[FFmpeg-devel] [PATCH 01/12] Implemented text to bitmap subtitles!

[Traian Coza <traian.coza@gmail.com>]

---
 fftools/ffmpeg.c           |  47 ++++++++++++++-
 libavfilter/vf_subtitles.c | 117 +++++++++++++++++++++++++++++++++++++
 2 files changed, 163 insertions(+), 1 deletion(-)

diff --git a/fftools/ffmpeg.c b/fftools/ffmpeg.c
index a85ed18b08..53717d3ebb 100644
--- a/fftools/ffmpeg.c
+++ b/fftools/ffmpeg.c
@@ -2325,6 +2325,45 @@ fail:
     return err < 0 ? err : ret;
 }
 
+void render_avsub_ass(InputStream *, AVSubtitle *);
+
+static void print_subtitle(AVSubtitle sub)
+{
+    printf("sub.format: %u\n", sub.format);
+    printf("sub.start_display_time: %u\n", sub.start_display_time);
+    printf("sub.end_display_time: %u\n", sub.end_display_time);
+    printf("sub.num_rects: %u\n", sub.num_rects);
+    printf("sub.pts: %ld\n", sub.pts);
+    for (int i = 0; i < sub.num_rects; i++)
+    {
+        printf("sub.rects[%d]->type: %d\n", i, sub.rects[i]->type);
+        printf("sub.rects[%d]->nb_colors: %d\n", i, sub.rects[i]->nb_colors);
+        printf("sub.rects[%d]->(x,y,w,h): (%d,%d,%d,%d)\n", i, sub.rects[i]->x, sub.rects[i]->y, sub.rects[i]->w, sub.rects[i]->h);
+        printf("sub.rects[%d]->linesize: [%d,%d,%d,%d]\n", i, sub.rects[i]->linesize[0], sub.rects[i]->linesize[1], sub.rects[i]->linesize[1], sub.rects[i]->linesize[1]);
+        switch (sub.rects[i]->type)
+        {
+            case SUBTITLE_TEXT:
+                printf("sub.rects[%d]->text: %s\n", i, sub.rects[i]->text);
+                break;
+            case SUBTITLE_ASS:
+                printf("sub.rects[%d]->ass: %s\n", i, sub.rects[i]->ass);
+                break;
+            case SUBTITLE_BITMAP:
+                for (int c = 0; c < sub.rects[i]->nb_colors; c++)
+                    printf("color %d: [%u,%u,%u,%u]\n", c,
+                           sub.rects[i]->data[1][c * 4 + 0],
+                           sub.rects[i]->data[1][c * 4 + 1],
+                           sub.rects[i]->data[1][c * 4 + 2],
+                           sub.rects[i]->data[1][c * 4 + 3]);
+                for (int y = 0; y < sub.rects[i]->h; y++, printf("\n"))
+                    for (int x = 0; x < sub.rects[i]->w; x++)
+                        printf("%d", sub.rects[i]->data[0][y * sub.rects[i]->w + x]);
+                break;
+        }
+    }
+    printf("\n");
+}
+
 static int transcode_subtitles(InputStream *ist, AVPacket *pkt, int *got_output,
                                int *decode_failed)
 {
@@ -2383,6 +2422,8 @@ static int transcode_subtitles(InputStream *ist, AVPacket *pkt, int *got_output,
         goto out;
 
     ist->frames_decoded++;
+    
+    render_avsub_ass(ist, &subtitle);
 
     for (i = 0; i < nb_output_streams; i++) {
         OutputStream *ost = output_streams[i];
@@ -3213,11 +3254,15 @@ static int init_output_stream(OutputStream *ost, AVFrame *frame,
                 input_props = input_descriptor->props & (AV_CODEC_PROP_TEXT_SUB | AV_CODEC_PROP_BITMAP_SUB);
             if (output_descriptor)
                 output_props = output_descriptor->props & (AV_CODEC_PROP_TEXT_SUB | AV_CODEC_PROP_BITMAP_SUB);
-            if (input_props && output_props && input_props != output_props) {
+            /*if (input_props && output_props && input_props != output_props) {
                 snprintf(error, error_len,
                          "Subtitle encoding currently only possible from text to text "
                          "or bitmap to bitmap");
                 return AVERROR_INVALIDDATA;
+            }*/
+            if (input_props == AV_CODEC_PROP_BITMAP_SUB && output_props == AV_CODEC_PROP_TEXT_SUB) {
+                snprintf(error, error_len, "Subtitle encoding from bitmap to text currently not possible");
+                return AVERROR_INVALIDDATA;
             }
         }
 
diff --git a/libavfilter/vf_subtitles.c b/libavfilter/vf_subtitles.c
index 82e140e986..f751786033 100644
--- a/libavfilter/vf_subtitles.c
+++ b/libavfilter/vf_subtitles.c
@@ -181,6 +181,24 @@ static void overlay_ass_image(AssContext *ass, AVFrame *picref,
     }
 }
 
+static void print_ass_image(const ASS_Image *image)
+{
+    int index = 0;
+    for (; image != NULL; image = image->next, index++)
+    {
+        printf("index: %d\n", index);
+        printf("image->(dst_x,dst_y): (%d,%d)\n", image->dst_x, image->dst_y);
+        printf("image->(w,h): (%d,%d)\n", image->w, image->h);
+        printf("image->stride: %d\n", image->stride);
+        printf("image->type: %d\n", image->type);
+        printf("image->color: [%u,%u,%u,%u]\n", AR(image->color), AG(image->color), AB(image->color), AA(image->color));
+        for (int y = 0; y < image->h; y++, printf("\n"))
+            for (int x = 0; x < image->w; x++)
+                printf("%02X", image->bitmap[y * image->stride + x]);
+        printf("\n");
+    }
+}
+
 static int filter_frame(AVFilterLink *inlink, AVFrame *picref)
 {
     AVFilterContext *ctx = inlink->dst;
@@ -199,6 +217,105 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *picref)
     return ff_filter_frame(outlink, picref);
 }
 
