path: root/libhb/rendersub.c

/* rendersub.c

   Copyright (c) 2003-2019 HandBrake Team
   This file is part of the HandBrake source code
   Homepage: <http://handbrake.fr/>.
   It may be used under the terms of the GNU General Public License v2.
   For full terms see the file COPYING file or visit http://www.gnu.org/licenses/gpl-2.0.html
 */

#include "handbrake.h"
#include "hbffmpeg.h"
#include <ass/ass.h>

#define ABS(a) ((a) > 0 ? (a) : (-(a)))

struct hb_filter_private_s
{
    // Common
    int                 crop[4];
    int                 type;
    struct SwsContext * sws;
    int                 sws_width;
    int                 sws_height;

    // VOBSUB
    hb_list_t         * sub_list; // List of active subs

    // SSA
    ASS_Library       * ssa;
    ASS_Renderer      * renderer;
    ASS_Track         * ssaTrack;
    uint8_t             script_initialized;

    // SRT
    int                 line;
    hb_buffer_t       * current_sub;

    hb_filter_init_t    input;
    hb_filter_init_t    output;
};

// VOBSUB
static int vobsub_post_init( hb_filter_object_t * filter, hb_job_t * job );

static int vobsub_work( hb_filter_object_t * filter,
                        hb_buffer_t ** buf_in,
                        hb_buffer_t ** buf_out );

static void vobsub_close( hb_filter_object_t * filter );


// SSA
static int ssa_post_init( hb_filter_object_t * filter, hb_job_t * job );

static int ssa_work( hb_filter_object_t * filter,
                     hb_buffer_t ** buf_in,
                     hb_buffer_t ** buf_out );

static void ssa_close( hb_filter_object_t * filter );


// SRT
static int textsub_post_init( hb_filter_object_t * filter, hb_job_t * job );
static int cc608sub_post_init( hb_filter_object_t * filter, hb_job_t * job );

static int textsub_work( hb_filter_object_t * filter,
                     hb_buffer_t ** buf_in,
                     hb_buffer_t ** buf_out );

static void textsub_close( hb_filter_object_t * filter );


// PGS
static int pgssub_post_init( hb_filter_object_t * filter, hb_job_t * job );

static int pgssub_work ( hb_filter_object_t * filter,
                      hb_buffer_t ** buf_in,
                      hb_buffer_t ** buf_out );

static void pgssub_close( hb_filter_object_t * filter );


// Entry points
static int hb_rendersub_init( hb_filter_object_t * filter,
                                 hb_filter_init_t * init );

static int hb_rendersub_post_init( hb_filter_object_t * filter, hb_job_t *job );

static int hb_rendersub_work( hb_filter_object_t * filter,
                                 hb_buffer_t ** buf_in,
                                 hb_buffer_t ** buf_out );

static void hb_rendersub_close( hb_filter_object_t * filter );

hb_filter_object_t hb_filter_render_sub =
{
    .id            = HB_FILTER_RENDER_SUB,
    .enforce_order = 1,
    .name          = "Subtitle renderer",
    .settings      = NULL,
    .init          = hb_rendersub_init,
    .post_init     = hb_rendersub_post_init,
    .work          = hb_rendersub_work,
    .close         = hb_rendersub_close,
};

// blends src YUVA420P buffer into dst
// dst is currently YUV420P, but in future will be other formats as well
static void blend( hb_buffer_t *dst, hb_buffer_t *src, int left, int top )
{
    int xx, yy;
    int ww, hh;
    int x0, y0;
    uint8_t *y_in, *y_out;
    uint8_t *u_in, *u_out;
    uint8_t *v_in, *v_out;
    uint8_t *a_in, alpha;

    x0 = y0 = 0;
    if( left < 0 )
    {
        x0 = -left;
    }
    if( top < 0 )
    {
        y0 = -top;
    }

    ww = src->f.width;
    if( src->f.width - x0 > dst->f.width - left )
    {
        ww = dst->f.width - left + x0;
    }
    hh = src->f.height;
    if( src->f.height - y0 > dst->f.height - top )
    {
        hh = dst->f.height - top + y0;
    }
    // Blend luma
    for( yy = y0; yy < hh; yy++ )
    {
        y_in   = src->plane[0].data + yy * src->plane[0].stride;
        y_out   = dst->plane[0].data + ( yy + top ) * dst->plane[0].stride;
        a_in = src->plane[3].data + yy * src->plane[3].stride;
        for( xx = x0; xx < ww; xx++ )
        {
            alpha = a_in[xx];
            /*
             * Merge the luminance and alpha with the picture
             */
            y_out[left + xx] =
                ( (uint16_t)y_out[left + xx] * ( 255 - alpha ) +
                     (uint16_t)y_in[xx] * alpha ) / 255;
        }
    }

