libx264: support aspect ratio switching
[libav.git] / libavfilter / af_asyncts.c
CommitLineData
9f26421b
AK
1/*
2 * This file is part of Libav.
3 *
4 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19#include "libavresample/avresample.h"
20#include "libavutil/audio_fifo.h"
21#include "libavutil/mathematics.h"
22#include "libavutil/opt.h"
23#include "libavutil/samplefmt.h"
24
25#include "audio.h"
26#include "avfilter.h"
803391f7 27#include "internal.h"
9f26421b
AK
28
29typedef struct ASyncContext {
30 const AVClass *class;
31
32 AVAudioResampleContext *avr;
33 int64_t pts; ///< timestamp in samples of the first sample in fifo
34 int min_delta; ///< pad/trim min threshold in samples
35
36 /* options */
37 int resample;
38 float min_delta_sec;
39 int max_comp;
40} ASyncContext;
41
42#define OFFSET(x) offsetof(ASyncContext, x)
43#define A AV_OPT_FLAG_AUDIO_PARAM
44static const AVOption options[] = {
45 { "compensate", "Stretch/squeeze the data to make it match the timestamps", OFFSET(resample), AV_OPT_TYPE_INT, { 0 }, 0, 1, A },
46 { "min_delta", "Minimum difference between timestamps and audio data "
47 "(in seconds) to trigger padding/trimmin the data.", OFFSET(min_delta_sec), AV_OPT_TYPE_FLOAT, { 0.1 }, 0, INT_MAX, A },
48 { "max_comp", "Maximum compensation in samples per second.", OFFSET(max_comp), AV_OPT_TYPE_INT, { 500 }, 0, INT_MAX, A },
49 { NULL },
50};
51
52static const AVClass async_class = {
53 .class_name = "asyncts filter",
54 .item_name = av_default_item_name,
55 .option = options,
56 .version = LIBAVUTIL_VERSION_INT,
57};
58
a5e8c41c 59static int init(AVFilterContext *ctx, const char *args)
9f26421b
AK
60{
61 ASyncContext *s = ctx->priv;
62 int ret;
63
64 s->class = &async_class;
65 av_opt_set_defaults(s);
66
67 if ((ret = av_set_options_string(s, args, "=", ":")) < 0) {
68 av_log(ctx, AV_LOG_ERROR, "Error parsing options string '%s'.\n", args);
69 return ret;
70 }
71 av_opt_free(s);
72
73 s->pts = AV_NOPTS_VALUE;
74
75 return 0;
76}
77
78static void uninit(AVFilterContext *ctx)
79{
80 ASyncContext *s = ctx->priv;
81
82 if (s->avr) {
83 avresample_close(s->avr);
84 avresample_free(&s->avr);
85 }
86}
87
88static int config_props(AVFilterLink *link)
89{
90 ASyncContext *s = link->src->priv;
91 int ret;
92
93 s->min_delta = s->min_delta_sec * link->sample_rate;
94 link->time_base = (AVRational){1, link->sample_rate};
95
96 s->avr = avresample_alloc_context();
97 if (!s->avr)
98 return AVERROR(ENOMEM);
99
100 av_opt_set_int(s->avr, "in_channel_layout", link->channel_layout, 0);
101 av_opt_set_int(s->avr, "out_channel_layout", link->channel_layout, 0);
102 av_opt_set_int(s->avr, "in_sample_fmt", link->format, 0);
103 av_opt_set_int(s->avr, "out_sample_fmt", link->format, 0);
104 av_opt_set_int(s->avr, "in_sample_rate", link->sample_rate, 0);
105 av_opt_set_int(s->avr, "out_sample_rate", link->sample_rate, 0);
106
107 if (s->resample)
108 av_opt_set_int(s->avr, "force_resampling", 1, 0);
109
110 if ((ret = avresample_open(s->avr)) < 0)
111 return ret;
112
113 return 0;
114}
115
116static int request_frame(AVFilterLink *link)
117{
118 AVFilterContext *ctx = link->src;
119 ASyncContext *s = ctx->priv;
803391f7 120 int ret = ff_request_frame(ctx->inputs[0]);
9f26421b
AK
121 int nb_samples;
122
123 /* flush the fifo */
124 if (ret == AVERROR_EOF && (nb_samples = avresample_get_delay(s->avr))) {
125 AVFilterBufferRef *buf = ff_get_audio_buffer(link, AV_PERM_WRITE,
126 nb_samples);
127 if (!buf)
128 return AVERROR(ENOMEM);
