@@ -593,6 +593,7 @@ OBJS-$(CONFIG_CENSOR_FILTER) += sf_textmod.o
OBJS-$(CONFIG_SHOW_SPEAKER_FILTER) += sf_textmod.o
OBJS-$(CONFIG_SPLITCC_FILTER) += sf_splitcc.o
OBJS-$(CONFIG_STRIPSTYLES_FILTER) += sf_stripstyles.o
+OBJS-$(CONFIG_SUBFEED_FILTER) += sf_subfeed.o
OBJS-$(CONFIG_SUBSCALE_FILTER) += sf_subscale.o
OBJS-$(CONFIG_TEXTMOD_FILTER) += sf_textmod.o
@@ -581,6 +581,7 @@ extern const AVFilter ff_sf_graphicsub2text;
extern const AVFilter ff_sf_showspeaker;
extern const AVFilter ff_sf_splitcc;
extern const AVFilter ff_sf_stripstyles;
+extern const AVFilter ff_sf_subfeed;
extern const AVFilter ff_sf_subscale;
extern const AVFilter ff_sf_textmod;
new file mode 100644
@@ -0,0 +1,412 @@
+/*
+ * Copyright (c) 2021 softworkz
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * subtitle filter for feeding subtitle frames into a filtergraph in a contiguous way
+ *
+ *
+ * also supports
+ * - duration fixup
+ * delaying a subtitle event with unknown duration and infer duration from the
+ * start time of the subsequent subtitle
+ * - scattering
+ * splitting a subtitle event with unknown duration into multiple ones with
+ * a short and fixed duration
+ *
+ */
+
+#include "filters.h"
+#include "libavutil/opt.h"
+#include "subtitles.h"
+#include "libavutil/avassert.h"
+
+enum SubFeedMode {
+ FM_REPEAT,
+ FM_SCATTER,
+ FM_FORWARD,
+};
+
+typedef struct SubFeedContext {
+ const AVClass *class;
+ enum AVSubtitleType format;
+ enum SubFeedMode mode;
+
+ AVRational frame_rate;
+ int fix_durations;
+ int fix_overlap;
+
+ int current_frame_isnew;
+ int eof;
+ int got_first_input;
+ int need_frame;
+ int64_t next_pts_offset;
+ int64_t recent_subtitle_pts;
+
+ int64_t counter;
+
+ /**
+ * Queue of frames waiting to be filtered.
+ */
+ FFFrameQueue fifo;
+
+} SubFeedContext;
+
+static int64_t ms_to_avtb(int64_t ms)
+{
+ return av_rescale_q(ms, (AVRational){ 1, 1000 }, AV_TIME_BASE_Q);
+}
+
+static int64_t avtb_to_ms(int64_t avtb)
+{
+ return av_rescale_q(avtb, AV_TIME_BASE_Q, (AVRational){ 1, 1000 });
+}
+
+static int init(AVFilterContext *ctx)
+{
+ SubFeedContext *s = ctx->priv;
+
+ ff_framequeue_init(&s->fifo, NULL);
+
+ return 0;
+}
+
+static void uninit(AVFilterContext *ctx)
+{
+ SubFeedContext *s = ctx->priv;
+ ff_framequeue_free(&s->fifo);
+}
+
+static int config_input(AVFilterLink *link)
+{
+ ////const subfeedContext *context = link->dst->priv;
+
+ return 0;
+}
+
+static int query_formats(AVFilterContext *ctx)
+{
+ AVFilterFormats *formats;
+ AVFilterLink *inlink0 = ctx->inputs[0];
+ AVFilterLink *outlink0 = ctx->outputs[0];
+ static const enum AVSubtitleType subtitle_fmts[] = { AV_SUBTITLE_FMT_BITMAP, AV_SUBTITLE_FMT_ASS, AV_SUBTITLE_FMT_NB };
+ int ret;
+
+ formats = ff_make_format_list(subtitle_fmts);
+
+ if ((ret = ff_formats_ref(formats, &inlink0->outcfg.formats)) < 0)
+ return ret;
+
+ if ((ret = ff_formats_ref(formats, &outlink0->incfg.formats)) < 0)
+ return ret;
+
+ return 0;
+}
+
+static int config_output(AVFilterLink *outlink)
+{
+ SubFeedContext *s = outlink->src->priv;
+ const AVFilterLink *inlink = outlink->src->inputs[0];
+
+ outlink->time_base = AV_TIME_BASE_Q;
+ outlink->format = inlink->format;
+ outlink->w = inlink->w;
+ outlink->h = inlink->h;
+
+ if (s->mode == FM_FORWARD)
+ outlink->frame_rate = (AVRational) { 1, 0 };
