blob: 68134033ec05ee494ead0d0b47043f0d6f37e324
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 | |
66 | static const AVOption volume_options[] = { |
67 | { "volume", "set volume adjustment expression", |
68 | OFFSET(volume_expr), AV_OPT_TYPE_STRING, { .str = "1.0" }, .flags = A|F }, |
69 | { "precision", "select mathematical precision", |
70 | OFFSET(precision), AV_OPT_TYPE_INT, { .i64 = PRECISION_FLOAT }, PRECISION_FIXED, PRECISION_DOUBLE, A|F, "precision" }, |
71 | { "fixed", "select 8-bit fixed-point", 0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_FIXED }, INT_MIN, INT_MAX, A|F, "precision" }, |
72 | { "float", "select 32-bit floating-point", 0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_FLOAT }, INT_MIN, INT_MAX, A|F, "precision" }, |
73 | { "double", "select 64-bit floating-point", 0, AV_OPT_TYPE_CONST, { .i64 = PRECISION_DOUBLE }, INT_MIN, INT_MAX, A|F, "precision" }, |
74 | { "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" }, |
75 | { "once", "eval volume expression once", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_ONCE}, .flags = A|F, .unit = "eval" }, |
76 | { "frame", "eval volume expression per-frame", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_FRAME}, .flags = A|F, .unit = "eval" }, |
77 | { "replaygain", "Apply replaygain side data when present", |
78 | OFFSET(replaygain), AV_OPT_TYPE_INT, { .i64 = REPLAYGAIN_DROP }, REPLAYGAIN_DROP, REPLAYGAIN_ALBUM, A|F, "replaygain" }, |
79 | { "drop", "replaygain side data is dropped", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_DROP }, 0, 0, A|F, "replaygain" }, |
80 | { "ignore", "replaygain side data is ignored", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_IGNORE }, 0, 0, A|F, "replaygain" }, |
81 | { "track", "track gain is preferred", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_TRACK }, 0, 0, A|F, "replaygain" }, |
82 | { "album", "album gain is preferred", 0, AV_OPT_TYPE_CONST, { .i64 = REPLAYGAIN_ALBUM }, 0, 0, A|F, "replaygain" }, |
83 | { "replaygain_preamp", "Apply replaygain pre-amplification", |
84 | OFFSET(replaygain_preamp), AV_OPT_TYPE_DOUBLE, { .dbl = 0.0 }, -15.0, 15.0, A|F }, |
85 | { "replaygain_noclip", "Apply replaygain clipping prevention", |
86 | OFFSET(replaygain_noclip), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, A|F }, |
87 | { NULL } |
88 | }; |
89 | |
90 | AVFILTER_DEFINE_CLASS(volume); |
91 | |
92 | static int set_expr(AVExpr **pexpr, const char *expr, void *log_ctx) |
93 | { |
94 | int ret; |
95 | AVExpr *old = NULL; |
96 | |
97 | if (*pexpr) |
98 | old = *pexpr; |
99 | ret = av_expr_parse(pexpr, expr, var_names, |
100 | NULL, NULL, NULL, NULL, 0, log_ctx); |
101 | if (ret < 0) { |
102 | av_log(log_ctx, AV_LOG_ERROR, |
103 | "Error when evaluating the volume expression '%s'\n", expr); |
104 | *pexpr = old; |
105 | return ret; |
106 | } |
107 | |
108 | av_expr_free(old); |
109 | return 0; |
110 | } |
111 | |
112 | static av_cold int init(AVFilterContext *ctx) |
113 | { |
114 | VolumeContext *vol = ctx->priv; |
115 | |
116 | vol->fdsp = avpriv_float_dsp_alloc(0); |
117 | if (!vol->fdsp) |
118 | return AVERROR(ENOMEM); |
119 | |
