• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (c) 2011 Stefano Sabatini
3  * Copyright (c) 2012 Justin Ruggles <justin.ruggles@gmail.com>
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 volume filter
25  */
26 
27 #include "libavutil/channel_layout.h"
28 #include "libavutil/common.h"
29 #include "libavutil/eval.h"
30 #include "libavutil/ffmath.h"
31 #include "libavutil/float_dsp.h"
32 #include "libavutil/intreadwrite.h"
33 #include "libavutil/opt.h"
34 #include "libavutil/replaygain.h"
35 
36 #include "audio.h"
37 #include "avfilter.h"
38 #include "formats.h"
39 #include "internal.h"
40 #include "af_volume.h"
41 
42 static const char * const precision_str[] = {
43     "fixed", "float", "double"
44 };
45 
46 static const char *const var_names[] = {
47     "n",                   ///< frame number (starting at zero)
48     "nb_channels",         ///< number of channels
49     "nb_consumed_samples", ///< number of samples consumed by the filter
50     "nb_samples",          ///< number of samples in the current frame
51     "pos",                 ///< position in the file of the frame
52     "pts",                 ///< frame presentation timestamp
53     "sample_rate",         ///< sample rate
54     "startpts",            ///< PTS at start of stream
55     "startt",              ///< time at start of stream
56     "t",                   ///< time in the file of the frame
57     "tb",                  ///< timebase
58     "volume",              ///< last set value
59     NULL
60 };
61 
62 #define OFFSET(x) offsetof(VolumeContext, x)
63 #define A AV_OPT_FLAG_AUDIO_PARAM
64 #define F AV_OPT_FLAG_FILTERING_PARAM
65 #define T AV_OPT_FLAG_RUNTIME_PARAM
66 
67 static const AVOption volume_options[] = {
68     { "volume", "set volume adjustment expression",
69             OFFSET(volume_expr), AV_OPT_TYPE_STRING, { .str = "1.0" }, .flags = A|F|T },
70     { "precision", "select mathematical precision",
71             OFFSET(precision), AV_OPT_TYPE_INT, { .i64 = PRECISION_FLOAT }, PRECISION_FIXED, PRECISION_DOUBLE, A|F, "precision" },
72         { "fixed",  "select 8-bit fixed-point",     0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_FIXED  }, INT_MIN, INT_MAX, A|F, "precision" },
73         { "float",  "select 32-bit floating-point", 0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_FLOAT  }, INT_MIN, INT_MAX, A|F, "precision" },
74         { "double", "select 64-bit floating-point", 0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_DOUBLE }, INT_MIN, INT_MAX, A|F, "precision" },
75     { "eval", "specify when to evaluate expressions", OFFSET(eval_mode), AV_OPT_TYPE_INT, {.i64 = EVAL_MODE_ONCE}, 0, EVAL_MODE_NB-1, .flags = A|F, "eval" },
76          { "once",  "eval volume expression once", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_ONCE},  .flags = A|F, .unit = "eval" },
77          { "frame", "eval volume expression per-frame",                  0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_FRAME}, .flags = A|F, .unit = "eval" },
78     { "replaygain", "Apply replaygain side data when present",
79             OFFSET(replaygain), AV_OPT_TYPE_INT, { .i64 = REPLAYGAIN_DROP }, REPLAYGAIN_DROP, REPLAYGAIN_ALBUM, A|F, "replaygain" },
80         { "drop",   "replaygain side data is dropped", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_DROP   }, 0, 0, A|F, "replaygain" },
81         { "ignore", "replaygain side data is ignored", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_IGNORE }, 0, 0, A|F, "replaygain" },
82         { "track",  "track gain is preferred",         0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_TRACK  }, 0, 0, A|F, "replaygain" },
83         { "album",  "album gain is preferred",         0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_ALBUM  }, 0, 0, A|F, "replaygain" },
