2 * This file is part of FFmpeg.
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.
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.
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
21 * sample format and channel layout conversion audio filter
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"
31 #include "libavresample/avresample.h"
38 typedef struct ResampleContext {
40 AVAudioResampleContext *avr;
41 AVDictionary *options;
46 /* set by filter_frame() to signal an output frame to request_frame() */
50 static av_cold int init(AVFilterContext *ctx, AVDictionary **opts)
52 ResampleContext *s = ctx->priv;
53 const AVClass *avr_class = avresample_get_class();
54 AVDictionaryEntry *e = NULL;
56 while ((e = av_dict_get(*opts, "", e, AV_DICT_IGNORE_SUFFIX))) {
57 if (av_opt_find(&avr_class, e->key, NULL, 0,
58 AV_OPT_SEARCH_FAKE_OBJ | AV_OPT_SEARCH_CHILDREN))
59 av_dict_set(&s->options, e->key, e->value, 0);
63 while ((e = av_dict_get(s->options, "", e, AV_DICT_IGNORE_SUFFIX)))
64 av_dict_set(opts, e->key, NULL, 0);
66 /* do not allow the user to override basic format options */
67 av_dict_set(&s->options, "in_channel_layout", NULL, 0);
68 av_dict_set(&s->options, "out_channel_layout", NULL, 0);
69 av_dict_set(&s->options, "in_sample_fmt", NULL, 0);
70 av_dict_set(&s->options, "out_sample_fmt", NULL, 0);
71 av_dict_set(&s->options, "in_sample_rate", NULL, 0);
72 av_dict_set(&s->options, "out_sample_rate", NULL, 0);
77 static av_cold void uninit(AVFilterContext *ctx)
79 ResampleContext *s = ctx->priv;
82 avresample_close(s->avr);
83 avresample_free(&s->avr);
85 av_dict_free(&s->options);
88 static int query_formats(AVFilterContext *ctx)
90 AVFilterLink *inlink = ctx->inputs[0];
91 AVFilterLink *outlink = ctx->outputs[0];
93 AVFilterFormats *in_formats = ff_all_formats(AVMEDIA_TYPE_AUDIO);
94 AVFilterFormats *out_formats = ff_all_formats(AVMEDIA_TYPE_AUDIO);
95 AVFilterFormats *in_samplerates = ff_all_samplerates();
96 AVFilterFormats *out_samplerates = ff_all_samplerates();
97 AVFilterChannelLayouts *in_layouts = ff_all_channel_layouts();
98 AVFilterChannelLayouts *out_layouts = ff_all_channel_layouts();
100 ff_formats_ref(in_formats, &inlink->out_formats);
101 ff_formats_ref(out_formats, &outlink->in_formats);
103 ff_formats_ref(in_samplerates, &inlink->out_samplerates);
104 ff_formats_ref(out_samplerates, &outlink->in_samplerates);
106 ff_channel_layouts_ref(in_layouts, &inlink->out_channel_layouts);
107 ff_channel_layouts_ref(out_layouts, &outlink->in_channel_layouts);
112 static int config_output(AVFilterLink *outlink)
114 AVFilterContext *ctx = outlink->src;
115 AVFilterLink *inlink = ctx->inputs[0];
116 ResampleContext *s = ctx->priv;
117 char buf1[64], buf2[64];
121 avresample_close(s->avr);
122 avresample_free(&s->avr);
125 if (inlink->channel_layout == outlink->channel_layout &&
126 inlink->sample_rate == outlink->sample_rate &&
127 (inlink->format == outlink->format ||
128 (av_get_channel_layout_nb_channels(inlink->channel_layout) == 1 &&
129 av_get_channel_layout_nb_channels(outlink->channel_layout) == 1 &&
130 av_get_planar_sample_fmt(inlink->format) ==
131 av_get_planar_sample_fmt(outlink->format))))
134 if (!(s->avr = avresample_alloc_context()))
135 return AVERROR(ENOMEM);
139 AVDictionaryEntry *e = NULL;
140 while ((e = av_dict_get(s->options, "", e, AV_DICT_IGNORE_SUFFIX)))
141 av_log(ctx, AV_LOG_VERBOSE, "lavr option: %s=%s\n", e->key, e->value);
143 ret = av_opt_set_dict(s->avr, &s->options);
148 av_opt_set_int(s->avr, "in_channel_layout", inlink ->channel_layout, 0);
149 av_opt_set_int(s->avr, "out_channel_layout", outlink->channel_layout, 0);
150 av_opt_set_int(s->avr, "in_sample_fmt", inlink ->format, 0);
151 av_opt_set_int(s->avr, "out_sample_fmt", outlink->format, 0);
152 av_opt_set_int(s->avr, "in_sample_rate", inlink ->sample_rate, 0);
153 av_opt_set_int(s->avr, "out_sample_rate", outlink->sample_rate, 0);
155 if ((ret = avresample_open(s->avr)) < 0)
