[FFmpeg-devel] [PATCH] ffmpeg: insert bitmap subtitles as video in filters.
Stefano Sabatini
stefasab at gmail.com
Wed Aug 1 14:33:10 CEST 2012
On date Tuesday 2012-07-31 16:54:34 +0200, Nicolas George encoded:
> With this feature, it becomes possible to perform commonly
> requested tasks, such as hardcoding bitmap subtitles.
>
> This will be reverted once libavfilter has proper support
> for subtitles. All the changes have the string "sub2video"
> in them, it makes it easy to spot the parts.
>
> Signed-off-by: Nicolas George <nicolas.george at normalesup.org>
> ---
> Changelog | 1 +
> doc/ffmpeg.texi | 14 +++++
> ffmpeg.c | 165 +++++++++++++++++++++++++++++++++++++++++++++++++++++--
> 3 files changed, 176 insertions(+), 4 deletions(-)
>
>
> Added a comment, as requested by Clément, and rebased on top of current Git.
> Otherwise unchanged.
>
> Does anyone want to comment?
>
>
> diff --git a/Changelog b/Changelog
> index 79a8f4c..64676a7 100644
> --- a/Changelog
> +++ b/Changelog
> @@ -38,6 +38,7 @@ version next:
> - alphaextract and alphamerge filters
> - concat filter
> - flite filter
> +- bitmap subtitles in filters (experimental and temporary)
>
>
> version 0.11:
> diff --git a/doc/ffmpeg.texi b/doc/ffmpeg.texi
> index 904a505..6d71532 100644
> --- a/doc/ffmpeg.texi
> +++ b/doc/ffmpeg.texi
> @@ -989,6 +989,20 @@ ffmpeg -i video.mkv -i image.png -filter_complex 'overlay' out.mkv
> @end example
> @end table
>
> +As a special exception, you can use a bitmap subtitle stream as input: it
> +will be converted into a video with the same size as the largest video in
> +the file, or 720×576 if no video is present. Note that this is an
> +experimental and temporary solution. It will be removed once libavfilter has
> +proper support for subtitles.
> +
> +For example, to hardcode subtitles on top of a DVB-T recording stored in
> +MPEG-TS format, delaying the subtitles by 1 second:
> + at example
> +ffmpeg -i input.ts -filter_complex \
> + '[#0x2ef] setpts=PTS+1/TB [sub] ; [#0x2d0] [sub] overlay' \
> + -sn -map '#0x2dc' output.mkv
> + at end example
Maybe mention what the funny labels stand for.
> +
> @section Preset files
> A preset file contains a sequence of @var{option}=@var{value} pairs,
> one for each line, specifying a sequence of options which would be
> diff --git a/ffmpeg.c b/ffmpeg.c
> index c2ea5bd..206580b 100644
> --- a/ffmpeg.c
> +++ b/ffmpeg.c
> @@ -249,6 +249,12 @@ typedef struct InputStream {
> int resample_channels;
> uint64_t resample_channel_layout;
>
> + struct sub2video {
> + int64_t last_pts;
> + AVFilterBufferRef *ref;
> + int w, h;
> + } sub2video;
> +
> /* a pool of free buffers for decoded data */
> FrameBuffer *buffer_pool;
> int dr1;
> @@ -504,6 +510,143 @@ static void update_benchmark(const char *fmt, ...)
> }
> }
>
> +/* sub2video hack:
> + Convert subtitles to video with alpha to insert them in filter graphs.
> + This is a temporary solution until libavfilter gets real subtitles support.
> + */
> +
> +
> +static int sub2video_prepare(InputStream *ist)
> +{
> + AVFormatContext *avf = input_files[ist->file_index]->ctx;
> + int i, ret, w, h;
> + uint8_t *image[4];
> + int linesize[4];
> +
> + w = ist->st->codec->width;
> + h = ist->st->codec->height;
> + if (!(w && h)) {
> + for (i = 0; i < avf->nb_streams; i++) {
> + if (avf->streams[i]->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
> + w = FFMAX(w, avf->streams[i]->codec->width);
> + h = FFMAX(h, avf->streams[i]->codec->height);
> + }
> + }
> + if (!(w && h)) {
> + w = FFMAX(w, 720);
> + h = FFMAX(h, 576);
> + }
> + av_log(avf, AV_LOG_INFO, "sub2video: using %dx%d canvas\n", w, h);
> + }
> + ist->sub2video.w = ist->st->codec->width = w;
> + ist->sub2video.h = ist->st->codec->height = h;
Comment this block with something along the line:
|Compute the size for the video subtitles stream. This will be the
|subtitles stream size if defined, otherwise the minimum size which
|will contain all the video streams specified in the corresponding
|input file.
or a shortened version.