129 avresample_convert(s->avr, (void**)buf->extended_data, buf->linesize[0],
130 nb_samples, NULL, 0, 0);
131 buf->pts = s->pts;
132 ff_filter_samples(link, buf);
133 return 0;
134 }
135
136 return ret;
137}
138
139static void write_to_fifo(ASyncContext *s, AVFilterBufferRef *buf)
140{
141 avresample_convert(s->avr, NULL, 0, 0, (void**)buf->extended_data,
142 buf->linesize[0], buf->audio->nb_samples);
143 avfilter_unref_buffer(buf);
144}
145
146/* get amount of data currently buffered, in samples */
147static int64_t get_delay(ASyncContext *s)
148{
149 return avresample_available(s->avr) + avresample_get_delay(s->avr);
150}
151
152static void filter_samples(AVFilterLink *inlink, AVFilterBufferRef *buf)
153{
154 AVFilterContext *ctx = inlink->dst;
155 ASyncContext *s = ctx->priv;
156 AVFilterLink *outlink = ctx->outputs[0];
157 int nb_channels = av_get_channel_layout_nb_channels(buf->audio->channel_layout);
158 int64_t pts = (buf->pts == AV_NOPTS_VALUE) ? buf->pts :
159 av_rescale_q(buf->pts, inlink->time_base, outlink->time_base);
160 int out_size;
161 int64_t delta;
162
163 /* buffer data until we get the first timestamp */
164 if (s->pts == AV_NOPTS_VALUE) {
165 if (pts != AV_NOPTS_VALUE) {
166 s->pts = pts - get_delay(s);
167 }
168 write_to_fifo(s, buf);
169 return;
170 }
171
172 /* now wait for the next timestamp */
173 if (pts == AV_NOPTS_VALUE) {
174 write_to_fifo(s, buf);
175 return;
176 }
177
178 /* when we have two timestamps, compute how many samples would we have
179 * to add/remove to get proper sync between data and timestamps */
180 delta = pts - s->pts - get_delay(s);
181 out_size = avresample_available(s->avr);
182
183 if (labs(delta) > s->min_delta) {
184 av_log(ctx, AV_LOG_VERBOSE, "Discontinuity - %"PRId64" samples.\n", delta);
185 out_size += delta;
f297dd38
AK
186 } else {
187 if (s->resample) {
188 int comp = av_clip(delta, -s->max_comp, s->max_comp);
189 av_log(ctx, AV_LOG_VERBOSE, "Compensating %d samples per second.\n", comp);
190 avresample_set_compensation(s->avr, delta, inlink->sample_rate);
191 }
192 delta = 0;
9f26421b
AK
193 }
194
195 if (out_size > 0) {
196 AVFilterBufferRef *buf_out = ff_get_audio_buffer(outlink, AV_PERM_WRITE,
197 out_size);
198 if (!buf_out)
199 return;
200
201 avresample_read(s->avr, (void**)buf_out->extended_data, out_size);
202 buf_out->pts = s->pts;
203
204 if (delta > 0) {
205 av_samples_set_silence(buf_out->extended_data, out_size - delta,
206 delta, nb_channels, buf->format);
207 }
208 ff_filter_samples(outlink, buf_out);
209 } else {
210 av_log(ctx, AV_LOG_WARNING, "Non-monotonous timestamps, dropping "
211 "whole buffer.\n");
212 }
213
214 /* drain any remaining buffered data */
215 avresample_read(s->avr, NULL, avresample_available(s->avr));
216
217 s->pts = pts - avresample_get_delay(s->avr);
218 avresample_convert(s->avr, NULL, 0, 0, (void**)buf->extended_data,
219 buf->linesize[0], buf->audio->nb_samples);
220 avfilter_unref_buffer(buf);
221}
222
223AVFilter avfilter_af_asyncts = {
224 .name = "asyncts",
225 .description = NULL_IF_CONFIG_SMALL("Sync audio data to timestamps"),
226
227 .init = init,
228 .uninit = uninit,
229
230 .priv_size = sizeof(ASyncContext),
231
232 .inputs = (const AVFilterPad[]) {{ .name = "default",
233 .type = AVMEDIA_TYPE_AUDIO,
234 .filter_samples = filter_samples },
235 { NULL }},
236 .outputs = (const AVFilterPad[]) {{ .name = "default",
237 .type = AVMEDIA_TYPE_AUDIO,
238 .config_props = config_props,
239 .request_frame = request_frame },
240 { NULL }},
241};