+ else
+ outlink->frame_rate = s->frame_rate;
+
+ return 0;
+}
+
+static int request_frame(AVFilterLink *outlink)
+{
+ SubFeedContext *s = outlink->src->priv;
+ AVFilterLink *inlink = outlink->src->inputs[0];
+ int64_t last_pts = outlink->current_pts;
+ int64_t next_pts;
+ int64_t interval = ms_to_avtb((int64_t)(av_q2d(av_inv_q(outlink->frame_rate)) * 1000));
+ AVFrame *out;
+ int status;
+
+ if (s->mode == FM_FORWARD)
+ return ff_request_frame(inlink);
+
+ s->counter++;
+ if (interval == 0)
+ interval = ms_to_avtb(200);
+
+ status = ff_outlink_get_status(inlink);
+ if (status == AVERROR_EOF)
+ s->eof = 1;
+
+ if (s->eof)
+ return AVERROR_EOF;
+
+ if (!s->got_first_input && inlink->current_pts != AV_NOPTS_VALUE) {
+
+ s->got_first_input = 1;
+ next_pts = av_rescale_q(inlink->current_pts, inlink->time_base, AV_TIME_BASE_Q);
+ if (next_pts < last_pts)
+ next_pts = last_pts + interval;
+
+ } else if (last_pts == AV_NOPTS_VALUE)
+ next_pts = av_rescale_q(inlink->current_pts, inlink->time_base, AV_TIME_BASE_Q);
+ else
+ next_pts = last_pts + interval;
+
+ if (next_pts == AV_NOPTS_VALUE)
+ next_pts = 0;
+
+ if (s->next_pts_offset) {
+ av_log(outlink->src, AV_LOG_VERBOSE, "Subtracting next_pts_offset: %"PRId64" \n", s->next_pts_offset);
+ next_pts -= s->next_pts_offset;
+ s->next_pts_offset = 0;
+ }
+
+retry:
+ if (ff_framequeue_queued_frames(&s->fifo) && !s->current_frame_isnew) {
+
+ const AVFrame *current_frame = ff_framequeue_peek(&s->fifo, 0);
+ const int64_t sub_end_time = current_frame->subtitle_timing.start_pts + current_frame->subtitle_timing.duration;
+
+ if (ff_framequeue_queued_frames(&s->fifo) > 1) {
+ const AVFrame *next_frame = ff_framequeue_peek(&s->fifo, 1);
+ if (next_pts + interval > next_frame->subtitle_timing.start_pts) {
+ AVFrame *remove_frame = ff_framequeue_take(&s->fifo);
+ av_frame_free(&remove_frame);
+ s->current_frame_isnew = 1;
+ goto retry;
+ }
+ }
+
+ if (next_pts > sub_end_time) {
+ AVFrame *remove_frame = ff_framequeue_take(&s->fifo);
+ av_frame_free(&remove_frame);
+ s->current_frame_isnew = 1;
+ goto retry;
+ }
+ }
+
+ if (ff_framequeue_queued_frames(&s->fifo)) {
+ AVFrame *current_frame = ff_framequeue_peek(&s->fifo, 0);
+
+ if (current_frame && current_frame->subtitle_timing.start_pts <= next_pts + interval) {
+ if (!s->current_frame_isnew)
+ current_frame->repeat_sub++;
+
+ out = av_frame_clone(current_frame);
+
+ if (!out)
+ return AVERROR(ENOMEM);
+
+ if (!s->current_frame_isnew) {
+ out->pts = next_pts;
+ } else {
+ out->pts = out->subtitle_timing.start_pts;
+
+ if (out->pts < next_pts)
+ out->pts = next_pts;
+
+ s->next_pts_offset = (out->pts - next_pts) % interval;
+ }
+
+ if (s->mode == FM_SCATTER) {
+ const int64_t sub_end_time = current_frame->subtitle_timing.start_pts + current_frame->subtitle_timing.duration;
+
+ if (s->current_frame_isnew == 1 && current_frame->subtitle_timing.start_pts < out->pts) {
+ const int64_t diff = out->pts - current_frame->subtitle_timing.start_pts;
+ current_frame->subtitle_timing.duration -= diff;
+ }
+
+ out->repeat_sub = 0;
+ out->subtitle_timing.start_pts = out->pts;
+ out->subtitle_timing.duration = interval;
+ av_assert1(out->pts >= next_pts);
+ av_assert1(out->pts < next_pts + interval);
+ av_assert1(out->pts < sub_end_time);
+