    // Blend U & V
    // Assumes source and dest are the same PIX_FMT
    int hshift = 0;
    int wshift = 0;
    if( dst->plane[1].height < dst->plane[0].height )
        hshift = 1;
    if( dst->plane[1].width < dst->plane[0].width )
        wshift = 1;

    for( yy = y0 >> hshift; yy < hh >> hshift; yy++ )
    {
        u_in = src->plane[1].data + yy * src->plane[1].stride;
        u_out = dst->plane[1].data + ( yy + ( top >> hshift ) ) * dst->plane[1].stride;
        v_in = src->plane[2].data + yy * src->plane[2].stride;
        v_out = dst->plane[2].data + ( yy + ( top >> hshift ) ) * dst->plane[2].stride;
        a_in = src->plane[3].data + ( yy << hshift ) * src->plane[3].stride;

        for( xx = x0 >> wshift; xx < ww >> wshift; xx++ )
        {
            alpha = a_in[xx << wshift];

            // Blend averge U and alpha
            u_out[(left >> wshift) + xx] =
                ( (uint16_t)u_out[(left >> wshift) + xx] * ( 255 - alpha ) +
                  (uint16_t)u_in[xx] * alpha ) / 255;

            // Blend V and alpha
            v_out[(left >> wshift) + xx] =
                ( (uint16_t)v_out[(left >> wshift) + xx] * ( 255 - alpha ) +
                  (uint16_t)v_in[xx] * alpha ) / 255;
        }
    }
}

// applies subtitle 'sub' YUVA420P buffer into destination 'buf'
// 'buf' is currently YUV420P, but in future will be other formats as well
// Assumes that the input destination buffer has the same dimensions
// as the original title dimensions
static void ApplySub( hb_filter_private_t * pv, hb_buffer_t * buf, hb_buffer_t * sub )
{
    blend( buf, sub, sub->f.x, sub->f.y );
}

static hb_buffer_t * ScaleSubtitle(hb_filter_private_t *pv,
                                   hb_buffer_t *sub, hb_buffer_t *buf)
{
    hb_buffer_t * scaled;
    double xfactor = 1., yfactor = 1.;

    // Do we need to rescale subtitles?
    if (sub->f.window_width > 0 && sub->f.window_height > 0)
    {

        // TODO: Factor aspect ratio
        // For now, assume subtitle and video PAR is the same.
        xfactor     = (double)buf->f.width  / sub->f.window_width;
        yfactor     = (double)buf->f.height / sub->f.window_height;
        // The video may have been cropped.  This will make xfactor != yfactor
        // even though video and subtitles are the same PAR.  So use the
        // larger of as the scale factor.
        if (xfactor > yfactor)
        {
            yfactor = xfactor;
        }
        else
        {
            xfactor = yfactor;
        }
    }
    if (ABS(xfactor - 1) > 0.01 || ABS(yfactor - 1) > 0.01)
    {
        uint8_t * in_data[4], * out_data[4];
        int       in_stride[4], out_stride[4];
        int       width, height;

        width       = sub->f.width  * xfactor;
        height      = sub->f.height * yfactor;
        // Note that subtitle frame buffer is YUVA420P, not YUV420P, it has alpha
        scaled      = hb_frame_buffer_init(AV_PIX_FMT_YUVA420P, width, height);
        if (scaled == NULL)
            return NULL;

        scaled->f.x = sub->f.x * xfactor;
        scaled->f.y = sub->f.y * yfactor;

        hb_picture_fill(in_data,  in_stride,  sub);
        hb_picture_fill(out_data, out_stride, scaled);

        if (pv->sws        == NULL   ||
            pv->sws_width  != width  ||
            pv->sws_height != height)
        {
            if (pv->sws!= NULL)
                sws_freeContext(pv->sws);
            pv->sws = hb_sws_get_context(
                                sub->f.width, sub->f.height, sub->f.fmt,
                                scaled->f.width, scaled->f.height, sub->f.fmt,
                                SWS_LANCZOS|SWS_ACCURATE_RND, SWS_CS_DEFAULT);
            pv->sws_width   = width;
            pv->sws_height  = height;
        }
        sws_scale(pv->sws, (const uint8_t* const *)in_data, in_stride,
                  0, sub->f.height, out_data, out_stride);
    }
    else
    {
        scaled = hb_buffer_dup(sub);
    }

    int top, left, margin_top, margin_percent;

