1 /*
2 * This file is part of FFmpeg.
3 *
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
8 *
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
13 *
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19 /**
20 * @file
21 * sample format and channel layout conversion audio filter
22 */
23
24 #include "libavutil/avassert.h"
25 #include "libavutil/avstring.h"
26 #include "libavutil/common.h"
27 #include "libavutil/dict.h"
28 #include "libavutil/mathematics.h"
29 #include "libavutil/opt.h"
30
31 #include "libavresample/avresample.h"
32
33 #include "audio.h"
34 #include "avfilter.h"
35 #include "formats.h"
36 #include "internal.h"
37
38 typedef struct ResampleContext {
39 const AVClass *class;
40 AVAudioResampleContext *avr;
41 AVDictionary *options;
42
43 int resampling;
44 int64_t next_pts;
45 int64_t next_in_pts;
46
47 /* set by filter_frame() to signal an output frame to request_frame() */
48 int got_output;
49 } ResampleContext;
50
init(AVFilterContext * ctx,AVDictionary ** opts)51 static av_cold int init(AVFilterContext *ctx, AVDictionary **opts)
52 {
53 ResampleContext *s = ctx->priv;
54 const AVClass *avr_class = avresample_get_class();
55 AVDictionaryEntry *e = NULL;
56
57 while ((e = av_dict_get(*opts, "", e, AV_DICT_IGNORE_SUFFIX))) {
58 if (av_opt_find(&avr_class, e->key, NULL, 0,
59 AV_OPT_SEARCH_FAKE_OBJ | AV_OPT_SEARCH_CHILDREN))
60 av_dict_set(&s->options, e->key, e->value, 0);
61 }
62
63 e = NULL;
64 while ((e = av_dict_get(s->options, "", e, AV_DICT_IGNORE_SUFFIX)))
65 av_dict_set(opts, e->key, NULL, 0);
66
67 /* do not allow the user to override basic format options */
68 av_dict_set(&s->options, "in_channel_layout", NULL, 0);
69 av_dict_set(&s->options, "out_channel_layout", NULL, 0);
70 av_dict_set(&s->options, "in_sample_fmt", NULL, 0);
71 av_dict_set(&s->options, "out_sample_fmt", NULL, 0);
72 av_dict_set(&s->options, "in_sample_rate", NULL, 0);
73 av_dict_set(&s->options, "out_sample_rate", NULL, 0);
74
75 return 0;
76 }
77
uninit(AVFilterContext * ctx)78 static av_cold void uninit(AVFilterContext *ctx)
79 {
80 ResampleContext *s = ctx->priv;
81
82 if (s->avr) {
83 avresample_close(s->avr);
84 avresample_free(&s->avr);
85 }
86 av_dict_free(&s->options);
87 }
88
query_formats(AVFilterContext * ctx)89 static int query_formats(AVFilterContext *ctx)
90 {
91 AVFilterLink *inlink = ctx->inputs[0];
92 AVFilterLink *outlink = ctx->outputs[0];
93 AVFilterFormats *in_formats, *out_formats, *in_samplerates, *out_samplerates;
94 AVFilterChannelLayouts *in_layouts, *out_layouts;
95 int ret;
96
97 if (!(in_formats = ff_all_formats (AVMEDIA_TYPE_AUDIO)) ||
98 !(out_formats = ff_all_formats (AVMEDIA_TYPE_AUDIO)) ||
99 !(in_samplerates = ff_all_samplerates ( )) ||
100 !(out_samplerates = ff_all_samplerates ( )) ||
101 !(in_layouts = ff_all_channel_layouts ( )) ||
102 !(out_layouts = ff_all_channel_layouts ( )))
103 return AVERROR(ENOMEM);
104
105 if ((ret = ff_formats_ref (in_formats, &inlink->outcfg.formats )) < 0 ||
106 (ret = ff_formats_ref (out_formats, &outlink->incfg.formats )) < 0 ||
107 (ret = ff_formats_ref (in_samplerates, &inlink->outcfg.samplerates )) < 0 ||
108 (ret = ff_formats_ref (out_samplerates, &outlink->incfg.samplerates )) < 0 ||
109 (ret = ff_channel_layouts_ref (in_layouts, &inlink->outcfg.channel_layouts)) < 0 ||
110 (ret = ff_channel_layouts_ref (out_layouts, &outlink->incfg.channel_layouts)) < 0)
111 return ret;
112
113 return 0;
114 }
115
config_output(AVFilterLink * outlink)116 static int config_output(AVFilterLink *outlink)
117 {
118 AVFilterContext *ctx = outlink->src;
119 AVFilterLink *inlink = ctx->inputs[0];
120 ResampleContext *s = ctx->priv;
121 char buf1[64], buf2[64];