120 | return set_expr(&vol->volume_pexpr, vol->volume_expr, ctx); |
121 | } |
122 | |
123 | static av_cold void uninit(AVFilterContext *ctx) |
124 | { |
125 | VolumeContext *vol = ctx->priv; |
126 | av_expr_free(vol->volume_pexpr); |
127 | av_opt_free(vol); |
128 | av_freep(&vol->fdsp); |
129 | } |
130 | |
131 | static int query_formats(AVFilterContext *ctx) |
132 | { |
133 | VolumeContext *vol = ctx->priv; |
134 | AVFilterFormats *formats = NULL; |
135 | AVFilterChannelLayouts *layouts; |
136 | static const enum AVSampleFormat sample_fmts[][7] = { |
137 | [PRECISION_FIXED] = { |
138 | AV_SAMPLE_FMT_U8, |
139 | AV_SAMPLE_FMT_U8P, |
140 | AV_SAMPLE_FMT_S16, |
141 | AV_SAMPLE_FMT_S16P, |
142 | AV_SAMPLE_FMT_S32, |
143 | AV_SAMPLE_FMT_S32P, |
144 | AV_SAMPLE_FMT_NONE |
145 | }, |
146 | [PRECISION_FLOAT] = { |
147 | AV_SAMPLE_FMT_FLT, |
148 | AV_SAMPLE_FMT_FLTP, |
149 | AV_SAMPLE_FMT_NONE |
150 | }, |
151 | [PRECISION_DOUBLE] = { |
152 | AV_SAMPLE_FMT_DBL, |
153 | AV_SAMPLE_FMT_DBLP, |
154 | AV_SAMPLE_FMT_NONE |
155 | } |
156 | }; |
157 | int ret; |
158 | |
159 | layouts = ff_all_channel_counts(); |
160 | if (!layouts) |
161 | return AVERROR(ENOMEM); |
162 | ret = ff_set_common_channel_layouts(ctx, layouts); |
163 | if (ret < 0) |
164 | return ret; |
165 | |
166 | formats = ff_make_format_list(sample_fmts[vol->precision]); |
167 | if (!formats) |
168 | return AVERROR(ENOMEM); |
169 | ret = ff_set_common_formats(ctx, formats); |
170 | if (ret < 0) |
171 | return ret; |
172 | |
173 | formats = ff_all_samplerates(); |
174 | if (!formats) |
175 | return AVERROR(ENOMEM); |
176 | return ff_set_common_samplerates(ctx, formats); |
177 | } |
178 | |
179 | static inline void scale_samples_u8(uint8_t *dst, const uint8_t *src, |
180 | int nb_samples, int volume) |
181 | { |
182 | int i; |
183 | for (i = 0; i < nb_samples; i++) |
184 | dst[i] = av_clip_uint8(((((int64_t)src[i] - 128) * volume + 128) >> 8) + 128); |
185 | } |
186 | |
187 | static inline void scale_samples_u8_small(uint8_t *dst, const uint8_t *src, |
188 | int nb_samples, int volume) |
189 | { |
190 | int i; |
191 | for (i = 0; i < nb_samples; i++) |
192 | dst[i] = av_clip_uint8((((src[i] - 128) * volume + 128) >> 8) + 128); |
193 | } |
194 | |
195 | static inline void scale_samples_s16(uint8_t *dst, const uint8_t *src, |
196 | int nb_samples, int volume) |
197 | { |
198 | int i; |
199 | int16_t *smp_dst = (int16_t *)dst; |
200 | const int16_t *smp_src = (const int16_t *)src; |
201 | for (i = 0; i < nb_samples; i++) |
202 | smp_dst[i] = av_clip_int16(((int64_t)smp_src[i] * volume + 128) >> 8); |
203 | } |
204 | |
205 | static inline void scale_samples_s16_small(uint8_t *dst, const uint8_t *src, |
206 | int nb_samples, int volume) |
207 | { |
208 | int i; |
209 | int16_t *smp_dst = (int16_t *)dst; |
210 | const int16_t *smp_src = (const int16_t *)src; |
211 | for (i = 0; i < nb_samples; i++) |
212 | smp_dst[i] = av_clip_int16((smp_src[i] * volume + 128) >> 8); |
213 | } |
214 | |
215 | static inline void scale_samples_s32(uint8_t *dst, const uint8_t *src, |
216 | int nb_samples, int volume) |
217 | { |
218 | int i; |
219 | int32_t *smp_dst = (int32_t *)dst; |