+#include "../fftools/ffmpeg.h"
+#include "libavcodec/avcodec.h"
+#include "../libavcodec/ass_split.h"
+
+#define ALPHA_THRESHOLD 0b10000000
+
+void render_avsub_ass(InputStream *, AVSubtitle *);
+void render_avsub_ass(InputStream *ist, AVSubtitle *sub)
+{
+    ASS_Library *library = ass_library_init();
+    ass_set_extract_fonts(library, 1);
+    ASS_Renderer *renderer = ass_renderer_init(library);
+    int w = 1920, h = 1080;
+    ass_set_frame_size(renderer, w, h);
+    ass_set_pixel_aspect(renderer, 1);
+    ass_set_storage_size(renderer, w, h);
+    ass_set_shaper(renderer, 0);
+    ass_set_fonts(renderer, NULL, NULL, 1, NULL, 1);
+
+    for (int r = 0; r < sub->num_rects; r++)
+    {
+        AVSubtitleRect *rect = sub->rects[r];
+        if (rect->data[0]) continue;
+
+        ASSSplitContext *ass_context = ff_ass_split((char *)ist->dec_ctx->subtitle_header);
+        ASSDialog *dialog = ff_ass_split_dialog(ass_context, rect->ass);
+
+        ASS_Track *track = ass_read_memory(library, (char *)ist->dec_ctx->subtitle_header, ist->dec_ctx->subtitle_header_size, NULL);
+        track->n_events = track->max_events = 1;
+        track->events = (ASS_Event *)malloc(sizeof(ASS_Event));
+        track->events[0].Start = sub->start_display_time + sub->pts / (AV_TIME_BASE / 1000);
+        track->events[0].Duration = sub->end_display_time - sub->start_display_time;
+        track->events[0].Effect = strdup(dialog->effect);
+        track->events[0].Layer = dialog->layer;
+        track->events[0].MarginL = dialog->margin_l;
+        track->events[0].MarginR = dialog->margin_r;
+        track->events[0].MarginV = dialog->margin_v;
+        track->events[0].Name = strdup(dialog->name);
+        track->events[0].Text = strdup(dialog->text);
+        track->events[0].ReadOrder = dialog->readorder;
+        track->events[0].Style = 0;
+        for (int style = 0; style < track->n_styles; style++)
+            if (!strcmp(track->styles[style].Name, dialog->style))
+                track->events[0].Style = style;
+        track->events[0].render_priv = NULL;
+        ff_ass_free_dialog(&dialog);
+        ff_ass_split_free(ass_context);
+
+        int change;
+        ASS_Image *image = ass_render_frame(renderer, track, track->events[0].Start + 1, &change);   // Don't have to free it for some reason
+        printf("image: %p\n", image);
+
+        rect->x = image->dst_x; rect->w = 0;
+        rect->y = image->dst_y; rect->h = 0;
+        rect->nb_colors = 1;    // Transparent background counts as a color
+        for (ASS_Image *img = image; img != NULL; img = img->next)
+        {
+            // Set image bounds to encompass all images
+            if (img->dst_x < rect->x) rect->x = img->dst_x;
+            if (img->dst_y < rect->y) rect->y = img->dst_y;
+            if (img->dst_x + img->w > rect->x + rect->w)
+                rect->w = img->dst_x + img->w - rect->x;
+            if (img->dst_y + img->h > rect->y + rect->h)
+                rect->h = img->dst_y + img->h - rect->y;
+            rect->nb_colors++;
+        }
+        rect->linesize[0] = rect->w;
+        rect->data[0] = (uint8_t *)malloc(rect->w * rect->h * sizeof(uint8_t));
+        rect->data[1] = (uint8_t *)malloc(4 * rect->nb_colors * sizeof(uint8_t));
+        memset(rect->data[0], 0, rect->w * rect->h);        // Set all to transparent
+        memset(rect->data[1], 0, 4);                        // Set transparent color
+        memset(&rect->linesize[1], 0, 3 * sizeof(int));
+        rect->data[2] = rect->data[3] = NULL;
+        for (int color = 1; image != NULL; image = image->next, color++)
+        {
+            // Set color
+            rect->data[1][4 * color + 0] = AR(image->color);
+            rect->data[1][4 * color + 1] = AG(image->color);
+            rect->data[1][4 * color + 2] = AB(image->color);
+            rect->data[1][4 * color + 3] = AA(image->color);
+            // Set pixels
+            for (int y = 0; y < image->h; y++)
+                for (int x = 0; x < image->w; x++)
+                    if (image->bitmap[y * image->stride + x] >= ALPHA_THRESHOLD)
+                    {
+                        int x_rect = image->dst_x + x - rect->x;
+                        int y_rect = image->dst_y + y - rect->y;
+                        rect->data[0][y_rect * rect->w + x_rect] = color;
+                    }
+        }
+        rect->type = SUBTITLE_BITMAP;
+
+        ass_free_track(track);
+    }
+
+    ass_renderer_done(renderer);
+    ass_library_done(library);
+}
+
 static const AVFilterPad ass_inputs[] = {
     {
         .name             = "default",
-- 
2.34.1

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".