> +
> + /* rectangles are PIX_FMT_PAL8, but we have no guarantee that the
> + palettes for all rectangles are identical or compatible */
> + ist->st->codec->pix_fmt = PIX_FMT_RGB32;
> +
> + ret = av_image_alloc(image, linesize, w, h, PIX_FMT_RGB32, 32);
> + if (ret < 0)
> + return ret;
> + memset(image[0], 0, h * linesize[0]);
> + ist->sub2video.ref = avfilter_get_video_buffer_ref_from_arrays(
> + image, linesize, AV_PERM_READ | AV_PERM_PRESERVE,
> + w, h, PIX_FMT_RGB32);
> + if (!ist->sub2video.ref) {
> + av_free(image[0]);
> + return AVERROR(ENOMEM);
> + }
> + return 0;
> +}
> +
> +static void sub2video_copy_rect(uint8_t *dst, int dst_linesize, int w, int h,
> + AVSubtitleRect *r)
> +{
> + uint32_t *pal, *dst2;
> + uint8_t *src, *src2;
> + int x, y;
> +
> + if (r->type != SUBTITLE_BITMAP) {
> + av_log(NULL, AV_LOG_WARNING, "sub2video: non-bitmap subtitle\n");
> + return;
> + }
> + if (r->x < 0 || r->x + r->w > w || r->y < 0 || r->y + r->h > h) {
> + av_log(NULL, AV_LOG_WARNING, "sub2video: rectangle overflowing\n");
> + return;
> + }
Maybe you could complain and write only in the overlapping region.
> +
> + dst += r->y * dst_linesize + r->x * 4;
> + src = r->pict.data[0];
> + pal = (uint32_t *)r->pict.data[1];
> + for (y = 0; y < r->h; y++) {
> + dst2 = (uint32_t *)dst;
> + src2 = src;
> + for (x = 0; x < r->w; x++)
> + *(dst2++) = pal[*(src2++)];
> + dst += dst_linesize;
> + src += r->pict.linesize[0];
> + }
> +}
> +
> +static void sub2video_push_ref(InputStream *ist, int64_t pts)
> +{
> + AVFilterBufferRef *ref = ist->sub2video.ref;
> + int i;
> +
> + ist->sub2video.last_pts = ref->pts = pts;
> + for (i = 0; i < ist->nb_filters; i++)
> + av_buffersrc_add_ref(ist->filters[i]->filter,
> + avfilter_ref_buffer(ref, ~0),
> + AV_BUFFERSRC_FLAG_NO_CHECK_FORMAT |
> + AV_BUFFERSRC_FLAG_NO_COPY);
> +}
> +
What's the implication of AV_BUFFERSRC_FLAG_NO_CHECK_FORMAT? (I mean
can it happens that the size of the buffer changes, in that case what
happens of the filtergraph?, can it cope with it?).
> +static void sub2video_update(InputStream *ist, AVSubtitle *sub, int64_t pts)
> +{
> + int w = ist->sub2video.w, h = ist->sub2video.h;
> + AVFilterBufferRef *ref = ist->sub2video.ref;
> + int8_t *dst = ref->data [0];
> + int dst_linesize = ref->linesize[0];
> + int i;
> +
> + memset(dst, 0, h * dst_linesize);
> + for (i = 0; i < sub->num_rects; i++)
> + sub2video_copy_rect(dst, dst_linesize, w, h, sub->rects[i]);
> + sub2video_push_ref(ist, pts);
> +}
> +
> +static void sub2video_heartbeat(InputStream *ist, int64_t pts)
> +{
> + InputFile *inf = input_files[ist->file_index];
nit: "inf" reminds "info", what about "infile"?
> + int i, j, nb_req;
nit: nb_reqs since it is plural
> + int64_t pts2;
> +
> + for (i = 0; i < inf->nb_streams; i++) {
> + InputStream *ist2 = input_streams[inf->ist_index + i];
> + if (!ist2->sub2video.ref)
> + continue;
> + /* subtitles seem to be usually muxed ahead of other streams;
> + if not, substracting a reasonable time here is necessary */
> + pts2 = av_rescale_q(pts, ist->st->time_base, ist2->st->time_base);
> + if (pts2 <= ist2->sub2video.last_pts)
> + continue;
> + for (j = 0, nb_req = 0; j < ist2->nb_filters; j++)
> + nb_req += av_buffersrc_get_nb_failed_requests(ist2->filters[j]->filter);
> + if (nb_req)
> + sub2video_push_ref(ist2, pts2);
> + }
> +}
Please can you explain what this does?
> +static void sub2video_flush(InputStream *ist)
> +{
> + int i;
> +
> + for (i = 0; i < ist->nb_filters; i++)
> + av_buffersrc_add_ref(ist->filters[i]->filter, NULL, 0);
> +}
> +
> +/* end of sub2video hack */
> +
> static void reset_options(OptionsContext *o, int is_input)
> {
> const OptionDef *po = options;
> @@ -745,7 +888,9 @@ static void init_input_filter(FilterGraph *fg, AVFilterInOut *in)
> s = input_files[file_idx]->ctx;
>
> for (i = 0; i < s->nb_streams; i++) {
> - if (s->streams[i]->codec->codec_type != type)
> + if (s->streams[i]->codec->codec_type != type &&
> + !(s->streams[i]->codec->codec_type == AVMEDIA_TYPE_SUBTITLE &&
> + type == AVMEDIA_TYPE_VIDEO /* sub2video hack */))
> continue;
s->streams[i]->codec->codec_type could be factored.
> if (check_stream_specifier(s, s->streams[i], *p == ':' ? p + 1 : p) == 1) {
> st = s->streams[i];
[...]
--
FFmpeg = Fierce and Fundamental Meaningful Prodigious Ecumenical Generator
More information about the ffmpeg-devel
mailing list