1 /*
2 * Copyright (c) 2013 Paul B Mahol
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 #include "libavutil/avassert.h"
22 #include "libavutil/avstring.h"
23 #include "libavutil/opt.h"
24 #include "libavutil/samplefmt.h"
25 #include "avfilter.h"
26 #include "audio.h"
27 #include "filters.h"
28 #include "internal.h"
29
30 typedef struct AudioEchoContext {
31 const AVClass *class;
32 float in_gain, out_gain;
33 char *delays, *decays;
34 float *delay, *decay;
35 int nb_echoes;
36 int delay_index;
37 uint8_t **delayptrs;
38 int max_samples, fade_out;
39 int *samples;
40 int eof;
41 int64_t next_pts;
42
43 void (*echo_samples)(struct AudioEchoContext *ctx, uint8_t **delayptrs,
44 uint8_t * const *src, uint8_t **dst,
45 int nb_samples, int channels);
46 } AudioEchoContext;
47
48 #define OFFSET(x) offsetof(AudioEchoContext, x)
49 #define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
50
51 static const AVOption aecho_options[] = {
52 { "in_gain", "set signal input gain", OFFSET(in_gain), AV_OPT_TYPE_FLOAT, {.dbl=0.6}, 0, 1, A },
53 { "out_gain", "set signal output gain", OFFSET(out_gain), AV_OPT_TYPE_FLOAT, {.dbl=0.3}, 0, 1, A },
54 { "delays", "set list of signal delays", OFFSET(delays), AV_OPT_TYPE_STRING, {.str="1000"}, 0, 0, A },
55 { "decays", "set list of signal decays", OFFSET(decays), AV_OPT_TYPE_STRING, {.str="0.5"}, 0, 0, A },
56 { NULL }
57 };
58
59 AVFILTER_DEFINE_CLASS(aecho);
60
count_items(char * item_str,int * nb_items)61 static void count_items(char *item_str, int *nb_items)
62 {
63 char *p;
64
65 *nb_items = 1;
66 for (p = item_str; *p; p++) {
67 if (*p == '|')
68 (*nb_items)++;
69 }
70
71 }
72
fill_items(char * item_str,int * nb_items,float * items)73 static void fill_items(char *item_str, int *nb_items, float *items)
74 {
75 char *p, *saveptr = NULL;
76 int i, new_nb_items = 0;
77
78 p = item_str;
79 for (i = 0; i < *nb_items; i++) {
80 char *tstr = av_strtok(p, "|", &saveptr);
81 p = NULL;
82 if (tstr)
83 new_nb_items += av_sscanf(tstr, "%f", &items[new_nb_items]) == 1;
84 }
85
86 *nb_items = new_nb_items;
87 }
88
uninit(AVFilterContext * ctx)89 static av_cold void uninit(AVFilterContext *ctx)
90 {
91 AudioEchoContext *s = ctx->priv;
92
93 av_freep(&s->delay);
94 av_freep(&s->decay);
95 av_freep(&s->samples);
96
97 if (s->delayptrs)
98 av_freep(&s->delayptrs[0]);
99 av_freep(&s->delayptrs);
100 }
101
init(AVFilterContext * ctx)102 static av_cold int init(AVFilterContext *ctx)
103 {
104 AudioEchoContext *s = ctx->priv;
105 int nb_delays, nb_decays, i;
106
107 if (!s->delays || !s->decays) {
108 av_log(ctx, AV_LOG_ERROR, "Missing delays and/or decays.\n");
109 return AVERROR(EINVAL);
110 }
111
112 count_items(s->delays, &nb_delays);
113 count_items(s->decays, &nb_decays);
114
115 s->delay = av_realloc_f(s->delay, nb_delays, sizeof(*s->delay));
116 s->decay = av_realloc_f(s->decay, nb_decays, sizeof(*s->decay));
117 if (!s->delay || !s->decay)
118 return AVERROR(ENOMEM);
119
120 fill_items(s->delays, &nb_delays, s->delay);
121 fill_items(s->decays, &nb_decays, s->decay);
122
123 if (nb_delays != nb_decays) {
124 av_log(ctx, AV_LOG_ERROR, "Number of delays %d differs from number of decays %d.\n", nb_delays, nb_decays);
125 return AVERROR(EINVAL);
126 }
127
128 s->nb_echoes = nb_delays;
129 if (!s->nb_echoes) {
130 av_log(ctx, AV_LOG_ERROR, "At least one decay & delay must be set.\n");
131 return AVERROR(EINVAL);
132 }
133
134 s->samples = av_realloc_f(s->samples, nb_delays, sizeof(*s->samples));
135 if (!s->samples)
136 return AVERROR(ENOMEM);
137
138 for (i = 0; i < nb_delays; i++) {
139 if (s->delay[i] <= 0 || s->delay[i] > 90000) {