158 outlink->time_base = (AVRational){ 1, outlink->sample_rate };
159 s->next_pts = AV_NOPTS_VALUE;
160 s->next_in_pts = AV_NOPTS_VALUE;
162 av_get_channel_layout_string(buf1, sizeof(buf1),
163 -1, inlink ->channel_layout);
164 av_get_channel_layout_string(buf2, sizeof(buf2),
165 -1, outlink->channel_layout);
166 av_log(ctx, AV_LOG_VERBOSE,
167 "fmt:%s srate:%d cl:%s -> fmt:%s srate:%d cl:%s\n",
168 av_get_sample_fmt_name(inlink ->format), inlink ->sample_rate, buf1,
169 av_get_sample_fmt_name(outlink->format), outlink->sample_rate, buf2);
174 static int request_frame(AVFilterLink *outlink)
176 AVFilterContext *ctx = outlink->src;
177 ResampleContext *s = ctx->priv;
181 while (ret >= 0 && !s->got_output)
182 ret = ff_request_frame(ctx->inputs[0]);
184 /* flush the lavr delay buffer */
185 if (ret == AVERROR_EOF && s->avr) {
187 int nb_samples = avresample_get_out_samples(s->avr, 0);
192 frame = ff_get_audio_buffer(outlink, nb_samples);
194 return AVERROR(ENOMEM);
196 ret = avresample_convert(s->avr, frame->extended_data,
197 frame->linesize[0], nb_samples,
200 av_frame_free(&frame);
201 return (ret == 0) ? AVERROR_EOF : ret;
204 frame->nb_samples = ret;
205 frame->pts = s->next_pts;
206 return ff_filter_frame(outlink, frame);
211 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
213 AVFilterContext *ctx = inlink->dst;
214 ResampleContext *s = ctx->priv;
215 AVFilterLink *outlink = ctx->outputs[0];
220 int delay, nb_samples;
222 /* maximum possible samples lavr can output */
223 delay = avresample_get_delay(s->avr);
224 nb_samples = avresample_get_out_samples(s->avr, in->nb_samples);
226 out = ff_get_audio_buffer(outlink, nb_samples);
228 ret = AVERROR(ENOMEM);
232 ret = avresample_convert(s->avr, out->extended_data, out->linesize[0],
233 nb_samples, in->extended_data, in->linesize[0],
241 av_assert0(!avresample_available(s->avr));
243 if (s->next_pts == AV_NOPTS_VALUE) {
244 if (in->pts == AV_NOPTS_VALUE) {
245 av_log(ctx, AV_LOG_WARNING, "First timestamp is missing, "
249 s->next_pts = av_rescale_q(in->pts, inlink->time_base,
254 out->nb_samples = ret;
256 ret = av_frame_copy_props(out, in);
262 out->sample_rate = outlink->sample_rate;
263 /* Only convert in->pts if there is a discontinuous jump.
264 This ensures that out->pts tracks the number of samples actually
265 output by the resampler in the absence of such a jump.
266 Otherwise, the rounding in av_rescale_q() and av_rescale()
267 causes off-by-1 errors. */
268 if (in->pts != AV_NOPTS_VALUE && in->pts != s->next_in_pts) {
269 out->pts = av_rescale_q(in->pts, inlink->time_base,
270 outlink->time_base) -
271 av_rescale(delay, outlink->sample_rate,
272 inlink->sample_rate);
274 out->pts = s->next_pts;
276 s->next_pts = out->pts + out->nb_samples;
277 s->next_in_pts = in->pts + in->nb_samples;
279 ret = ff_filter_frame(outlink, out);
286 in->format = outlink->format;
287 ret = ff_filter_frame(outlink, in);
294 static const AVClass *resample_child_class_next(const AVClass *prev)
296 return prev ? NULL : avresample_get_class();
299 static void *resample_child_next(void *obj, void *prev)
301 ResampleContext *s = obj;
302 return prev ? NULL : s->avr;
305 static const AVClass resample_class = {
306 .class_name = "resample",
307 .item_name = av_default_item_name,
308 .version = LIBAVUTIL_VERSION_INT,
309 .child_class_next = resample_child_class_next,
310 .child_next = resample_child_next,
313 static const AVFilterPad avfilter_af_resample_inputs[] = {
316 .type = AVMEDIA_TYPE_AUDIO,
317 .filter_frame = filter_frame,
322 static const AVFilterPad avfilter_af_resample_outputs[] = {
325 .type = AVMEDIA_TYPE_AUDIO,
326 .config_props = config_output,
327 .request_frame = request_frame
332 AVFilter ff_af_resample = {
334 .description = NULL_IF_CONFIG_SMALL("Audio resampling and conversion."),
335 .priv_size = sizeof(ResampleContext),
336 .priv_class = &resample_class,
339 .query_formats = query_formats,
340 .inputs = avfilter_af_resample_inputs,
341 .outputs = avfilter_af_resample_outputs,