+ if (out->pts > next_pts)
+ out->subtitle_timing.duration -= out->pts - next_pts;
+
+ if (sub_end_time < next_pts + interval) {
+ const int64_t diff = next_pts + interval - sub_end_time;
+ av_assert1(diff <= out->subtitle_timing.duration);
+ out->subtitle_timing.duration -= diff;
+ }
+ }
+
+ s->current_frame_isnew = 0;
+ s->recent_subtitle_pts = out->subtitle_timing.start_pts;
+
+ av_log(outlink->src, AV_LOG_DEBUG, "Output1 frame pts: %"PRId64" subtitle_pts: %"PRId64" repeat_frame: %d\n",
+ out->pts, out->subtitle_timing.start_pts, out->repeat_sub);
+
+ return ff_filter_frame(outlink, out);
+ }
+ }
+
+ if (ff_framequeue_queued_frames(&s->fifo) == 0) {
+ status = ff_request_frame(inlink);
+ if (status == AVERROR_EOF) {
+ s->eof = 1;
+ return status;
+ }
+
+ if (s->counter > 1 && s->counter % 2)
+ return 0;
+ }
+
+ out = ff_get_subtitles_buffer(outlink, outlink->format);
+ out->pts = next_pts;
+ out->repeat_sub = 1;
+ out->subtitle_timing.start_pts = s->recent_subtitle_pts;
+
+ av_log(outlink->src, AV_LOG_DEBUG, "Output2 frame pts: %"PRId64" subtitle_pts: %"PRId64" repeat_frame: %d\n",
+ out->pts, out->subtitle_timing.start_pts, out->repeat_sub);
+
+ return ff_filter_frame(outlink, out);
+}
+
+static int filter_frame(AVFilterLink *inlink, AVFrame *frame)
+{
+ AVFilterContext *ctx = inlink->dst;
+ SubFeedContext *s = inlink->dst->priv;
+ AVFilterLink *outlink = inlink->dst->outputs[0];
+ const int64_t index = (int64_t)ff_framequeue_queued_frames(&s->fifo) - 1;
+ size_t nb_queued_frames;
+ int ret = 0;
+
+ av_log(ctx, AV_LOG_VERBOSE, "frame.pts: %"PRId64" (AVTB: %"PRId64") - subtitle_timing.start_pts: %"PRId64" subtitle_timing.duration: %"PRId64" - format: %d\n",
+ frame->pts, av_rescale_q(frame->pts, inlink->time_base, AV_TIME_BASE_Q), frame->subtitle_timing.start_pts, frame->subtitle_timing.duration, frame->format);
+
+ frame->pts = av_rescale_q(frame->pts, inlink->time_base, AV_TIME_BASE_Q);
+
+ if (index < 0) {
+ s->current_frame_isnew = 1;
+ } else if (s->fix_durations || s->fix_overlap) {
+ AVFrame *previous_frame = ff_framequeue_peek(&s->fifo, index);
+ const int64_t pts_diff = frame->subtitle_timing.start_pts - previous_frame->subtitle_timing.start_pts;
+ nb_queued_frames = ff_framequeue_queued_frames(&s->fifo);
+
+ if (s->fix_durations && pts_diff > 0 && previous_frame->subtitle_timing.duration > ms_to_avtb(29000)) {
+ av_log(ctx, AV_LOG_VERBOSE, "Previous frame (index #%"PRId64") has a duration of %"PRId64" ms, setting to %"PRId64" ms\n",
+ index, avtb_to_ms(previous_frame->subtitle_timing.duration), avtb_to_ms(frame->subtitle_timing.start_pts - previous_frame->subtitle_timing.start_pts));
+ previous_frame->subtitle_timing.duration = frame->subtitle_timing.start_pts - previous_frame->subtitle_timing.start_pts;
+ }
+
+ if (s->fix_overlap && pts_diff > 0 && previous_frame->subtitle_timing.duration > pts_diff) {
+ av_log(ctx, AV_LOG_VERBOSE, "Detected overlap from previous frame (index #%"PRId64") which had a duration of %"PRId64" ms, setting to the pts_diff which is %"PRId64" ms\n",
+ index, avtb_to_ms(previous_frame->subtitle_timing.duration), avtb_to_ms(pts_diff));
+ previous_frame->subtitle_timing.duration = pts_diff;
+ }
+
+ if (pts_diff <= 0) {
