git.sesse.net Git - ffmpeg/blob - libavfilter/af_resample.c

   1 /*
   2  *
   3  * This file is part of Libav.
   4  *
   5  * Libav is free software; you can redistribute it and/or
   6  * modify it under the terms of the GNU Lesser General Public
   7  * License as published by the Free Software Foundation; either
   8  * version 2.1 of the License, or (at your option) any later version.
   9  *
  10  * Libav is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * Lesser General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU Lesser General Public
  16  * License along with Libav; if not, write to the Free Software
  17  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  18  */
  19
  20 /**
  21  * @file
  22  * sample format and channel layout conversion audio filter
  23  */
  24
  25 #include "libavutil/avassert.h"
  26 #include "libavutil/avstring.h"
  27 #include "libavutil/common.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     AVAudioResampleContext *avr;
  40
  41     int64_t next_pts;
  42
  43     /* set by filter_samples() to signal an output frame to request_frame() */
  44     int got_output;
  45 } ResampleContext;
  46
  47 static av_cold void uninit(AVFilterContext *ctx)
  48 {
  49     ResampleContext *s = ctx->priv;
  50
  51     if (s->avr) {
  52         avresample_close(s->avr);
  53         avresample_free(&s->avr);
  54     }
  55 }
  56
  57 static int query_formats(AVFilterContext *ctx)
  58 {
  59     AVFilterLink *inlink  = ctx->inputs[0];
  60     AVFilterLink *outlink = ctx->outputs[0];
  61
  62     AVFilterFormats        *in_formats      = ff_all_formats(AVMEDIA_TYPE_AUDIO);
  63     AVFilterFormats        *out_formats     = ff_all_formats(AVMEDIA_TYPE_AUDIO);
  64     AVFilterFormats        *in_samplerates  = ff_all_samplerates();
  65     AVFilterFormats        *out_samplerates = ff_all_samplerates();
  66     AVFilterChannelLayouts *in_layouts      = ff_all_channel_layouts();
  67     AVFilterChannelLayouts *out_layouts     = ff_all_channel_layouts();
  68
  69     ff_formats_ref(in_formats,  &inlink->out_formats);
  70     ff_formats_ref(out_formats, &outlink->in_formats);
  71
  72     ff_formats_ref(in_samplerates,  &inlink->out_samplerates);
  73     ff_formats_ref(out_samplerates, &outlink->in_samplerates);
  74
  75     ff_channel_layouts_ref(in_layouts,  &inlink->out_channel_layouts);
  76     ff_channel_layouts_ref(out_layouts, &outlink->in_channel_layouts);
  77
  78     return 0;
  79 }
  80
  81 static int config_output(AVFilterLink *outlink)
  82 {
  83     AVFilterContext *ctx = outlink->src;
  84     AVFilterLink *inlink = ctx->inputs[0];
  85     ResampleContext   *s = ctx->priv;
  86     char buf1[64], buf2[64];
  87     int ret;
  88
  89     if (s->avr) {
  90         avresample_close(s->avr);
  91         avresample_free(&s->avr);
  92     }
  93
  94     if (inlink->channel_layout == outlink->channel_layout &&
  95         inlink->sample_rate    == outlink->sample_rate    &&
  96         inlink->format         == outlink->format)
  97         return 0;
  98
  99     if (!(s->avr = avresample_alloc_context()))
 100         return AVERROR(ENOMEM);
 101
 102     av_opt_set_int(s->avr,  "in_channel_layout", inlink ->channel_layout, 0);
 103     av_opt_set_int(s->avr, "out_channel_layout", outlink->channel_layout, 0);
 104     av_opt_set_int(s->avr,  "in_sample_fmt",     inlink ->format,         0);
 105     av_opt_set_int(s->avr, "out_sample_fmt",     outlink->format,         0);
 106     av_opt_set_int(s->avr,  "in_sample_rate",    inlink ->sample_rate,    0);
 107     av_opt_set_int(s->avr, "out_sample_rate",    outlink->sample_rate,    0);
 108
 109     if ((ret = avresample_open(s->avr)) < 0)
 110         return ret;
 111
 112     outlink->time_base = (AVRational){ 1, outlink->sample_rate };
 113     s->next_pts        = AV_NOPTS_VALUE;
 114
 115     av_get_channel_layout_string(buf1, sizeof(buf1),
 116                                  -1, inlink ->channel_layout);
 117     av_get_channel_layout_string(buf2, sizeof(buf2),
 118                                  -1, outlink->channel_layout);
 119     av_log(ctx, AV_LOG_VERBOSE,
 120            "fmt:%s srate:%d cl:%s -> fmt:%s srate:%d cl:%s\n",
 121            av_get_sample_fmt_name(inlink ->format), inlink ->sample_rate, buf1,
 122            av_get_sample_fmt_name(outlink->format), outlink->sample_rate, buf2);
 123
 124     return 0;
 125 }
 126
 127 static int request_frame(AVFilterLink *outlink)
 128 {
 129     AVFilterContext *ctx = outlink->src;
 130     ResampleContext   *s = ctx->priv;
 131     int ret = 0;
 132
 133     s->got_output = 0;
 134     while (ret >= 0 && !s->got_output)
 135         ret = ff_request_frame(ctx->inputs[0]);
 136
 137     /* flush the lavr delay buffer */
 138     if (ret == AVERROR_EOF && s->avr) {