220 | const int32_t *smp_src = (const int32_t *)src; |
221 | for (i = 0; i < nb_samples; i++) |
222 | smp_dst[i] = av_clipl_int32((((int64_t)smp_src[i] * volume + 128) >> 8)); |
223 | } |
224 | |
225 | static av_cold void volume_init(VolumeContext *vol) |
226 | { |
227 | vol->samples_align = 1; |
228 | |
229 | switch (av_get_packed_sample_fmt(vol->sample_fmt)) { |
230 | case AV_SAMPLE_FMT_U8: |
231 | if (vol->volume_i < 0x1000000) |
232 | vol->scale_samples = scale_samples_u8_small; |
233 | else |
234 | vol->scale_samples = scale_samples_u8; |
235 | break; |
236 | case AV_SAMPLE_FMT_S16: |
237 | if (vol->volume_i < 0x10000) |
238 | vol->scale_samples = scale_samples_s16_small; |
239 | else |
240 | vol->scale_samples = scale_samples_s16; |
241 | break; |
242 | case AV_SAMPLE_FMT_S32: |
243 | vol->scale_samples = scale_samples_s32; |
244 | break; |
245 | case AV_SAMPLE_FMT_FLT: |
246 | vol->samples_align = 4; |
247 | break; |
248 | case AV_SAMPLE_FMT_DBL: |
249 | vol->samples_align = 8; |
250 | break; |
251 | } |
252 | |
253 | if (ARCH_X86) |
254 | ff_volume_init_x86(vol); |
255 | } |
256 | |
257 | static int set_volume(AVFilterContext *ctx) |
258 | { |
259 | VolumeContext *vol = ctx->priv; |
260 | |
261 | vol->volume = av_expr_eval(vol->volume_pexpr, vol->var_values, NULL); |
262 | if (isnan(vol->volume)) { |
263 | if (vol->eval_mode == EVAL_MODE_ONCE) { |
264 | av_log(ctx, AV_LOG_ERROR, "Invalid value NaN for volume\n"); |
265 | return AVERROR(EINVAL); |
266 | } else { |
267 | av_log(ctx, AV_LOG_WARNING, "Invalid value NaN for volume, setting to 0\n"); |
268 | vol->volume = 0; |
269 | } |
270 | } |
271 | vol->var_values[VAR_VOLUME] = vol->volume; |
272 | |
273 | av_log(ctx, AV_LOG_VERBOSE, "n:%f t:%f pts:%f precision:%s ", |
274 | vol->var_values[VAR_N], vol->var_values[VAR_T], vol->var_values[VAR_PTS], |
275 | precision_str[vol->precision]); |
276 | |
277 | if (vol->precision == PRECISION_FIXED) { |
278 | vol->volume_i = (int)(vol->volume * 256 + 0.5); |
279 | vol->volume = vol->volume_i / 256.0; |
280 | av_log(ctx, AV_LOG_VERBOSE, "volume_i:%d/255 ", vol->volume_i); |
281 | } |
282 | av_log(ctx, AV_LOG_VERBOSE, "volume:%f volume_dB:%f\n", |
283 | vol->volume, 20.0*log10(vol->volume)); |
284 | |
285 | volume_init(vol); |
286 | return 0; |
287 | } |
288 | |
289 | static int config_output(AVFilterLink *outlink) |
290 | { |
291 | AVFilterContext *ctx = outlink->src; |
292 | VolumeContext *vol = ctx->priv; |
293 | AVFilterLink *inlink = ctx->inputs[0]; |
294 | |
295 | vol->sample_fmt = inlink->format; |
296 | vol->channels = inlink->channels; |
297 | vol->planes = av_sample_fmt_is_planar(inlink->format) ? vol->channels : 1; |
298 | |
299 | vol->var_values[VAR_N] = |
300 | vol->var_values[VAR_NB_CONSUMED_SAMPLES] = |
301 | vol->var_values[VAR_NB_SAMPLES] = |
302 | vol->var_values[VAR_POS] = |
303 | vol->var_values[VAR_PTS] = |
304 | vol->var_values[VAR_STARTPTS] = |
305 | vol->var_values[VAR_STARTT] = |
306 | vol->var_values[VAR_T] = |
307 | vol->var_values[VAR_VOLUME] = NAN; |
308 | |
309 | vol->var_values[VAR_NB_CHANNELS] = inlink->channels; |
310 | vol->var_values[VAR_TB] = av_q2d(inlink->time_base); |