    /*
     * Percent of height of picture that form a margin that subtitles
     * should not be displayed within.
     */
    margin_percent = 2;

    /*
     * If necessary, move the subtitle so it is not in a cropped zone.
     * When it won't fit, we center it so we lose as much on both ends.
     * Otherwise we try to leave a 20px or 2% margin around it.
     */
    margin_top = ( ( buf->f.height - pv->crop[0] - pv->crop[1] ) *
                   margin_percent ) / 100;

    if( margin_top > 20 )
    {
        /*
         * A maximum margin of 20px regardless of height of the picture.
         */
        margin_top = 20;
    }

    if( scaled->f.height > buf->f.height - pv->crop[0] - pv->crop[1] -
        ( margin_top * 2 ) )
    {
        /*
         * The subtitle won't fit in the cropped zone, so center
         * it vertically so we fit in as much as we can.
         */
        top = pv->crop[0] + ( buf->f.height - pv->crop[0] -
                                      pv->crop[1] - scaled->f.height ) / 2;
    }
    else if( scaled->f.y < pv->crop[0] + margin_top )
    {
        /*
         * The subtitle fits in the cropped zone, but is currently positioned
         * within our top margin, so move it outside of our margin.
         */
        top = pv->crop[0] + margin_top;
    }
    else if( scaled->f.y > buf->f.height - pv->crop[1] - margin_top - scaled->f.height )
    {
        /*
         * The subtitle fits in the cropped zone, and is not within the top
         * margin but is within the bottom margin, so move it to be above
         * the margin.
         */
        top = buf->f.height - pv->crop[1] - margin_top - scaled->f.height;
    }
    else
    {
        /*
         * The subtitle is fine where it is.
         */
        top = scaled->f.y;
    }

    if( scaled->f.width > buf->f.width - pv->crop[2] - pv->crop[3] - 40 )
        left = pv->crop[2] + ( buf->f.width - pv->crop[2] -
                pv->crop[3] - scaled->f.width ) / 2;
    else if( scaled->f.x < pv->crop[2] + 20 )
        left = pv->crop[2] + 20;
    else if( scaled->f.x > buf->f.width - pv->crop[3] - 20 - scaled->f.width )
        left = buf->f.width - pv->crop[3] - 20 - scaled->f.width;
    else
        left = scaled->f.x;

    scaled->f.x = left;
    scaled->f.y = top;

    return scaled;
}

// Assumes that the input buffer has the same dimensions
// as the original title dimensions
static void ApplyVOBSubs( hb_filter_private_t * pv, hb_buffer_t * buf )
{
    int ii;
    hb_buffer_t *sub, *next;

    for( ii = 0; ii < hb_list_count(pv->sub_list); )
    {
        sub = hb_list_item( pv->sub_list, ii );
        if (ii + 1 < hb_list_count(pv->sub_list))
            next = hb_list_item( pv->sub_list, ii + 1 );
        else
            next = NULL;

        if ((sub->s.stop != AV_NOPTS_VALUE && sub->s.stop <= buf->s.start) ||
            (next != NULL && sub->s.stop == AV_NOPTS_VALUE && next->s.start <= buf->s.start))
        {
            // Subtitle stop is in the past, delete it
            hb_list_rem( pv->sub_list, sub );
            hb_buffer_close( &sub );
        }
        else if( sub->s.start <= buf->s.start )
        {
            // The subtitle has started before this frame and ends
            // after it.  Render the subtitle into the frame.
            while ( sub )
            {
                hb_buffer_t *scaled = ScaleSubtitle(pv, sub, buf);
                ApplySub( pv, buf, scaled );
                hb_buffer_close(&scaled);
                sub = sub->next;
            }
            ii++;
        }
        else
        {
            // The subtitle starts in the future.  No need to continue.
            break;
        }
    }
}

