2 * Copyright (c) 2011 Stefano Sabatini
3 * Copyright (c) 2010 S.N. Hemanth Meenakshisundaram
4 * Copyright (c) 2003 Gustavo Sverzut Barbieri <gsbarbieri@yahoo.com.br>
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 * drawtext filter, based on the original vhook/drawtext.c
26 * filter by Gustavo Sverzut Barbieri
34 #include <sys/types.h>
42 #if CONFIG_LIBFONTCONFIG
43 #include <fontconfig/fontconfig.h>
46 #include "libavutil/avstring.h"
47 #include "libavutil/bprint.h"
48 #include "libavutil/common.h"
49 #include "libavutil/file.h"
50 #include "libavutil/eval.h"
51 #include "libavutil/opt.h"
52 #include "libavutil/random_seed.h"
53 #include "libavutil/parseutils.h"
54 #include "libavutil/timecode.h"
55 #include "libavutil/tree.h"
56 #include "libavutil/lfg.h"
58 #include "drawutils.h"
68 #include FT_FREETYPE_H
72 static const char *const var_names[] = {
75 "line_h", "lh", ///< line height, same as max_glyph_h
76 "main_h", "h", "H", ///< height of the input video
77 "main_w", "w", "W", ///< width of the input video
78 "max_glyph_a", "ascent", ///< max glyph ascent
79 "max_glyph_d", "descent", ///< min glyph descent
80 "max_glyph_h", ///< max glyph height
81 "max_glyph_w", ///< max glyph width
82 "n", ///< number of frame
84 "t", ///< timestamp expressed in seconds
85 "text_h", "th", ///< height of the rendered text
86 "text_w", "tw", ///< width of the rendered text
93 static const char *const fun2_names[] = {
97 static double drand(void *opaque, double min, double max)
99 return min + (max-min) / UINT_MAX * av_lfg_get(opaque);
102 typedef double (*eval_func2)(void *, double a, double b);
104 static const eval_func2 fun2[] = {
113 VAR_MAIN_H, VAR_h, VAR_H,
114 VAR_MAIN_W, VAR_w, VAR_W,
115 VAR_MAX_GLYPH_A, VAR_ASCENT,
116 VAR_MAX_GLYPH_D, VAR_DESCENT,
130 enum expansion_mode {
136 typedef struct DrawTextContext {
137 const AVClass *class;
138 enum expansion_mode exp_mode; ///< expansion mode to use for the text
139 int reinit; ///< tells if the filter is being reinited
140 #if CONFIG_LIBFONTCONFIG
141 uint8_t *font; ///< font to be used
143 uint8_t *fontfile; ///< font to be used
144 uint8_t *text; ///< text to be drawn
145 AVBPrint expanded_text; ///< used to contain the expanded text
146 uint8_t *fontcolor_expr; ///< fontcolor expression to evaluate
147 AVBPrint expanded_fontcolor; ///< used to contain the expanded fontcolor spec
148 int ft_load_flags; ///< flags used for loading fonts, see FT_LOAD_*
149 FT_Vector *positions; ///< positions for each element in the text
150 size_t nb_positions; ///< number of elements of positions array
151 char *textfile; ///< file with text to be drawn
152 int x; ///< x position to start drawing text
153 int y; ///< y position to start drawing text
154 int max_glyph_w; ///< max glyph width
155 int max_glyph_h; ///< max glyph height
156 int shadowx, shadowy;
157 int borderw; ///< border width
158 unsigned int fontsize; ///< font size to use
160 short int draw_box; ///< draw box around text - true or false
161 int use_kerning; ///< font kerning is used - true/false
162 int tabsize; ///< tab size
163 int fix_bounds; ///< do we let it go out of frame bounds - t/f
166 FFDrawColor fontcolor; ///< foreground color
167 FFDrawColor shadowcolor; ///< shadow color
168 FFDrawColor bordercolor; ///< border color
169 FFDrawColor boxcolor; ///< background color
171 FT_Library library; ///< freetype font library handle
172 FT_Face face; ///< freetype font face handle
173 FT_Stroker stroker; ///< freetype stroker handle
174 struct AVTreeNode *glyphs; ///< rendered glyphs, stored using the UTF-32 char code
175 char *x_expr; ///< expression for x position
176 char *y_expr; ///< expression for y position
177 AVExpr *x_pexpr, *y_pexpr; ///< parsed expressions for x and y
178 int64_t basetime; ///< base pts time in the real world for display
179 double var_values[VAR_VARS_NB];
180 AVLFG prng; ///< random
181 char *tc_opt_string; ///< specified timecode option string
182 AVRational tc_rate; ///< frame rate for timecode
183 AVTimecode tc; ///< timecode context
184 int tc24hmax; ///< 1 if timecode is wrapped to 24 hours, 0 otherwise
185 int reload; ///< reload text file for each frame
186 int start_number; ///< starting frame number for n/frame_num var
187 #if CONFIG_LIBFRIBIDI
188 int text_shaping; ///< 1 to shape the text before drawing it
190 AVDictionary *metadata;
193 #define OFFSET(x) offsetof(DrawTextContext, x)
194 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
196 static const AVOption drawtext_options[]= {