84     { "replaygain_preamp", "Apply replaygain pre-amplification",
85             OFFSET(replaygain_preamp), AV_OPT_TYPE_DOUBLE, { .dbl = 0.0 }, -15.0, 15.0, A|F },
86     { "replaygain_noclip", "Apply replaygain clipping prevention",
87             OFFSET(replaygain_noclip), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, A|F },
88     { NULL }
89 };
90 
91 AVFILTER_DEFINE_CLASS(volume);
92 
set_expr(AVExpr ** pexpr,const char * expr,void * log_ctx)93 static int set_expr(AVExpr **pexpr, const char *expr, void *log_ctx)
94 {
95     int ret;
96     AVExpr *old = NULL;
97 
98     if (*pexpr)
99         old = *pexpr;
100     ret = av_expr_parse(pexpr, expr, var_names,
101                         NULL, NULL, NULL, NULL, 0, log_ctx);
102     if (ret < 0) {
103         av_log(log_ctx, AV_LOG_ERROR,
104                "Error when evaluating the volume expression '%s'\n", expr);
105         *pexpr = old;
106         return ret;
107     }
108 
109     av_expr_free(old);
110     return 0;
111 }
112 
init(AVFilterContext * ctx)113 static av_cold int init(AVFilterContext *ctx)
114 {
115     VolumeContext *vol = ctx->priv;
116 
117     vol->fdsp = avpriv_float_dsp_alloc(0);
118     if (!vol->fdsp)
119         return AVERROR(ENOMEM);
120 
121     return set_expr(&vol->volume_pexpr, vol->volume_expr, ctx);
122 }
123 
uninit(AVFilterContext * ctx)124 static av_cold void uninit(AVFilterContext *ctx)
125 {
126     VolumeContext *vol = ctx->priv;
127     av_expr_free(vol->volume_pexpr);
128     av_opt_free(vol);
129     av_freep(&vol->fdsp);
130 }
131 
query_formats(AVFilterContext * ctx)132 static int query_formats(AVFilterContext *ctx)
133 {
134     VolumeContext *vol = ctx->priv;
135     static const enum AVSampleFormat sample_fmts[][7] = {
136         [PRECISION_FIXED] = {
137             AV_SAMPLE_FMT_U8,
138             AV_SAMPLE_FMT_U8P,
139             AV_SAMPLE_FMT_S16,
140             AV_SAMPLE_FMT_S16P,
141             AV_SAMPLE_FMT_S32,
142             AV_SAMPLE_FMT_S32P,
143             AV_SAMPLE_FMT_NONE
144         },
145         [PRECISION_FLOAT] = {
146             AV_SAMPLE_FMT_FLT,
147             AV_SAMPLE_FMT_FLTP,
148             AV_SAMPLE_FMT_NONE
149         },
150         [PRECISION_DOUBLE] = {
151             AV_SAMPLE_FMT_DBL,
152             AV_SAMPLE_FMT_DBLP,
153             AV_SAMPLE_FMT_NONE
154         }
155     };
156     int ret = ff_set_common_all_channel_counts(ctx);
157     if (ret < 0)
158         return ret;
159 
160     ret = ff_set_common_formats_from_list(ctx, sample_fmts[vol->precision]);
161     if (ret < 0)
162         return ret;
163 
164     return ff_set_common_all_samplerates(ctx);
165 }
166 
scale_samples_u8(uint8_t * dst,const uint8_t * src,int nb_samples,int volume)167 static inline void scale_samples_u8(uint8_t *dst, const uint8_t *src,
168                                     int nb_samples, int volume)
169 {
170     int i;
171     for (i = 0; i < nb_samples; i++)
172         dst[i] = av_clip_uint8(((((int64_t)src[i] - 128) * volume + 128) >> 8) + 128);
173 }
174 
scale_samples_u8_small(uint8_t * dst,const uint8_t * src,int nb_samples,int volume)175 static inline void scale_samples_u8_small(uint8_t *dst, const uint8_t *src,
176                                           int nb_samples, int volume)
177 {
178     int i;
179     for (i = 0; i < nb_samples; i++)
180         dst[i] = av_clip_uint8((((src[i] - 128) * volume + 128) >> 8) + 128);
181 }
182 
scale_samples_s16(uint8_t * dst,const uint8_t * src,int nb_samples,int volume)183 static inline void scale_samples_s16(uint8_t *dst, const uint8_t *src,
184                                      int nb_samples, int volume)
185 {
186     int i;
187     int16_t *smp_dst       = (int16_t *)dst;