static int vobsub_post_init( hb_filter_object_t * filter, hb_job_t * job )
{
    hb_filter_private_t * pv = filter->private_data;

    pv->sub_list = hb_list_init();

    return 0;
}

static void vobsub_close( hb_filter_object_t * filter )
{
    hb_filter_private_t * pv = filter->private_data;

    if( !pv )
    {
        return;
    }

    if( pv->sub_list )
        hb_list_empty( &pv->sub_list );

    free( pv );
    filter->private_data = NULL;
}

static int vobsub_work( hb_filter_object_t * filter,
                        hb_buffer_t ** buf_in,
                        hb_buffer_t ** buf_out )
{
    hb_filter_private_t * pv = filter->private_data;
    hb_buffer_t * in = *buf_in;
    hb_buffer_t * sub;

    if (in->s.flags & HB_BUF_FLAG_EOF)
    {
        *buf_in = NULL;
        *buf_out = in;
        return HB_FILTER_DONE;
    }

    // Get any pending subtitles and add them to the active
    // subtitle list
    while( ( sub = hb_fifo_get( filter->subtitle->fifo_out ) ) )
    {
        if (sub->s.flags & HB_BUF_FLAG_EOF)
        {
            hb_buffer_close(&sub);
            break;
        }
        hb_list_add( pv->sub_list, sub );
    }

    ApplyVOBSubs( pv, in );
    *buf_in = NULL;
    *buf_out = in;

    return HB_FILTER_OK;
}

static uint8_t ssaAlpha( ASS_Image *frame, int x, int y )
{
    unsigned frameA = ( frame->color ) & 0xff;
    unsigned gliphA = frame->bitmap[y*frame->stride + x];

    // Alpha for this pixel is the frame opacity (255 - frameA)
    // multiplied by the gliph alfa (gliphA) for this pixel
    unsigned alpha = (255 - frameA) * gliphA >> 8;

    return (uint8_t)alpha;
}

// Returns a subtitle rendered to a YUVA420P frame 
static hb_buffer_t * RenderSSAFrame( hb_filter_private_t * pv, ASS_Image * frame )
{
    hb_buffer_t *sub;
    int xx, yy;

    unsigned r = ( frame->color >> 24 ) & 0xff;
    unsigned g = ( frame->color >> 16 ) & 0xff;
    unsigned b = ( frame->color >>  8 ) & 0xff;

    int yuv = hb_rgb2yuv((r << 16) | (g << 8) | b );

    unsigned frameY = (yuv >> 16) & 0xff;
    unsigned frameV = (yuv >> 8 ) & 0xff;
    unsigned frameU = (yuv >> 0 ) & 0xff;

    // Note that subtitle frame buffer is YUVA420P, not YUV420P, it has alpha
    sub = hb_frame_buffer_init(AV_PIX_FMT_YUVA420P, frame->w, frame->h);
    if (sub == NULL)
        return NULL;

    uint8_t *y_out, *u_out, *v_out, *a_out;
    y_out = sub->plane[0].data;
    u_out = sub->plane[1].data;
    v_out = sub->plane[2].data;
    a_out = sub->plane[3].data;

    for( yy = 0; yy < frame->h; yy++ )
    {
        for( xx = 0; xx < frame->w; xx++ )
        {
            y_out[xx] = frameY;
            if( ( yy & 1 ) == 0 )
            {
                u_out[xx>>1] = frameU;
                v_out[xx>>1] = frameV;
            }
            a_out[xx] = ssaAlpha( frame, xx, yy );;
        }
        y_out += sub->plane[0].stride;
        if( ( yy & 1 ) == 0 )
        {
            u_out += sub->plane[1].stride;
            v_out += sub->plane[2].stride;
        }
        a_out += sub->plane[3].stride;
    }
    sub->f.width = frame->w;
    sub->f.height = frame->h;
    sub->f.x = frame->dst_x + pv->crop[2];
    sub->f.y = frame->dst_y + pv->crop[0];

    return sub;
}

static void ApplySSASubs( hb_filter_private_t * pv, hb_buffer_t * buf )
{
    ASS_Image *frameList;
    hb_buffer_t *sub;

    frameList = ass_render_frame( pv->renderer, pv->ssaTrack,
                                  buf->s.start / 90, NULL );
    if ( !frameList )
        return;