140 av_log(ctx, AV_LOG_ERROR, "delay[%d]: %f is out of allowed range: (0, 90000]\n", i, s->delay[i]);
141 return AVERROR(EINVAL);
142 }
143 if (s->decay[i] <= 0 || s->decay[i] > 1) {
144 av_log(ctx, AV_LOG_ERROR, "decay[%d]: %f is out of allowed range: (0, 1]\n", i, s->decay[i]);
145 return AVERROR(EINVAL);
146 }
147 }
148
149 s->next_pts = AV_NOPTS_VALUE;
150
151 av_log(ctx, AV_LOG_DEBUG, "nb_echoes:%d\n", s->nb_echoes);
152 return 0;
153 }
154
query_formats(AVFilterContext * ctx)155 static int query_formats(AVFilterContext *ctx)
156 {
157 AVFilterChannelLayouts *layouts;
158 AVFilterFormats *formats;
159 static const enum AVSampleFormat sample_fmts[] = {
160 AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_S32P,
161 AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_DBLP,
162 AV_SAMPLE_FMT_NONE
163 };
164 int ret;
165
166 layouts = ff_all_channel_counts();
167 if (!layouts)
168 return AVERROR(ENOMEM);
169 ret = ff_set_common_channel_layouts(ctx, layouts);
170 if (ret < 0)
171 return ret;
172
173 formats = ff_make_format_list(sample_fmts);
174 if (!formats)
175 return AVERROR(ENOMEM);
176 ret = ff_set_common_formats(ctx, formats);
177 if (ret < 0)
178 return ret;
179
180 formats = ff_all_samplerates();
181 if (!formats)
182 return AVERROR(ENOMEM);
183 return ff_set_common_samplerates(ctx, formats);
184 }
185
186 #define MOD(a, b) (((a) >= (b)) ? (a) - (b) : (a))
187
188 #define ECHO(name, type, min, max) \
189 static void echo_samples_## name ##p(AudioEchoContext *ctx, \
190 uint8_t **delayptrs, \
191 uint8_t * const *src, uint8_t **dst, \
192 int nb_samples, int channels) \
193 { \
194 const double out_gain = ctx->out_gain; \
195 const double in_gain = ctx->in_gain; \
196 const int nb_echoes = ctx->nb_echoes; \
197 const int max_samples = ctx->max_samples; \
198 int i, j, chan, av_uninit(index); \
199 \
200 av_assert1(channels > 0); /* would corrupt delay_index */ \
201 \
202 for (chan = 0; chan < channels; chan++) { \
203 const type *s = (type *)src[chan]; \
204 type *d = (type *)dst[chan]; \
205 type *dbuf = (type *)delayptrs[chan]; \
206 \
207 index = ctx->delay_index; \
208 for (i = 0; i < nb_samples; i++, s++, d++) { \
209 double out, in; \
210 \
211 in = *s; \
212 out = in * in_gain; \
213 for (j = 0; j < nb_echoes; j++) { \
214 int ix = index + max_samples - ctx->samples[j]; \
215 ix = MOD(ix, max_samples); \
216 out += dbuf[ix] * ctx->decay[j]; \
217 } \
218 out *= out_gain; \
219 \
220 *d = av_clipd(out, min, max); \
221 dbuf[index] = in; \
222 \
223 index = MOD(index + 1, max_samples); \
224 } \
225 } \
226 ctx->delay_index = index; \
227 }
228
229 ECHO(dbl, double, -1.0, 1.0 )
230 ECHO(flt, float, -1.0, 1.0 )
ECHO(s16,int16_t,INT16_MIN,INT16_MAX)231 ECHO(s16, int16_t, INT16_MIN, INT16_MAX)
232 ECHO(s32, int32_t, INT32_MIN, INT32_MAX)
233
234 static int config_output(AVFilterLink *outlink)
235 {
236 AVFilterContext *ctx = outlink->src;
237 AudioEchoContext *s = ctx->priv;
238 float volume = 1.0;
239 int i;
240
241 for (i = 0; i < s->nb_echoes; i++) {
242 s->samples[i] = s->delay[i] * outlink->sample_rate / 1000.0;
243 s->max_samples = FFMAX(s->max_samples, s->samples[i]);
244 volume += s->decay[i];
245 }
246
247 if (s->max_samples <= 0) {
248 av_log(ctx, AV_LOG_ERROR, "Nothing to echo - missing delay samples.\n");
249 return AVERROR(EINVAL);
250 }
251 s->fade_out = s->max_samples;
252
253 if (volume * s->in_gain * s->out_gain > 1.0)
254 av_log(ctx, AV_LOG_WARNING,
255 "out_gain %f can cause saturation of output\n", s->out_gain);
256
257 switch (outlink->format) {
258 case AV_SAMPLE_FMT_DBLP: s->echo_samples = echo_samples_dblp; break;
259 case AV_SAMPLE_FMT_FLTP: s->echo_samples = echo_samples_fltp; break;