122 int ret;
123
124 int64_t resampling_forced;
125
126 if (s->avr) {
127 avresample_close(s->avr);
128 avresample_free(&s->avr);
129 }
130
131 if (inlink->channel_layout == outlink->channel_layout &&
132 inlink->sample_rate == outlink->sample_rate &&
133 (inlink->format == outlink->format ||
134 (av_get_channel_layout_nb_channels(inlink->channel_layout) == 1 &&
135 av_get_channel_layout_nb_channels(outlink->channel_layout) == 1 &&
136 av_get_planar_sample_fmt(inlink->format) ==
137 av_get_planar_sample_fmt(outlink->format))))
138 return 0;
139
140 if (!(s->avr = avresample_alloc_context()))
141 return AVERROR(ENOMEM);
142
143 if (s->options) {
144 int ret;
145 AVDictionaryEntry *e = NULL;
146 while ((e = av_dict_get(s->options, "", e, AV_DICT_IGNORE_SUFFIX)))
147 av_log(ctx, AV_LOG_VERBOSE, "lavr option: %s=%s\n", e->key, e->value);
148
149 ret = av_opt_set_dict(s->avr, &s->options);
150 if (ret < 0)
151 return ret;
152 }
153
154 av_opt_set_int(s->avr, "in_channel_layout", inlink ->channel_layout, 0);
155 av_opt_set_int(s->avr, "out_channel_layout", outlink->channel_layout, 0);
156 av_opt_set_int(s->avr, "in_sample_fmt", inlink ->format, 0);
157 av_opt_set_int(s->avr, "out_sample_fmt", outlink->format, 0);
158 av_opt_set_int(s->avr, "in_sample_rate", inlink ->sample_rate, 0);
159 av_opt_set_int(s->avr, "out_sample_rate", outlink->sample_rate, 0);
160
161 if ((ret = avresample_open(s->avr)) < 0)
162 return ret;
163
164 av_opt_get_int(s->avr, "force_resampling", 0, &resampling_forced);
165 s->resampling = resampling_forced || (inlink->sample_rate != outlink->sample_rate);
166
167 if (s->resampling) {
168 outlink->time_base = (AVRational){ 1, outlink->sample_rate };
169 s->next_pts = AV_NOPTS_VALUE;
170 s->next_in_pts = AV_NOPTS_VALUE;
171 } else
172 outlink->time_base = inlink->time_base;
173
174 av_get_channel_layout_string(buf1, sizeof(buf1),
175 -1, inlink ->channel_layout);
176 av_get_channel_layout_string(buf2, sizeof(buf2),
177 -1, outlink->channel_layout);
178 av_log(ctx, AV_LOG_VERBOSE,
179 "fmt:%s srate:%d cl:%s -> fmt:%s srate:%d cl:%s\n",
180 av_get_sample_fmt_name(inlink ->format), inlink ->sample_rate, buf1,
181 av_get_sample_fmt_name(outlink->format), outlink->sample_rate, buf2);
182
183 return 0;
184 }
185
request_frame(AVFilterLink * outlink)186 static int request_frame(AVFilterLink *outlink)
187 {
188 AVFilterContext *ctx = outlink->src;
189 ResampleContext *s = ctx->priv;
190 int ret = 0;
191
192 s->got_output = 0;
193 while (ret >= 0 && !s->got_output)
194 ret = ff_request_frame(ctx->inputs[0]);
195
196 /* flush the lavr delay buffer */
197 if (ret == AVERROR_EOF && s->avr) {
198 AVFrame *frame;
199 int nb_samples = avresample_get_out_samples(s->avr, 0);
200
201 if (!nb_samples)
202 return ret;
203
204 frame = ff_get_audio_buffer(outlink, nb_samples);
205 if (!frame)
206 return AVERROR(ENOMEM);
207
208 ret = avresample_convert(s->avr, frame->extended_data,
209 frame->linesize[0], nb_samples,
210 NULL, 0, 0);
211 if (ret <= 0) {
212 av_frame_free(&frame);
213 return (ret == 0) ? AVERROR_EOF : ret;
214 }
215
216 frame->nb_samples = ret;
217 frame->pts = s->next_pts;
218 return ff_filter_frame(outlink, frame);
219 }
220 return ret;
221 }
222
filter_frame(AVFilterLink * inlink,AVFrame * in)223 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
224 {
225 AVFilterContext *ctx = inlink->dst;
226 ResampleContext *s = ctx->priv;
227 AVFilterLink *outlink = ctx->outputs[0];
228 int ret;
229
230 if (s->avr) {
231 AVFrame *out;
232 int delay, nb_samples;
233
234 /* maximum possible samples lavr can output */
235 delay = avresample_get_delay(s->avr);
236 nb_samples = avresample_get_out_samples(s->avr, in->nb_samples);
