 0.5 * (sin(x - PI/2) + 1)
146
+    //
147
+    // whereas x = (fabs(f(in) - k) - aa1) * PI / aa
148
+    // for both cases.
149
+
150
+    switch (s->mode) {
151
+    case 0:
152
+    default:
153
+        // linear
154
+        y = in * coeff;
155
+        k = roundf(y);
156
+        if (k - aa1 <= y && y <= k + aa1) {
157
+            k /= coeff;
158
+        } else if (y > k + aa1) {
159
+            k = k / coeff + ((k + 1) / coeff - k / coeff) *
160
+                factor(y, k, aa1, aa);
161
+        } else {
162
+            k = k / coeff - (k / coeff - (k - 1) / coeff) *
163
+                factor(y, k, aa1, aa);
164
+        }
165
+        break;
166
+    case 1:
167
+        // logarithmic
168
+        y = sqr * log(fabs(in)) + sqr * sqr;
169
+        k = roundf(y);
170
+        if(!in) {
171
+            k = 0;
172
+        } else if (k - aa1 <= y && y <= k + aa1) {
173
+            k = in / fabs(in) * exp(k / sqr - sqr);
174
+        } else if (y > k + aa1) {
175
+            double x = exp(k / sqr - sqr);
176
+            k = FFSIGN(in) * (x + (exp((k + 1) / sqr - sqr) - x) *
177
+                factor(y, k, aa1, aa));
178
+        } else {
179
+            double x = exp(k / sqr - sqr);
180
+            k = in / fabs(in) * (x - (x - exp((k - 1) / sqr - sqr)) *
181
+                factor(y, k, aa1, aa));
182
+        }
183
+        break;
184
+    }
185
+
186
+    // mix between dry and wet signal
187
+    k += (in - k) * s->mix;
188
+
189
+    // remove dc
190
+    k = remove_dc(k, s->dc, s->idc);
191
+
192
+    return k;
193
+}
194
+
195
+static double lfo_get(LFOContext *lfo)
196
+{
197
+    double phs = FFMIN(100., lfo->phase / FFMIN(1.99, FFMAX(0.01, lfo->pwidth)) + lfo->offset);
198
+    double val;
199
+
200
+    if (phs > 1)
201
+        phs = fmod(phs, 1.);
202
+
203
+    val = sin((phs * 360.) * M_PI / 180);
204
+
205
+    return val * lfo->amount;
206
+}
207
+
208
+static void lfo_advance(LFOContext *lfo, unsigned count)
209
+{
210
+    lfo->phase = fabs(lfo->phase + count * lfo->freq * (1. / lfo->srate));
211
+    if (lfo->phase >= 1.)
212
+        lfo->phase = fmod(lfo->phase, 1.);
213
+}
214
+
215
+static int filter_frame(AVFilterLink *inlink, AVFrame *in)
216
+{
217
+    AVFilterContext *ctx = inlink->dst;
218
+    ACrusherContext *s = ctx->priv;
219
+    AVFilterLink *outlink = ctx->outputs[0];
220
+    AVFrame *out;
221
+    const double *src = (const double *)in->data[0];
222
+    double *dst;
223
+    const double level_in = s->level_in;
224
+    const double level_out = s->level_out;
225
+    const double mix = s->mix;
226
+    int n, c;
227
+
228
+    if (av_frame_is_writable(in)) {
229
+        out = in;
230
+    } else {
231
+        out = ff_get_audio_buffer(inlink, in->nb_samples);
232
+        if (!out) {
233
+            av_frame_free(&in);
234
+            return AVERROR(ENOMEM);
235
+        }
236
+        av_frame_copy_props(out, in);
237
+    }
238
+
239
+    dst = (double *)out->data[0];
240
+    for (n = 0; n < in->nb_samples; n++) {
241
+        if (s->is_lfo) {
242
+            s->samples = s->smin + s->sdiff * (lfo_get(&s->lfo) + 0.5);
243
+            s->round = round(s->samples);
244
+        }
245
+
246
+        for (c = 0; c < inlink->channels; c++) {
247
+            double sample = src[c] * level_in;
248
+
249
+            sample = mix * samplereduction(s, &s->sr[c], sample) + src[c] * (1. - mix) * level_in;
250
+            dst[c] = bitreduction(s, sample) * level_out;
251
+        }
252
+        src += c;
253
+        dst += c;
254
+
255
+        if (s->is_lfo)
256
+            lfo_advance(&s->lfo, 1);
257
+    }
258
+
259
+    if (in != out)
260
+        av_frame_free(&in);
261
+
262
+    return ff_filter_frame(outlink, out);
263
+}
264
+
265