    ASS_Image *frame;
    for (frame = frameList; frame; frame = frame->next) {
        sub = RenderSSAFrame( pv, frame );
        if( sub )
        {
            ApplySub( pv, buf, sub );
            hb_buffer_close( &sub );
        }
    }
}

static void ssa_log(int level, const char *fmt, va_list args, void *data)
{
    if ( level < 5 )      // same as default verbosity when no callback is set
    {
        hb_valog( 1, "[ass]", fmt, args );
    }
}

static int ssa_post_init( hb_filter_object_t * filter, hb_job_t * job )
{
    hb_filter_private_t * pv = filter->private_data;

    pv->ssa = ass_library_init();
    if ( !pv->ssa ) {
        hb_error( "decssasub: libass initialization failed\n" );
        return 1;
    }

    // Redirect libass output to hb_log
    ass_set_message_cb( pv->ssa, ssa_log, NULL );

    // Load embedded fonts
    hb_list_t * list_attachment = job->list_attachment;
    int i;
    for ( i = 0; i < hb_list_count(list_attachment); i++ )
    {
        hb_attachment_t * attachment = hb_list_item( list_attachment, i );

        if ( attachment->type == FONT_TTF_ATTACH ||
             attachment->type == FONT_OTF_ATTACH )
        {
            ass_add_font(
                pv->ssa,
                attachment->name,
                attachment->data,
                attachment->size );
        }
    }

    ass_set_extract_fonts( pv->ssa, 1 );
    ass_set_style_overrides( pv->ssa, NULL );

    pv->renderer = ass_renderer_init( pv->ssa );
    if ( !pv->renderer ) {
        hb_log( "decssasub: renderer initialization failed\n" );
        return 1;
    }

    ass_set_use_margins( pv->renderer, 0 );
    ass_set_hinting( pv->renderer, ASS_HINTING_LIGHT ); // VLC 1.0.4 uses this
    ass_set_font_scale( pv->renderer, 1.0 );
    ass_set_line_spacing( pv->renderer, 1.0 );

    // Setup default font family
    //
    // SSA v4.00 requires that "Arial" be the default font
    const char *font = NULL;
    const char *family = "Arial";
    // NOTE: This can sometimes block for several *seconds*.
    //       It seems that process_fontdata() for some embedded fonts is slow.
    ass_set_fonts( pv->renderer, font, family, /*haveFontConfig=*/1, NULL, 1 );

    // Setup track state
    pv->ssaTrack = ass_new_track( pv->ssa );
    if ( !pv->ssaTrack ) {
        hb_log( "decssasub: ssa track initialization failed\n" );
        return 1;
    }

    int height = job->title->geometry.height - job->crop[0] - job->crop[1];
    int width = job->title->geometry.width - job->crop[2] - job->crop[3];
    ass_set_frame_size( pv->renderer, width, height);

    double par = (double)job->par.num / job->par.den;
    ass_set_aspect_ratio( pv->renderer, 1, par );

    return 0;
}

static void ssa_close( hb_filter_object_t * filter )
{
    hb_filter_private_t * pv = filter->private_data;

    if( !pv )
    {
        return;
    }

    if ( pv->ssaTrack )
        ass_free_track( pv->ssaTrack );
    if ( pv->renderer )
        ass_renderer_done( pv->renderer );
    if ( pv->ssa )
        ass_library_done( pv->ssa );

    free( pv );
    filter->private_data = NULL;
}

static int ssa_work( hb_filter_object_t * filter,
                     hb_buffer_t ** buf_in,
                     hb_buffer_t ** buf_out )
{
    hb_filter_private_t * pv = filter->private_data;
    hb_buffer_t * in = *buf_in;
    hb_buffer_t * sub;

    if (!pv->script_initialized)
    {
        // NOTE: The codec extradata is expected to be in MKV format
        // I would like to initialize this in ssa_post_init, but when we are
        // transcoding text subtitles to SSA, the extradata does not
        // get initialized until the decoder is initialized.  Since
        // decoder initialization happens after filter initialization,
        // we need to postpone this.
        ass_process_codec_private(pv->ssaTrack,
                                  (char*)filter->subtitle->extradata,
                                  filter->subtitle->extradata_size);
        pv->script_initialized = 1;
    }
    if (in->s.flags & HB_BUF_FLAG_EOF)
    {
        *buf_in = NULL;
        *buf_out = in;
        return HB_FILTER_DONE;
    }