188     const int16_t *smp_src = (const int16_t *)src;
189     for (i = 0; i < nb_samples; i++)
190         smp_dst[i] = av_clip_int16(((int64_t)smp_src[i] * volume + 128) >> 8);
191 }
192 
scale_samples_s16_small(uint8_t * dst,const uint8_t * src,int nb_samples,int volume)193 static inline void scale_samples_s16_small(uint8_t *dst, const uint8_t *src,
194                                            int nb_samples, int volume)
195 {
196     int i;
197     int16_t *smp_dst       = (int16_t *)dst;
198     const int16_t *smp_src = (const int16_t *)src;
199     for (i = 0; i < nb_samples; i++)
200         smp_dst[i] = av_clip_int16((smp_src[i] * volume + 128) >> 8);
201 }
202 
scale_samples_s32(uint8_t * dst,const uint8_t * src,int nb_samples,int volume)203 static inline void scale_samples_s32(uint8_t *dst, const uint8_t *src,
204                                      int nb_samples, int volume)
205 {
206     int i;
207     int32_t *smp_dst       = (int32_t *)dst;
208     const int32_t *smp_src = (const int32_t *)src;
209     for (i = 0; i < nb_samples; i++)
210         smp_dst[i] = av_clipl_int32((((int64_t)smp_src[i] * volume + 128) >> 8));
211 }
212 
volume_init(VolumeContext * vol)213 static av_cold void volume_init(VolumeContext *vol)
214 {
215     vol->samples_align = 1;
216 
217     switch (av_get_packed_sample_fmt(vol->sample_fmt)) {
218     case AV_SAMPLE_FMT_U8:
219         if (vol->volume_i < 0x1000000)
220             vol->scale_samples = scale_samples_u8_small;
221         else
222             vol->scale_samples = scale_samples_u8;
223         break;
224     case AV_SAMPLE_FMT_S16:
225         if (vol->volume_i < 0x10000)
226             vol->scale_samples = scale_samples_s16_small;
227         else
228             vol->scale_samples = scale_samples_s16;
229         break;
230     case AV_SAMPLE_FMT_S32:
231         vol->scale_samples = scale_samples_s32;
232         break;
233     case AV_SAMPLE_FMT_FLT:
234         vol->samples_align = 4;
235         break;
236     case AV_SAMPLE_FMT_DBL:
237         vol->samples_align = 8;
238         break;
239     }
240 
241 #if ARCH_X86
242     ff_volume_init_x86(vol);
243 #endif
244 }
245 
set_volume(AVFilterContext * ctx)246 static int set_volume(AVFilterContext *ctx)
247 {
248     VolumeContext *vol = ctx->priv;
249 
250     vol->volume = av_expr_eval(vol->volume_pexpr, vol->var_values, NULL);
251     if (isnan(vol->volume)) {
252         if (vol->eval_mode == EVAL_MODE_ONCE) {
253             av_log(ctx, AV_LOG_ERROR, "Invalid value NaN for volume\n");
254             return AVERROR(EINVAL);
255         } else {
256             av_log(ctx, AV_LOG_WARNING, "Invalid value NaN for volume, setting to 0\n");
257             vol->volume = 0;
258         }
259     }
260     vol->var_values[VAR_VOLUME] = vol->volume;
261 
262     av_log(ctx, AV_LOG_VERBOSE, "n:%f t:%f pts:%f precision:%s ",
263            vol->var_values[VAR_N], vol->var_values[VAR_T], vol->var_values[VAR_PTS],
264            precision_str[vol->precision]);
265 
266     if (vol->precision == PRECISION_FIXED) {
267         vol->volume_i = (int)(vol->volume * 256 + 0.5);
268         vol->volume   = vol->volume_i / 256.0;
269         av_log(ctx, AV_LOG_VERBOSE, "volume_i:%d/255 ", vol->volume_i);
270     }
271     av_log(ctx, AV_LOG_VERBOSE, "volume:%f volume_dB:%f\n",
272            vol->volume, 20.0*log10(vol->volume));
273 
274     volume_init(vol);
275     return 0;
276 }
277 
config_output(AVFilterLink * outlink)278 static int config_output(AVFilterLink *outlink)
279 {
280     AVFilterContext *ctx = outlink->src;
281     VolumeContext *vol   = ctx->priv;
282     AVFilterLink *inlink = ctx->inputs[0];
283 
284     vol->sample_fmt = inlink->format;
285     vol->channels   = inlink->ch_layout.nb_channels;