+static int query_formats(AVFilterContext *ctx)
266
+{
267
+    AVFilterFormats *formats;
268
+    AVFilterChannelLayouts *layouts;
269
+    static const enum AVSampleFormat sample_fmts[] = {
270
+        AV_SAMPLE_FMT_DBL,
271
+        AV_SAMPLE_FMT_NONE
272
+    };
273
+    int ret;
274
+
275
+    layouts = ff_all_channel_counts();
276
+    if (!layouts)
277
+        return AVERROR(ENOMEM);
278
+    ret = ff_set_common_channel_layouts(ctx, layouts);
279
+    if (ret < 0)
280
+        return ret;
281
+
282
+    formats = ff_make_format_list(sample_fmts);
283
+    if (!formats)
284
+        return AVERROR(ENOMEM);
285
+    ret = ff_set_common_formats(ctx, formats);
286
+    if (ret < 0)
287
+        return ret;
288
+
289
+    formats = ff_all_samplerates();
290
+    if (!formats)
291
+        return AVERROR(ENOMEM);
292
+    return ff_set_common_samplerates(ctx, formats);
293
+}
294
+
295
+static av_cold void uninit(AVFilterContext *ctx)
296
+{
297
+    ACrusherContext *s = ctx->priv;
298
+
299
+    av_freep(&s->sr);
300
+}
301
+
302
+static int config_input(AVFilterLink *inlink)
303
+{
304
+    AVFilterContext *ctx = inlink->dst;
305
+    ACrusherContext *s = ctx->priv;
306
+    double rad, sun, smax, sov;
307
+
308
+    s->idc = 1. / s->dc;
309
+    s->coeff = exp2(s->bits) - 1;
310
+    s->sqr = sqrt(s->coeff / 2);
311
+    s->aa1 = (1. - s->aa) / 2.;
312
+    s->round = round(s->samples);
313
+    rad = s->lforange / 2.;
314
+    s->smin = FFMAX(s->samples - rad, 1.);
315
+    sun = s->samples - rad - s->smin;
316
+    smax = FFMIN(s->samples + rad, 250.);
317
+    sov = s->samples + rad - smax;
318
+    smax -= sun;
319
+    s->smin -= sov;
320
+    s->sdiff = smax - s->smin;
321
+
322
+    s->lfo.freq = s->lforate;
323
+    s->lfo.pwidth = 1.;
324
+    s->lfo.srate = inlink->sample_rate;
325
+    s->lfo.amount = .5;
326
+
327
+    s->sr = av_calloc(inlink->channels, sizeof(*s->sr));
328
+    if (!s->sr)
329
+        return AVERROR(ENOMEM);
330
+
331
+    return 0;
332
+}
333
+
334
+static const AVFilterPad avfilter_af_acrusher_inputs[] = {
335
+    {
336
+        .name         = "default",
337
+        .type         = AVMEDIA_TYPE_AUDIO,
338
+        .config_props = config_input,
339
+        .filter_frame = filter_frame,
340
+    },
341
+    { NULL }
342
+};
343
+
344
+static const AVFilterPad avfilter_af_acrusher_outputs[] = {
345
+    {
346
+        .name = "default",
347
+        .type = AVMEDIA_TYPE_AUDIO,
348
+    },
349
+    { NULL }
350
+};
351
+
352
+AVFilter ff_af_acrusher = {
353
+    .name          = "acrusher",
354
+    .description   = NULL_IF_CONFIG_SMALL("Reduce audio bit resolution."),
355
+    .priv_size     = sizeof(ACrusherContext),
356
+    .priv_class    = &acrusher_class,
357
+    .uninit        = uninit,
358
+    .query_formats = query_formats,
359
+    .inputs        = avfilter_af_acrusher_inputs,
360
+    .outputs       = avfilter_af_acrusher_outputs,
361
+};
... ...
@@ -48,6 +48,7 @@ void avfilter_register_all(void)
48 48
     REGISTER_FILTER(ABENCH,         abench,         af);
49 49
     REGISTER_FILTER(ACOMPRESSOR,    acompressor,    af);
50 50
     REGISTER_FILTER(ACROSSFADE,     acrossfade,     af);
51
+    REGISTER_FILTER(ACRUSHER,       acrusher,       af);
51 52
     REGISTER_FILTER(ADELAY,         adelay,         af);
52 53
     REGISTER_FILTER(AECHO,          aecho,          af);
53 54
     REGISTER_FILTER(AEMPHASIS,      aemphasis,      af);
... ...
@@ -30,7 +30,7 @@
30 30
 #include "libavutil/version.h"
31 31
 
32 32
 #define LIBAVFILTER_VERSION_MAJOR   6
33
-#define LIBAVFILTER_VERSION_MINOR  50
33
+#define LIBAVFILTER_VERSION_MINOR  51
34 34
 #define LIBAVFILTER_VERSION_MICRO 100
35 35
 
36 36
 #define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \