af_asyncts.c 8.61 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
/*
 * This file is part of Libav.
 *
 * Libav is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * Libav is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with Libav; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include "libavresample/avresample.h"
#include "libavutil/audio_fifo.h"
21
#include "libavutil/common.h"
22 23 24 25 26 27
#include "libavutil/mathematics.h"
#include "libavutil/opt.h"
#include "libavutil/samplefmt.h"

#include "audio.h"
#include "avfilter.h"
28
#include "internal.h"
29 30 31 32 33 34 35 36 37 38 39 40

typedef struct ASyncContext {
    const AVClass *class;

    AVAudioResampleContext *avr;
    int64_t pts;            ///< timestamp in samples of the first sample in fifo
    int min_delta;          ///< pad/trim min threshold in samples

    /* options */
    int resample;
    float min_delta_sec;
    int max_comp;
41 42 43

    /* set by filter_samples() to signal an output frame to request_frame() */
    int got_output;
44 45 46 47 48
} ASyncContext;

#define OFFSET(x) offsetof(ASyncContext, x)
#define A AV_OPT_FLAG_AUDIO_PARAM
static const AVOption options[] = {
49
    { "compensate", "Stretch/squeeze the data to make it match the timestamps", OFFSET(resample),      AV_OPT_TYPE_INT,   { .i64 = 0 },   0, 1,       A },
50
    { "min_delta",  "Minimum difference between timestamps and audio data "
51
                    "(in seconds) to trigger padding/trimmin the data.",        OFFSET(min_delta_sec), AV_OPT_TYPE_FLOAT, { .dbl = 0.1 }, 0, INT_MAX, A },
52
    { "max_comp",   "Maximum compensation in samples per second.",              OFFSET(max_comp),      AV_OPT_TYPE_INT,   { .i64 = 500 }, 0, INT_MAX, A },
53
    { "first_pts",  "Assume the first pts should be this value.",               OFFSET(pts),           AV_OPT_TYPE_INT64, { .i64 = AV_NOPTS_VALUE }, INT64_MIN, INT64_MAX, A },
54 55 56 57 58 59 60 61 62 63
    { NULL },
};

static const AVClass async_class = {
    .class_name = "asyncts filter",
    .item_name  = av_default_item_name,
    .option     = options,
    .version    = LIBAVUTIL_VERSION_INT,
};

64
static int init(AVFilterContext *ctx, const char *args)
65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122
{
    ASyncContext *s = ctx->priv;
    int ret;

    s->class = &async_class;
    av_opt_set_defaults(s);

    if ((ret = av_set_options_string(s, args, "=", ":")) < 0) {
        av_log(ctx, AV_LOG_ERROR, "Error parsing options string '%s'.\n", args);
        return ret;
    }
    av_opt_free(s);

    return 0;
}

static void uninit(AVFilterContext *ctx)
{
    ASyncContext *s = ctx->priv;

    if (s->avr) {
        avresample_close(s->avr);
        avresample_free(&s->avr);
    }
}

static int config_props(AVFilterLink *link)
{
    ASyncContext *s = link->src->priv;
    int ret;

    s->min_delta = s->min_delta_sec * link->sample_rate;
    link->time_base = (AVRational){1, link->sample_rate};

    s->avr = avresample_alloc_context();
    if (!s->avr)
        return AVERROR(ENOMEM);

    av_opt_set_int(s->avr,  "in_channel_layout", link->channel_layout, 0);
    av_opt_set_int(s->avr, "out_channel_layout", link->channel_layout, 0);
    av_opt_set_int(s->avr,  "in_sample_fmt",     link->format,         0);
    av_opt_set_int(s->avr, "out_sample_fmt",     link->format,         0);
    av_opt_set_int(s->avr,  "in_sample_rate",    link->sample_rate,    0);
    av_opt_set_int(s->avr, "out_sample_rate",    link->sample_rate,    0);

    if (s->resample)
        av_opt_set_int(s->avr, "force_resampling", 1, 0);

    if ((ret = avresample_open(s->avr)) < 0)
        return ret;

    return 0;
}

static int request_frame(AVFilterLink *link)
{
    AVFilterContext *ctx = link->src;
    ASyncContext      *s = ctx->priv;
123
    int ret = 0;
124 125
    int nb_samples;

126 127 128 129
    s->got_output = 0;
    while (ret >= 0 && !s->got_output)
        ret = ff_request_frame(ctx->inputs[0]);

130 131 132 133 134 135
    /* flush the fifo */
    if (ret == AVERROR_EOF && (nb_samples = avresample_get_delay(s->avr))) {
        AVFilterBufferRef *buf = ff_get_audio_buffer(link, AV_PERM_WRITE,
                                                     nb_samples);
        if (!buf)
            return AVERROR(ENOMEM);
136
        ret = avresample_convert(s->avr, buf->extended_data,
137 138 139 140 141 142
                                 buf->linesize[0], nb_samples, NULL, 0, 0);
        if (ret <= 0) {
            avfilter_unref_bufferp(&buf);
            return (ret < 0) ? ret : AVERROR_EOF;
        }