286     vol->planes     = av_sample_fmt_is_planar(inlink->format) ? vol->channels : 1;
287 
288     vol->var_values[VAR_N] =
289     vol->var_values[VAR_NB_CONSUMED_SAMPLES] =
290     vol->var_values[VAR_NB_SAMPLES] =
291     vol->var_values[VAR_POS] =
292     vol->var_values[VAR_PTS] =
293     vol->var_values[VAR_STARTPTS] =
294     vol->var_values[VAR_STARTT] =
295     vol->var_values[VAR_T] =
296     vol->var_values[VAR_VOLUME] = NAN;
297 
298     vol->var_values[VAR_NB_CHANNELS] = inlink->ch_layout.nb_channels;
299     vol->var_values[VAR_TB]          = av_q2d(inlink->time_base);
300     vol->var_values[VAR_SAMPLE_RATE] = inlink->sample_rate;
301 
302     av_log(inlink->src, AV_LOG_VERBOSE, "tb:%f sample_rate:%f nb_channels:%f\n",
303            vol->var_values[VAR_TB],
304            vol->var_values[VAR_SAMPLE_RATE],
305            vol->var_values[VAR_NB_CHANNELS]);
306 
307     return set_volume(ctx);
308 }
309 
process_command(AVFilterContext * ctx,const char * cmd,const char * args,char * res,int res_len,int flags)310 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
311                            char *res, int res_len, int flags)
312 {
313     VolumeContext *vol = ctx->priv;
314     int ret = AVERROR(ENOSYS);
315 
316     if (!strcmp(cmd, "volume")) {
317         if ((ret = set_expr(&vol->volume_pexpr, args, ctx)) < 0)
318             return ret;
319         if (vol->eval_mode == EVAL_MODE_ONCE)
320             set_volume(ctx);
321     }
322 
323     return ret;
324 }
325 
filter_frame(AVFilterLink * inlink,AVFrame * buf)326 static int filter_frame(AVFilterLink *inlink, AVFrame *buf)
327 {
328     AVFilterContext *ctx = inlink->dst;
329     VolumeContext *vol    = inlink->dst->priv;
330     AVFilterLink *outlink = inlink->dst->outputs[0];
331     int nb_samples        = buf->nb_samples;
332     AVFrame *out_buf;
333     int64_t pos;
334     AVFrameSideData *sd = av_frame_get_side_data(buf, AV_FRAME_DATA_REPLAYGAIN);
335     int ret;
336 
337     if (sd && vol->replaygain != REPLAYGAIN_IGNORE) {
338         if (vol->replaygain != REPLAYGAIN_DROP) {
339             AVReplayGain *replaygain = (AVReplayGain*)sd->data;
340             int32_t gain  = 100000;
341             uint32_t peak = 100000;
342             float g, p;
343 
344             if (vol->replaygain == REPLAYGAIN_TRACK &&
345                 replaygain->track_gain != INT32_MIN) {
346                 gain = replaygain->track_gain;
347 
348                 if (replaygain->track_peak != 0)
349                     peak = replaygain->track_peak;
350             } else if (replaygain->album_gain != INT32_MIN) {
351                 gain = replaygain->album_gain;
352 
353                 if (replaygain->album_peak != 0)
354                     peak = replaygain->album_peak;
355             } else {
356                 av_log(inlink->dst, AV_LOG_WARNING, "Both ReplayGain gain "
357                        "values are unknown.\n");
358             }
359             g = gain / 100000.0f;
360             p = peak / 100000.0f;
361 
362             av_log(inlink->dst, AV_LOG_VERBOSE,
363                    "Using gain %f dB from replaygain side data.\n", g);
364 
365             vol->volume   = ff_exp10((g + vol->replaygain_preamp) / 20);
366             if (vol->replaygain_noclip)
367                 vol->volume = FFMIN(vol->volume, 1.0 / p);
368             vol->volume_i = (int)(vol->volume * 256 + 0.5);
369 
370             volume_init(vol);
371         }
372         av_frame_remove_side_data(buf, AV_FRAME_DATA_REPLAYGAIN);
373     }
374 
375     if (isnan(vol->var_values[VAR_STARTPTS])) {
376         vol->var_values[VAR_STARTPTS] = TS2D(buf->pts);
377         vol->var_values[VAR_STARTT  ] = TS2T(buf->pts, inlink->time_base);
378     }
379     vol->var_values[VAR_PTS] = TS2D(buf->pts);