 139         AVFilterBufferRef *buf;
 140         int nb_samples = av_rescale_rnd(avresample_get_delay(s->avr),
 141                                         outlink->sample_rate,
 142                                         ctx->inputs[0]->sample_rate,
 143                                         AV_ROUND_UP);
 144
 145         if (!nb_samples)
 146             return ret;
 147
 148         buf = ff_get_audio_buffer(outlink, AV_PERM_WRITE, nb_samples);
 149         if (!buf)
 150             return AVERROR(ENOMEM);
 151
 152         ret = avresample_convert(s->avr, (void**)buf->extended_data,
 153                                  buf->linesize[0], nb_samples,
 154                                  NULL, 0, 0);
 155         if (ret <= 0) {
 156             avfilter_unref_buffer(buf);
 157             return (ret == 0) ? AVERROR_EOF : ret;
 158         }
 159
 160         buf->pts = s->next_pts;
 161         return ff_filter_samples(outlink, buf);
 162     }
 163     return ret;
 164 }
 165
 166 static int filter_samples(AVFilterLink *inlink, AVFilterBufferRef *buf)
 167 {
 168     AVFilterContext  *ctx = inlink->dst;
 169     ResampleContext    *s = ctx->priv;
 170     AVFilterLink *outlink = ctx->outputs[0];
 171     int ret;
 172
 173     if (s->avr) {
 174         AVFilterBufferRef *buf_out;
 175         int delay, nb_samples;
 176
 177         /* maximum possible samples lavr can output */
 178         delay      = avresample_get_delay(s->avr);
 179         nb_samples = av_rescale_rnd(buf->audio->nb_samples + delay,
 180                                     outlink->sample_rate, inlink->sample_rate,
 181                                     AV_ROUND_UP);
 182
 183         buf_out = ff_get_audio_buffer(outlink, AV_PERM_WRITE, nb_samples);
 184         if (!buf_out) {
 185             ret = AVERROR(ENOMEM);
 186             goto fail;
 187         }
 188
 189         ret     = avresample_convert(s->avr, (void**)buf_out->extended_data,
 190                                      buf_out->linesize[0], nb_samples,
 191                                      (void**)buf->extended_data, buf->linesize[0],
 192                                      buf->audio->nb_samples);
 193         if (ret < 0) {
 194             avfilter_unref_buffer(buf_out);
 195             goto fail;
 196         }
 197
 198         av_assert0(!avresample_available(s->avr));
 199
 200         if (s->next_pts == AV_NOPTS_VALUE) {
 201             if (buf->pts == AV_NOPTS_VALUE) {
 202                 av_log(ctx, AV_LOG_WARNING, "First timestamp is missing, "
 203                        "assuming 0.\n");
 204                 s->next_pts = 0;
 205             } else
 206                 s->next_pts = av_rescale_q(buf->pts, inlink->time_base,
 207                                            outlink->time_base);
 208         }
 209
 210         if (ret > 0) {
 211             buf_out->audio->nb_samples = ret;
 212             if (buf->pts != AV_NOPTS_VALUE) {
 213                 buf_out->pts = av_rescale_q(buf->pts, inlink->time_base,
 214                                             outlink->time_base) -
 215                                av_rescale(delay, outlink->sample_rate,
 216                                           inlink->sample_rate);
 217             } else
 218                 buf_out->pts = s->next_pts;
 219
 220             s->next_pts = buf_out->pts + buf_out->audio->nb_samples;
 221
 222             ret = ff_filter_samples(outlink, buf_out);
 223             s->got_output = 1;
 224         }
 225
 226 fail:
 227         avfilter_unref_buffer(buf);
 228     } else {
 229         ret = ff_filter_samples(outlink, buf);
 230         s->got_output = 1;
 231     }
 232
 233     return ret;
 234 }
 235
 236 AVFilter avfilter_af_resample = {
 237     .name          = "resample",
 238     .description   = NULL_IF_CONFIG_SMALL("Audio resampling and conversion."),
 239     .priv_size     = sizeof(ResampleContext),
 240
 241     .uninit         = uninit,
 242     .query_formats  = query_formats,
 243
 244     .inputs    = (const AVFilterPad[]) {{ .name            = "default",
 245                                           .type            = AVMEDIA_TYPE_AUDIO,
 246                                           .filter_samples  = filter_samples,
 247                                           .min_perms       = AV_PERM_READ },
 248                                         { .name = NULL}},
 249     .outputs   = (const AVFilterPad[]) {{ .name          = "default",
 250                                           .type          = AVMEDIA_TYPE_AUDIO,
 251                                           .config_props  = config_output,
 252                                           .request_frame = request_frame },
 253                                         { .name = NULL}},
 254 };