311 | vol->var_values[VAR_SAMPLE_RATE] = inlink->sample_rate; |
312 | |
313 | av_log(inlink->src, AV_LOG_VERBOSE, "tb:%f sample_rate:%f nb_channels:%f\n", |
314 | vol->var_values[VAR_TB], |
315 | vol->var_values[VAR_SAMPLE_RATE], |
316 | vol->var_values[VAR_NB_CHANNELS]); |
317 | |
318 | return set_volume(ctx); |
319 | } |
320 | |
321 | static int process_command(AVFilterContext *ctx, const char *cmd, const char *args, |
322 | char *res, int res_len, int flags) |
323 | { |
324 | VolumeContext *vol = ctx->priv; |
325 | int ret = AVERROR(ENOSYS); |
326 | |
327 | if (!strcmp(cmd, "volume")) { |
328 | if ((ret = set_expr(&vol->volume_pexpr, args, ctx)) < 0) |
329 | return ret; |
330 | if (vol->eval_mode == EVAL_MODE_ONCE) |
331 | set_volume(ctx); |
332 | } |
333 | |
334 | return ret; |
335 | } |
336 | |
337 | #define D2TS(d) (isnan(d) ? AV_NOPTS_VALUE : (int64_t)(d)) |
338 | #define TS2D(ts) ((ts) == AV_NOPTS_VALUE ? NAN : (double)(ts)) |
339 | #define TS2T(ts, tb) ((ts) == AV_NOPTS_VALUE ? NAN : (double)(ts)*av_q2d(tb)) |
340 | |
341 | static int filter_frame(AVFilterLink *inlink, AVFrame *buf) |
342 | { |
343 | AVFilterContext *ctx = inlink->dst; |
344 | VolumeContext *vol = inlink->dst->priv; |
345 | AVFilterLink *outlink = inlink->dst->outputs[0]; |
346 | int nb_samples = buf->nb_samples; |
347 | AVFrame *out_buf; |
348 | int64_t pos; |
349 | AVFrameSideData *sd = av_frame_get_side_data(buf, AV_FRAME_DATA_REPLAYGAIN); |
350 | int ret; |
351 | |
352 | if (sd && vol->replaygain != REPLAYGAIN_IGNORE) { |
353 | if (vol->replaygain != REPLAYGAIN_DROP) { |
354 | AVReplayGain *replaygain = (AVReplayGain*)sd->data; |
355 | int32_t gain = 100000; |
356 | uint32_t peak = 100000; |
357 | float g, p; |
358 | |
359 | if (vol->replaygain == REPLAYGAIN_TRACK && |
360 | replaygain->track_gain != INT32_MIN) { |
361 | gain = replaygain->track_gain; |
362 | |
363 | if (replaygain->track_peak != 0) |
364 | peak = replaygain->track_peak; |
365 | } else if (replaygain->album_gain != INT32_MIN) { |
366 | gain = replaygain->album_gain; |
367 | |
368 | if (replaygain->album_peak != 0) |
369 | peak = replaygain->album_peak; |
370 | } else { |
371 | av_log(inlink->dst, AV_LOG_WARNING, "Both ReplayGain gain " |
372 | "values are unknown.\n"); |
373 | } |
374 | g = gain / 100000.0f; |
375 | p = peak / 100000.0f; |
376 | |
377 | av_log(inlink->dst, AV_LOG_VERBOSE, |
378 | "Using gain %f dB from replaygain side data.\n", g); |
379 | |
380 | vol->volume = ff_exp10((g + vol->replaygain_preamp) / 20); |
381 | if (vol->replaygain_noclip) |
382 | vol->volume = FFMIN(vol->volume, 1.0 / p); |
383 | vol->volume_i = (int)(vol->volume * 256 + 0.5); |
384 | |
385 | volume_init(vol); |
386 | } |
387 | av_frame_remove_side_data(buf, AV_FRAME_DATA_REPLAYGAIN); |
388 | } |
389 | |
390 | if (isnan(vol->var_values[VAR_STARTPTS])) { |
391 | vol->var_values[VAR_STARTPTS] = TS2D(buf->pts); |
392 | vol->var_values[VAR_STARTT ] = TS2T(buf->pts, inlink->time_base); |
393 | } |
394 | vol->var_values[VAR_PTS] = TS2D(buf->pts); |
395 | vol->var_values[VAR_T ] = TS2T(buf->pts, inlink->time_base); |
396 | vol->var_values[VAR_N ] = inlink->frame_count_out; |