+ av_log(ctx, AV_LOG_WARNING, "The pts_diff to the previous frame (index #%"PRId64") is <= 0: %"PRId64" ms. The previous frame duration is %"PRId64" ms.\n",
+ index, avtb_to_ms(pts_diff), avtb_to_ms(previous_frame->subtitle_timing.duration));
+
+ if (s->fix_overlap) {
+ av_log(ctx, AV_LOG_VERBOSE, "Removing previous frame\n");
+ previous_frame = ff_framequeue_take(&s->fifo);
+ while (nb_queued_frames > 1) {
+ ff_framequeue_add(&s->fifo, previous_frame);
+ previous_frame = ff_framequeue_take(&s->fifo);
+ nb_queued_frames--;
+ }
+ }
+ }
+ }
+
+ ff_framequeue_add(&s->fifo, frame);
+
+ nb_queued_frames = ff_framequeue_queued_frames(&s->fifo);
+
+ if (nb_queued_frames > 3)
+ av_log(ctx, AV_LOG_WARNING, "frame queue count: %zu\n", nb_queued_frames);
+
+ if (s->mode == FM_FORWARD && nb_queued_frames) {
+
+ AVFrame *first_frame = ff_framequeue_peek(&s->fifo, 0);
+
+ if (s->fix_overlap && nb_queued_frames < 2) {
+ av_log(ctx, AV_LOG_VERBOSE, "Return no frame since we have less than 2\n");
+ return 0;
+ }
+
+ if (s->fix_durations && first_frame->subtitle_timing.duration > ms_to_avtb(29000)) {
+ av_log(ctx, AV_LOG_VERBOSE, "Return no frame because first frame duration is %"PRId64" ms\n", avtb_to_ms(first_frame->subtitle_timing.duration));
+ return 0;
+ }
+
+ first_frame = ff_framequeue_take(&s->fifo);
+ return ff_filter_frame(outlink, first_frame);
+ }
+
+ return ret;
+}
+
+#define OFFSET(x) offsetof(SubFeedContext, x)
+#define FLAGS (AV_OPT_FLAG_SUBTITLE_PARAM | AV_OPT_FLAG_FILTERING_PARAM)
+
+static const AVOption subfeed_options[] = {
+ { "fix_durations", "delay output and determine duration from next frame", OFFSET(fix_durations), AV_OPT_TYPE_BOOL, { .i64=1 }, 0, 1, FLAGS, NULL },
+ { "fix_overlap", "delay output and adjust durations to prevent overlap", OFFSET(fix_overlap), AV_OPT_TYPE_BOOL, { .i64=0 }, 0, 1, FLAGS, NULL },
+ { "mode", "set feed mode", OFFSET(mode), AV_OPT_TYPE_INT, { .i64=FM_REPEAT }, FM_REPEAT, FM_FORWARD, FLAGS, "mode" },
+ { "repeat", "repeat recent while valid, send empty otherwise", 0, AV_OPT_TYPE_CONST, { .i64=FM_REPEAT }, 0, 0, FLAGS, "mode" },
+ { "scatter", "subdivide subtitles into 1/framerate segments", 0, AV_OPT_TYPE_CONST, { .i64=FM_SCATTER }, 0, 0, FLAGS, "mode" },
+ { "forward", "forward only (clears output framerate)", 0, AV_OPT_TYPE_CONST, { .i64=FM_FORWARD }, 0, 0, FLAGS, "mode" },
+ { "rate", "output frame rate", OFFSET(frame_rate), AV_OPT_TYPE_VIDEO_RATE, { .str = "5"}, 0, INT_MAX, FLAGS, NULL },\
+ { "r", "output frame rate", OFFSET(frame_rate), AV_OPT_TYPE_VIDEO_RATE, { .str = "5"}, 0, INT_MAX, FLAGS, NULL },\
+ { NULL },
+};
+
+AVFILTER_DEFINE_CLASS(subfeed);
+
+static const AVFilterPad inputs[] = {
+ {
+ .name = "default",
+ .type = AVMEDIA_TYPE_SUBTITLE,
+ .filter_frame = filter_frame,
+ .config_props = config_input,
+ },
+};
+
+static const AVFilterPad outputs[] = {
+ {
+ .name = "default",
+ .type = AVMEDIA_TYPE_SUBTITLE,
+ .request_frame = request_frame,
+ .config_props = config_output,
+ },
+};
+
+const AVFilter ff_sf_subfeed = {
+ .name = "subfeed",
+ .description = NULL_IF_CONFIG_SMALL("Control subtitle frame timing and flow in a filtergraph"),
+ .init = init,
+ .uninit = uninit,
+ .priv_size = sizeof(SubFeedContext),
+ .priv_class = &subfeed_class,
+ FILTER_INPUTS(inputs),
+ FILTER_OUTPUTS(outputs),
+ FILTER_QUERY_FUNC(query_formats),
+};