    // Get any pending subtitles and add them to the active
    // subtitle list
    while( ( sub = hb_fifo_get( filter->subtitle->fifo_out ) ) )
    {
        if (sub->s.flags & HB_BUF_FLAG_EOF)
        {
            hb_buffer_close(&sub);
            break;
        }
        // Parse MKV-SSA packet
        // SSA subtitles always have an explicit stop time, so we
        // do not need to do special processing for stop == AV_NOPTS_VALUE
        ass_process_chunk( pv->ssaTrack, (char*)sub->data, sub->size,
                           sub->s.start / 90,
                           (sub->s.stop - sub->s.start) / 90 );
        hb_buffer_close(&sub);
    }

    ApplySSASubs( pv, in );
    *buf_in = NULL;
    *buf_out = in;

    return HB_FILTER_OK;
}

static int cc608sub_post_init( hb_filter_object_t * filter, hb_job_t * job )
{
    // Text subtitles for which we create a dummy ASS header need
    // to have the header rewritten with the correct dimensions.
    int height = job->title->geometry.height - job->crop[0] - job->crop[1];
    int width = job->title->geometry.width - job->crop[2] - job->crop[3];
    int safe_height = 0.8 * job->title->geometry.height;
    // Use fixed width font for CC
    hb_subtitle_add_ssa_header(filter->subtitle, HB_FONT_MONO,
                               .08 * safe_height, width, height);
    return ssa_post_init(filter, job);
}

static int textsub_post_init( hb_filter_object_t * filter, hb_job_t * job )
{
    // Text subtitles for which we create a dummy ASS header need
    // to have the header rewritten with the correct dimensions.
    int height = job->title->geometry.height - job->crop[0] - job->crop[1];
    int width = job->title->geometry.width - job->crop[2] - job->crop[3];
    hb_subtitle_add_ssa_header(filter->subtitle, HB_FONT_SANS,
                               .066 * job->title->geometry.height,
                               width, height);
    return ssa_post_init(filter, job);
}

static void textsub_close( hb_filter_object_t * filter )
{
    return ssa_close(filter);
}

static void process_sub(hb_filter_private_t *pv, hb_buffer_t *sub)
{
    int64_t start, dur;
    char *ssa, *tmp;

    // libass expects every chunk to have a unique sequence number
    // since we are repeating subs in some cases, we need to replace
    // the sequence number.
    tmp = strchr((char*)sub->data, ',');
    if (tmp == NULL)
        return;

    ssa = hb_strdup_printf("%d%s", ++pv->line, tmp);

    // Parse MKV-SSA packet
    // SSA subtitles always have an explicit stop time, so we
    // do not need to do special processing for stop == AV_NOPTS_VALUE
    start = sub->s.start;
    dur = sub->s.stop - sub->s.start;
    ass_process_chunk(pv->ssaTrack, ssa, sub->size, start, dur);
    free(ssa);
}

static int textsub_work(hb_filter_object_t * filter,
                    hb_buffer_t ** buf_in,
                    hb_buffer_t ** buf_out)
{
    hb_filter_private_t * pv = filter->private_data;
    hb_buffer_t * in = *buf_in;
    hb_buffer_t * sub;

    if (!pv->script_initialized)
    {
        ass_process_codec_private(pv->ssaTrack,
                                  (char*)filter->subtitle->extradata,
                                  filter->subtitle->extradata_size);
        pv->script_initialized = 1;
    }

    if (in->s.flags & HB_BUF_FLAG_EOF)
    {
        *buf_in = NULL;
        *buf_out = in;
        return HB_FILTER_DONE;
    }

    int in_start_ms = in->s.start / 90;

    // Get any pending subtitles and add them to the active
    // subtitle list
    while ((sub = hb_fifo_get(filter->subtitle->fifo_out)))
    {
        if (sub->s.flags & HB_BUF_FLAG_EOF)
        {
            hb_buffer_close(&sub);
            if (pv->current_sub != NULL)
            {
                // Make us some duration for final sub
                pv->current_sub->s.stop = pv->current_sub->s.start +
                                          90000LL * 10;
                process_sub(pv, pv->current_sub);
                hb_buffer_close(&pv->current_sub);
            }
            break;
        }