237
238 out = ff_get_audio_buffer(outlink, nb_samples);
239 if (!out) {
240 ret = AVERROR(ENOMEM);
241 goto fail;
242 }
243
244 ret = avresample_convert(s->avr, out->extended_data, out->linesize[0],
245 nb_samples, in->extended_data, in->linesize[0],
246 in->nb_samples);
247 if (ret <= 0) {
248 av_frame_free(&out);
249 if (ret < 0)
250 goto fail;
251 }
252
253 av_assert0(!avresample_available(s->avr));
254
255 if (s->resampling && s->next_pts == AV_NOPTS_VALUE) {
256 if (in->pts == AV_NOPTS_VALUE) {
257 av_log(ctx, AV_LOG_WARNING, "First timestamp is missing, "
258 "assuming 0.\n");
259 s->next_pts = 0;
260 } else
261 s->next_pts = av_rescale_q(in->pts, inlink->time_base,
262 outlink->time_base);
263 }
264
265 if (ret > 0) {
266 out->nb_samples = ret;
267
268 ret = av_frame_copy_props(out, in);
269 if (ret < 0) {
270 av_frame_free(&out);
271 goto fail;
272 }
273
274 if (s->resampling) {
275 out->sample_rate = outlink->sample_rate;
276 /* Only convert in->pts if there is a discontinuous jump.
277 This ensures that out->pts tracks the number of samples actually
278 output by the resampler in the absence of such a jump.
279 Otherwise, the rounding in av_rescale_q() and av_rescale()
280 causes off-by-1 errors. */
281 if (in->pts != AV_NOPTS_VALUE && in->pts != s->next_in_pts) {
282 out->pts = av_rescale_q(in->pts, inlink->time_base,
283 outlink->time_base) -
284 av_rescale(delay, outlink->sample_rate,
285 inlink->sample_rate);
286 } else
287 out->pts = s->next_pts;
288
289 s->next_pts = out->pts + out->nb_samples;
290 s->next_in_pts = in->pts + in->nb_samples;
291 } else
292 out->pts = in->pts;
293
294 ret = ff_filter_frame(outlink, out);
295 s->got_output = 1;
296 }
297
298 fail:
299 av_frame_free(&in);
300 } else {
301 in->format = outlink->format;
302 ret = ff_filter_frame(outlink, in);
303 s->got_output = 1;
304 }
305
306 return ret;
307 }
308
309 #if FF_API_CHILD_CLASS_NEXT
resample_child_class_next(const AVClass * prev)310 static const AVClass *resample_child_class_next(const AVClass *prev)
311 {
312 return prev ? NULL : avresample_get_class();
313 }
314 #endif
315
resample_child_class_iterate(void ** iter)316 static const AVClass *resample_child_class_iterate(void **iter)
317 {
318 const AVClass *c = *iter ? NULL : avresample_get_class();
319 *iter = (void*)(uintptr_t)c;
320 return c;
321 }
322
resample_child_next(void * obj,void * prev)323 static void *resample_child_next(void *obj, void *prev)
324 {
325 ResampleContext *s = obj;
326 return prev ? NULL : s->avr;
327 }
328
329 static const AVClass resample_class = {
330 .class_name = "resample",
331 .item_name = av_default_item_name,
332 .version = LIBAVUTIL_VERSION_INT,
333 #if FF_API_CHILD_CLASS_NEXT
334 .child_class_next = resample_child_class_next,
335 #endif
336 .child_class_iterate = resample_child_class_iterate,
337 .child_next = resample_child_next,
338 };
339
340 static const AVFilterPad avfilter_af_resample_inputs[] = {
341 {
342 .name = "default",
343 .type = AVMEDIA_TYPE_AUDIO,
344 .filter_frame = filter_frame,
345 },
346 { NULL }
347 };
348
349 static const AVFilterPad avfilter_af_resample_outputs[] = {
350 {
351 .name = "default",
352 .type = AVMEDIA_TYPE_AUDIO,
353 .config_props = config_output,
354 .request_frame = request_frame
355 },
356 { NULL }
357 };
358
359 AVFilter ff_af_resample = {
360 .name = "resample",
361 .description = NULL_IF_CONFIG_SMALL("Audio resampling and conversion."),
362 .priv_size = sizeof(ResampleContext),
363 .priv_class = &resample_class,
364 .init_dict = init,
365 .uninit = uninit,
366 .query_formats = query_formats,
367 .inputs = avfilter_af_resample_inputs,
368 .outputs = avfilter_af_resample_outputs,
369 };
370