397 | |
398 | pos = av_frame_get_pkt_pos(buf); |
399 | vol->var_values[VAR_POS] = pos == -1 ? NAN : pos; |
400 | if (vol->eval_mode == EVAL_MODE_FRAME) |
401 | set_volume(ctx); |
402 | |
403 | if (vol->volume == 1.0 || vol->volume_i == 256) { |
404 | out_buf = buf; |
405 | goto end; |
406 | } |
407 | |
408 | /* do volume scaling in-place if input buffer is writable */ |
409 | if (av_frame_is_writable(buf) |
410 | && (vol->precision != PRECISION_FIXED || vol->volume_i > 0)) { |
411 | out_buf = buf; |
412 | } else { |
413 | out_buf = ff_get_audio_buffer(inlink, nb_samples); |
414 | if (!out_buf) |
415 | return AVERROR(ENOMEM); |
416 | ret = av_frame_copy_props(out_buf, buf); |
417 | if (ret < 0) { |
418 | av_frame_free(&out_buf); |
419 | av_frame_free(&buf); |
420 | return ret; |
421 | } |
422 | } |
423 | |
424 | if (vol->precision != PRECISION_FIXED || vol->volume_i > 0) { |
425 | int p, plane_samples; |
426 | |
427 | if (av_sample_fmt_is_planar(buf->format)) |
428 | plane_samples = FFALIGN(nb_samples, vol->samples_align); |
429 | else |
430 | plane_samples = FFALIGN(nb_samples * vol->channels, vol->samples_align); |
431 | |
432 | if (vol->precision == PRECISION_FIXED) { |
433 | for (p = 0; p < vol->planes; p++) { |
434 | vol->scale_samples(out_buf->extended_data[p], |
435 | buf->extended_data[p], plane_samples, |
436 | vol->volume_i); |
437 | } |
438 | } else if (av_get_packed_sample_fmt(vol->sample_fmt) == AV_SAMPLE_FMT_FLT) { |
439 | for (p = 0; p < vol->planes; p++) { |
440 | vol->fdsp->vector_fmul_scalar((float *)out_buf->extended_data[p], |
441 | (const float *)buf->extended_data[p], |
442 | vol->volume, plane_samples); |
443 | } |
444 | } else { |
445 | for (p = 0; p < vol->planes; p++) { |
446 | vol->fdsp->vector_dmul_scalar((double *)out_buf->extended_data[p], |
447 | (const double *)buf->extended_data[p], |
448 | vol->volume, plane_samples); |
449 | } |
450 | } |
451 | } |
452 | |
453 | emms_c(); |
454 | |
455 | if (buf != out_buf) |
456 | av_frame_free(&buf); |
457 | |
458 | end: |
459 | vol->var_values[VAR_NB_CONSUMED_SAMPLES] += out_buf->nb_samples; |
460 | return ff_filter_frame(outlink, out_buf); |
461 | } |
462 | |
463 | static const AVFilterPad avfilter_af_volume_inputs[] = { |
464 | { |
465 | .name = "default", |
466 | .type = AVMEDIA_TYPE_AUDIO, |
467 | .filter_frame = filter_frame, |
468 | }, |
469 | { NULL } |
470 | }; |
471 | |
472 | static const AVFilterPad avfilter_af_volume_outputs[] = { |
473 | { |
474 | .name = "default", |
475 | .type = AVMEDIA_TYPE_AUDIO, |
476 | .config_props = config_output, |
477 | }, |
478 | { NULL } |
479 | }; |
480 | |
481 | AVFilter ff_af_volume = { |
482 | .name = "volume", |
483 | .description = NULL_IF_CONFIG_SMALL("Change input volume."), |
484 | .query_formats = query_formats, |
485 | .priv_size = sizeof(VolumeContext), |
486 | .priv_class = &volume_class, |
487 | .init = init, |
488 | .uninit = uninit, |
489 | .inputs = avfilter_af_volume_inputs, |
490 | .outputs = avfilter_af_volume_outputs, |
491 | .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC, |
492 | .process_command = process_command, |
493 | }; |
494 |