        // libass expects times in ms.  So to make the math easy,
        // convert to ms immediately.
        sub->s.start /= 90;
        if (sub->s.stop != AV_NOPTS_VALUE)
        {
            sub->s.stop /= 90;
        }

        // Subtitle formats such as CC can have stop times
        // that are not known until an "erase display" command
        // is encountered in the stream.  For these formats
        // current_sub is the currently active subtitle for which
        // we do not yet know the stop time.  We do not currently
        // support overlapping subtitles of this type.
        if (pv->current_sub != NULL)
        {
            // Next sub start time tells us the stop time of the
            // current sub when it is not known in advance.
            pv->current_sub->s.stop = sub->s.start;
            process_sub(pv, pv->current_sub);
            hb_buffer_close(&pv->current_sub);
        }
        if (sub->s.flags & HB_BUF_FLAG_EOS)
        {
            // marker used to "clear" previous sub that had
            // an unknown duration
            hb_buffer_close(&sub);
        }
        else if (sub->s.stop == AV_NOPTS_VALUE)
        {
            // We don't know the duration of this sub.  So we will
            // apply it to every video frame until we see a "clear" sub.
            pv->current_sub = sub;
            pv->current_sub->s.stop = pv->current_sub->s.start;
        }
        else
        {
            // Duration of this subtitle is known, so we can just
            // process it normally.
            process_sub(pv, sub);
            hb_buffer_close(&sub);
        }
    }
    if (pv->current_sub != NULL && pv->current_sub->s.start <= in_start_ms)
    {
        // We don't know the duration of this subtitle, but we know
        // that it started before the current video frame and that
        // it is still active.  So render it on this video frame.
        pv->current_sub->s.start = pv->current_sub->s.stop;
        pv->current_sub->s.stop = in_start_ms + 1;
        process_sub(pv, pv->current_sub);
    }

    ApplySSASubs(pv, in);
    *buf_in = NULL;
    *buf_out = in;

    return HB_FILTER_OK;
}

static void ApplyPGSSubs( hb_filter_private_t * pv, hb_buffer_t * buf )
{
    int index;
    hb_buffer_t * old_sub;
    hb_buffer_t * sub;

    // Each PGS subtitle supersedes anything that preceded it.
    // Find the active subtitle (if there is one), and delete
    // everything before it.
    for( index = hb_list_count( pv->sub_list ) - 1; index > 0; index-- )
    {
        sub = hb_list_item( pv->sub_list, index);
        if ( sub->s.start <= buf->s.start )
        {
            while ( index > 0 )
            {
                old_sub = hb_list_item( pv->sub_list, index - 1);
                hb_list_rem( pv->sub_list, old_sub );
                hb_buffer_close( &old_sub );
                index--;
            }
        }
    }

    // Some PGS subtitles have no content and only serve to clear
    // the screen. If any of these are at the front of our list,
    // we can now get rid of them.
    while ( hb_list_count( pv->sub_list ) > 0 )
    {
        sub = hb_list_item( pv->sub_list, 0 );
        if (sub->f.width != 0 && sub->f.height != 0)
            break;

        hb_list_rem( pv->sub_list, sub );
        hb_buffer_close( &sub );
    }

    // Check to see if there's an active subtitle, and apply it.
    if ( hb_list_count( pv->sub_list ) > 0)
    {
        sub = hb_list_item( pv->sub_list, 0 );
        if ( sub->s.start <= buf->s.start )
        {
            hb_buffer_t *scaled = ScaleSubtitle(pv, sub, buf);
            ApplySub( pv, buf, scaled );
            hb_buffer_close(&scaled);
        }
    }
}

static int pgssub_post_init( hb_filter_object_t * filter, hb_job_t * job )
{
    hb_filter_private_t * pv = filter->private_data;

    pv->sub_list = hb_list_init();

    return 0;
}

static void pgssub_close( hb_filter_object_t * filter )
{
    hb_filter_private_t * pv = filter->private_data;

    if ( !pv )
    {
        return;
    }

    if ( pv->sub_list )
        hb_list_empty( &pv->sub_list );

    free( pv );
    filter->private_data = NULL;
}

static int pgssub_work( hb_filter_object_t * filter,
                        hb_buffer_t ** buf_in,
                        hb_buffer_t ** buf_out)
{
    hb_filter_private_t * pv = filter->private_data;
    hb_buffer_t * in = *buf_in;
    hb_buffer_t * sub;

    if (in->s.flags & HB_BUF_FLAG_EOF)
    {
        *buf_in = NULL;
        *buf_out = in;
        return HB_FILTER_DONE;
    }