380     vol->var_values[VAR_T  ] = TS2T(buf->pts, inlink->time_base);
381     vol->var_values[VAR_N  ] = inlink->frame_count_out;
382 
383     pos = buf->pkt_pos;
384     vol->var_values[VAR_POS] = pos == -1 ? NAN : pos;
385     if (vol->eval_mode == EVAL_MODE_FRAME)
386         set_volume(ctx);
387 
388     if (vol->volume == 1.0 || vol->volume_i == 256) {
389         out_buf = buf;
390         goto end;
391     }
392 
393     /* do volume scaling in-place if input buffer is writable */
394     if (av_frame_is_writable(buf)
395             && (vol->precision != PRECISION_FIXED || vol->volume_i > 0)) {
396         out_buf = buf;
397     } else {
398         out_buf = ff_get_audio_buffer(outlink, nb_samples);
399         if (!out_buf) {
400             av_frame_free(&buf);
401             return AVERROR(ENOMEM);
402         }
403         ret = av_frame_copy_props(out_buf, buf);
404         if (ret < 0) {
405             av_frame_free(&out_buf);
406             av_frame_free(&buf);
407             return ret;
408         }
409     }
410 
411     if (vol->precision != PRECISION_FIXED || vol->volume_i > 0) {
412         int p, plane_samples;
413 
414         if (av_sample_fmt_is_planar(buf->format))
415             plane_samples = FFALIGN(nb_samples, vol->samples_align);
416         else
417             plane_samples = FFALIGN(nb_samples * vol->channels, vol->samples_align);
418 
419         if (vol->precision == PRECISION_FIXED) {
420             for (p = 0; p < vol->planes; p++) {
421                 vol->scale_samples(out_buf->extended_data[p],
422                                    buf->extended_data[p], plane_samples,
423                                    vol->volume_i);
424             }
425         } else if (av_get_packed_sample_fmt(vol->sample_fmt) == AV_SAMPLE_FMT_FLT) {
426             for (p = 0; p < vol->planes; p++) {
427                 vol->fdsp->vector_fmul_scalar((float *)out_buf->extended_data[p],
428                                              (const float *)buf->extended_data[p],
429                                              vol->volume, plane_samples);
430             }
431         } else {
432             for (p = 0; p < vol->planes; p++) {
433                 vol->fdsp->vector_dmul_scalar((double *)out_buf->extended_data[p],
434                                              (const double *)buf->extended_data[p],
435                                              vol->volume, plane_samples);
436             }
437         }
438     }
439 
440     emms_c();
441 
442     if (buf != out_buf)
443         av_frame_free(&buf);
444 
445 end:
446     vol->var_values[VAR_NB_CONSUMED_SAMPLES] += out_buf->nb_samples;
447     return ff_filter_frame(outlink, out_buf);
448 }
449 
450 static const AVFilterPad avfilter_af_volume_inputs[] = {
451     {
452         .name           = "default",
453         .type           = AVMEDIA_TYPE_AUDIO,
454         .filter_frame   = filter_frame,
455     },
456 };
457 
458 static const AVFilterPad avfilter_af_volume_outputs[] = {
459     {
460         .name         = "default",
461         .type         = AVMEDIA_TYPE_AUDIO,
462         .config_props = config_output,
463     },
464 };
465 
466 const AVFilter ff_af_volume = {
467     .name           = "volume",
468     .description    = NULL_IF_CONFIG_SMALL("Change input volume."),
469     .priv_size      = sizeof(VolumeContext),
470     .priv_class     = &volume_class,
471     .init           = init,
472     .uninit         = uninit,
473     FILTER_INPUTS(avfilter_af_volume_inputs),
474     FILTER_OUTPUTS(avfilter_af_volume_outputs),
475     FILTER_QUERY_FUNC(query_formats),
476     .flags          = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
477     .process_command = process_command,
478 };
479