143
        buf->pts = s->pts;
144
        return ff_filter_samples(link, buf);
145 146 147 148 149
    }

    return ret;
}

150
static int write_to_fifo(ASyncContext *s, AVFilterBufferRef *buf)
151
{
152
    int ret = avresample_convert(s->avr, NULL, 0, 0, buf->extended_data,
153
                                 buf->linesize[0], buf->audio->nb_samples);
154
    avfilter_unref_buffer(buf);
155
    return ret;
156 157 158 159 160 161 162 163
}

/* get amount of data currently buffered, in samples */
static int64_t get_delay(ASyncContext *s)
{
    return avresample_available(s->avr) + avresample_get_delay(s->avr);
}

164
static int filter_samples(AVFilterLink *inlink, AVFilterBufferRef *buf)
165 166 167 168 169 170 171
{
    AVFilterContext  *ctx = inlink->dst;
    ASyncContext       *s = ctx->priv;
    AVFilterLink *outlink = ctx->outputs[0];
    int nb_channels = av_get_channel_layout_nb_channels(buf->audio->channel_layout);
    int64_t pts = (buf->pts == AV_NOPTS_VALUE) ? buf->pts :
                  av_rescale_q(buf->pts, inlink->time_base, outlink->time_base);
172
    int out_size, ret;
173 174 175 176 177 178 179
    int64_t delta;

    /* buffer data until we get the first timestamp */
    if (s->pts == AV_NOPTS_VALUE) {
        if (pts != AV_NOPTS_VALUE) {
            s->pts = pts - get_delay(s);
        }
180
        return write_to_fifo(s, buf);
181 182 183 184
    }

    /* now wait for the next timestamp */
    if (pts == AV_NOPTS_VALUE) {
185
        return write_to_fifo(s, buf);
186 187 188 189 190 191 192 193 194
    }

    /* when we have two timestamps, compute how many samples would we have
     * to add/remove to get proper sync between data and timestamps */
    delta    = pts - s->pts - get_delay(s);
    out_size = avresample_available(s->avr);

    if (labs(delta) > s->min_delta) {
        av_log(ctx, AV_LOG_VERBOSE, "Discontinuity - %"PRId64" samples.\n", delta);
195
        out_size = av_clipl_int32((int64_t)out_size + delta);
196 197 198 199 200 201 202
    } else {
        if (s->resample) {
            int comp = av_clip(delta, -s->max_comp, s->max_comp);
            av_log(ctx, AV_LOG_VERBOSE, "Compensating %d samples per second.\n", comp);
            avresample_set_compensation(s->avr, delta, inlink->sample_rate);
        }
        delta = 0;
203 204 205 206 207
    }

    if (out_size > 0) {
        AVFilterBufferRef *buf_out = ff_get_audio_buffer(outlink, AV_PERM_WRITE,
                                                         out_size);
208 209 210 211
        if (!buf_out) {
            ret = AVERROR(ENOMEM);
            goto fail;
        }
212

213
        avresample_read(s->avr, buf_out->extended_data, out_size);
214 215 216 217 218 219
        buf_out->pts = s->pts;

        if (delta > 0) {
            av_samples_set_silence(buf_out->extended_data, out_size - delta,
                                   delta, nb_channels, buf->format);
        }
220 221 222
        ret = ff_filter_samples(outlink, buf_out);
        if (ret < 0)
            goto fail;
223
        s->got_output = 1;
224 225 226 227 228 229 230 231 232
    } else {
        av_log(ctx, AV_LOG_WARNING, "Non-monotonous timestamps, dropping "
               "whole buffer.\n");
    }

    /* drain any remaining buffered data */
    avresample_read(s->avr, NULL, avresample_available(s->avr));

    s->pts = pts - avresample_get_delay(s->avr);
233
    ret = avresample_convert(s->avr, NULL, 0, 0, buf->extended_data,
234 235 236
                             buf->linesize[0], buf->audio->nb_samples);

fail:
237
    avfilter_unref_buffer(buf);
238 239

    return ret;
240 241
}

242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
static const AVFilterPad avfilter_af_asyncts_inputs[] = {
    {
        .name           = "default",
        .type           = AVMEDIA_TYPE_AUDIO,
        .filter_samples = filter_samples
    },
    { NULL }
};

static const AVFilterPad avfilter_af_asyncts_outputs[] = {
    {
        .name          = "default",
        .type          = AVMEDIA_TYPE_AUDIO,
        .config_props  = config_props,
        .request_frame = request_frame
    },
    { NULL }
};

261 262 263 264 265 266 267 268 269
AVFilter avfilter_af_asyncts = {
    .name        = "asyncts",
    .description = NULL_IF_CONFIG_SMALL("Sync audio data to timestamps"),

    .init        = init,
    .uninit      = uninit,

    .priv_size   = sizeof(ASyncContext),

270 271
    .inputs      = avfilter_af_asyncts_inputs,
    .outputs     = avfilter_af_asyncts_outputs,
272
};