    // Get any pending subtitles and add them to the active
    // subtitle list
    while ( ( sub = hb_fifo_get( filter->subtitle->fifo_out ) ) )
    {
        if (sub->s.flags & HB_BUF_FLAG_EOF)
        {
            hb_buffer_close(&sub);
            break;
        }
        hb_list_add( pv->sub_list, sub );
    }

    ApplyPGSSubs( pv, in );
    *buf_in = NULL;
    *buf_out = in;

    return HB_FILTER_OK;
}

static int hb_rendersub_init( hb_filter_object_t * filter,
                              hb_filter_init_t * init )
{
    filter->private_data = calloc( 1, sizeof(struct hb_filter_private_s) );
    hb_filter_private_t * pv = filter->private_data;
    hb_subtitle_t *subtitle;
    int ii;

    pv->input = *init;

    // Find the subtitle we need
    for( ii = 0; ii < hb_list_count(init->job->list_subtitle); ii++ )
    {
        subtitle = hb_list_item( init->job->list_subtitle, ii );
        if( subtitle && subtitle->config.dest == RENDERSUB )
        {
            // Found it
            filter->subtitle = subtitle;
            pv->type = subtitle->source;
            break;
        }
    }
    if( filter->subtitle == NULL )
    {
        hb_log("rendersub: no subtitle marked for burn");
        return 1;
    }
    pv->output = *init;

    return 0;
}

static int hb_rendersub_post_init( hb_filter_object_t * filter, hb_job_t *job )
{
    hb_filter_private_t * pv = filter->private_data;

    pv->crop[0] = job->crop[0];
    pv->crop[1] = job->crop[1];
    pv->crop[2] = job->crop[2];
    pv->crop[3] = job->crop[3];

    switch( pv->type )
    {
        case VOBSUB:
        {
            return vobsub_post_init( filter, job );
        } break;

        case SSASUB:
        {
            return ssa_post_init( filter, job );
        } break;

        case IMPORTSRT:
        case IMPORTSSA:
        case UTF8SUB:
        case TX3GSUB:
        {
            return textsub_post_init( filter, job );
        } break;

        case CC608SUB:
        {
            return cc608sub_post_init( filter, job );
        } break;

        case PGSSUB:
        {
            return pgssub_post_init( filter, job );
        } break;

        default:
        {
            hb_log("rendersub: unsupported subtitle format %d", pv->type );
            return 1;
        } break;
    }
    return 0;
}

static int hb_rendersub_work( hb_filter_object_t * filter,
                                 hb_buffer_t ** buf_in,
                                 hb_buffer_t ** buf_out )
{
    hb_filter_private_t * pv = filter->private_data;
    switch( pv->type )
    {
        case VOBSUB:
        {
            return vobsub_work( filter, buf_in, buf_out );
        } break;

        case SSASUB:
        {
            return ssa_work( filter, buf_in, buf_out );
        } break;

        case IMPORTSRT:
        case IMPORTSSA:
        case CC608SUB:
        case UTF8SUB:
        case TX3GSUB:
        {
            return textsub_work( filter, buf_in, buf_out );
        } break;

        case PGSSUB:
        {
            return pgssub_work( filter, buf_in, buf_out );
        } break;

        default:
        {
            hb_error("rendersub: unsupported subtitle format %d", pv->type );
            return 1;
        } break;
    }
}

static void hb_rendersub_close( hb_filter_object_t * filter )
{
    hb_filter_private_t * pv = filter->private_data;

    if (pv->sws != NULL)
    {
        sws_freeContext(pv->sws);
    }
    switch( pv->type )
    {
        case VOBSUB:
        {
            vobsub_close( filter );
        } break;

        case SSASUB:
        {
            ssa_close( filter );
        } break;

        case IMPORTSRT:
        case IMPORTSSA:
        case CC608SUB:
        case UTF8SUB:
        case TX3GSUB:
        {
            textsub_close( filter );
        } break;

        case PGSSUB:
        {
            pgssub_close( filter );
        } break;

        default:
        {
            hb_error("rendersub: unsupported subtitle format %d", pv->type );
        } break;
    }
}