260 case AV_SAMPLE_FMT_S16P: s->echo_samples = echo_samples_s16p; break;
261 case AV_SAMPLE_FMT_S32P: s->echo_samples = echo_samples_s32p; break;
262 }
263
264
265 if (s->delayptrs)
266 av_freep(&s->delayptrs[0]);
267 av_freep(&s->delayptrs);
268
269 return av_samples_alloc_array_and_samples(&s->delayptrs, NULL,
270 outlink->channels,
271 s->max_samples,
272 outlink->format, 0);
273 }
274
filter_frame(AVFilterLink * inlink,AVFrame * frame)275 static int filter_frame(AVFilterLink *inlink, AVFrame *frame)
276 {
277 AVFilterContext *ctx = inlink->dst;
278 AudioEchoContext *s = ctx->priv;
279 AVFrame *out_frame;
280
281 if (av_frame_is_writable(frame)) {
282 out_frame = frame;
283 } else {
284 out_frame = ff_get_audio_buffer(ctx->outputs[0], frame->nb_samples);
285 if (!out_frame) {
286 av_frame_free(&frame);
287 return AVERROR(ENOMEM);
288 }
289 av_frame_copy_props(out_frame, frame);
290 }
291
292 s->echo_samples(s, s->delayptrs, frame->extended_data, out_frame->extended_data,
293 frame->nb_samples, inlink->channels);
294
295 s->next_pts = frame->pts + av_rescale_q(frame->nb_samples, (AVRational){1, inlink->sample_rate}, inlink->time_base);
296
297 if (frame != out_frame)
298 av_frame_free(&frame);
299
300 return ff_filter_frame(ctx->outputs[0], out_frame);
301 }
302
request_frame(AVFilterLink * outlink)303 static int request_frame(AVFilterLink *outlink)
304 {
305 AVFilterContext *ctx = outlink->src;
306 AudioEchoContext *s = ctx->priv;
307 int nb_samples = FFMIN(s->fade_out, 2048);
308 AVFrame *frame = ff_get_audio_buffer(outlink, nb_samples);
309
310 if (!frame)
311 return AVERROR(ENOMEM);
312 s->fade_out -= nb_samples;
313
314 av_samples_set_silence(frame->extended_data, 0,
315 frame->nb_samples,
316 outlink->channels,
317 frame->format);
318
319 s->echo_samples(s, s->delayptrs, frame->extended_data, frame->extended_data,
320 frame->nb_samples, outlink->channels);
321
322 frame->pts = s->next_pts;
323 if (s->next_pts != AV_NOPTS_VALUE)
324 s->next_pts += av_rescale_q(nb_samples, (AVRational){1, outlink->sample_rate}, outlink->time_base);
325
326 return ff_filter_frame(outlink, frame);
327 }
328
activate(AVFilterContext * ctx)329 static int activate(AVFilterContext *ctx)
330 {
331 AVFilterLink *inlink = ctx->inputs[0];
332 AVFilterLink *outlink = ctx->outputs[0];
333 AudioEchoContext *s = ctx->priv;
334 AVFrame *in;
335 int ret, status;
336 int64_t pts;
337
338 FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink);
339
340 ret = ff_inlink_consume_frame(inlink, &in);
341 if (ret < 0)
342 return ret;
343 if (ret > 0)
344 return filter_frame(inlink, in);
345
346 if (!s->eof && ff_inlink_acknowledge_status(inlink, &status, &pts)) {
347 if (status == AVERROR_EOF)
348 s->eof = 1;
349 }
350
351 if (s->eof && s->fade_out <= 0) {
352 ff_outlink_set_status(outlink, AVERROR_EOF, s->next_pts);
353 return 0;
354 }
355
356 if (!s->eof)
357 FF_FILTER_FORWARD_WANTED(outlink, inlink);
358
359 return request_frame(outlink);
360 }
361
362 static const AVFilterPad aecho_inputs[] = {
363 {
364 .name = "default",
365 .type = AVMEDIA_TYPE_AUDIO,
366 },
367 { NULL }
368 };
369
370 static const AVFilterPad aecho_outputs[] = {
371 {
372 .name = "default",
373 .config_props = config_output,
374 .type = AVMEDIA_TYPE_AUDIO,
375 },
376 { NULL }
377 };
378
379 AVFilter ff_af_aecho = {
380 .name = "aecho",
381 .description = NULL_IF_CONFIG_SMALL("Add echoing to the audio."),
382 .query_formats = query_formats,
383 .priv_size = sizeof(AudioEchoContext),
384 .priv_class = &aecho_class,
385 .init = init,
386 .activate = activate,
387 .uninit = uninit,
388 .inputs = aecho_inputs,
389 .outputs = aecho_outputs,
390 };
391