197 {"fontfile", "set font file", OFFSET(fontfile), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
198 {"text", "set text", OFFSET(text), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
199 {"textfile", "set text file", OFFSET(textfile), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
200 {"fontcolor", "set foreground color", OFFSET(fontcolor.rgba), AV_OPT_TYPE_COLOR, {.str="black"}, CHAR_MIN, CHAR_MAX, FLAGS},
201 {"fontcolor_expr", "set foreground color expression", OFFSET(fontcolor_expr), AV_OPT_TYPE_STRING, {.str=""}, CHAR_MIN, CHAR_MAX, FLAGS},
202 {"boxcolor", "set box color", OFFSET(boxcolor.rgba), AV_OPT_TYPE_COLOR, {.str="white"}, CHAR_MIN, CHAR_MAX, FLAGS},
203 {"bordercolor", "set border color", OFFSET(bordercolor.rgba), AV_OPT_TYPE_COLOR, {.str="black"}, CHAR_MIN, CHAR_MAX, FLAGS},
204 {"shadowcolor", "set shadow color", OFFSET(shadowcolor.rgba), AV_OPT_TYPE_COLOR, {.str="black"}, CHAR_MIN, CHAR_MAX, FLAGS},
205 {"box", "set box", OFFSET(draw_box), AV_OPT_TYPE_INT, {.i64=0}, 0, 1 , FLAGS},
206 {"fontsize", "set font size", OFFSET(fontsize), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX , FLAGS},
207 {"x", "set x expression", OFFSET(x_expr), AV_OPT_TYPE_STRING, {.str="0"}, CHAR_MIN, CHAR_MAX, FLAGS},
208 {"y", "set y expression", OFFSET(y_expr), AV_OPT_TYPE_STRING, {.str="0"}, CHAR_MIN, CHAR_MAX, FLAGS},
209 {"shadowx", "set x", OFFSET(shadowx), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX , FLAGS},
210 {"shadowy", "set y", OFFSET(shadowy), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX , FLAGS},
211 {"borderw", "set border width", OFFSET(borderw), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX , FLAGS},
212 {"tabsize", "set tab size", OFFSET(tabsize), AV_OPT_TYPE_INT, {.i64=4}, 0, INT_MAX , FLAGS},
213 {"basetime", "set base time", OFFSET(basetime), AV_OPT_TYPE_INT64, {.i64=AV_NOPTS_VALUE}, INT64_MIN, INT64_MAX , FLAGS},
214 #if CONFIG_LIBFONTCONFIG
215 { "font", "Font name", OFFSET(font), AV_OPT_TYPE_STRING, { .str = "Sans" }, .flags = FLAGS },
218 {"expansion", "set the expansion mode", OFFSET(exp_mode), AV_OPT_TYPE_INT, {.i64=EXP_NORMAL}, 0, 2, FLAGS, "expansion"},
219 {"none", "set no expansion", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_NONE}, 0, 0, FLAGS, "expansion"},
220 {"normal", "set normal expansion", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_NORMAL}, 0, 0, FLAGS, "expansion"},
221 {"strftime", "set strftime expansion (deprecated)", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_STRFTIME}, 0, 0, FLAGS, "expansion"},
223 {"timecode", "set initial timecode", OFFSET(tc_opt_string), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
224 {"tc24hmax", "set 24 hours max (timecode only)", OFFSET(tc24hmax), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS},
225 {"timecode_rate", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
226 {"r", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
227 {"rate", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
228 {"reload", "reload text file for each frame", OFFSET(reload), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS},
229 {"fix_bounds", "if true, check and fix text coords to avoid clipping", OFFSET(fix_bounds), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, FLAGS},
230 {"start_number", "start frame number for n/frame_num variable", OFFSET(start_number), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS},
232 #if CONFIG_LIBFRIBIDI
233 {"text_shaping", "attempt to shape text before drawing", OFFSET(text_shaping), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, FLAGS},
236 /* FT_LOAD_* flags */
237 { "ft_load_flags", "set font loading flags for libfreetype", OFFSET(ft_load_flags), AV_OPT_TYPE_FLAGS, { .i64 = FT_LOAD_DEFAULT }, 0, INT_MAX, FLAGS, "ft_load_flags" },
238 { "default", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_DEFAULT }, .flags = FLAGS, .unit = "ft_load_flags" },
239 { "no_scale", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_SCALE }, .flags = FLAGS, .unit = "ft_load_flags" },
240 { "no_hinting", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_HINTING }, .flags = FLAGS, .unit = "ft_load_flags" },
241 { "render", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_RENDER }, .flags = FLAGS, .unit = "ft_load_flags" },
242 { "no_bitmap", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_BITMAP }, .flags = FLAGS, .unit = "ft_load_flags" },
243 { "vertical_layout", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_VERTICAL_LAYOUT }, .flags = FLAGS, .unit = "ft_load_flags" },
244 { "force_autohint", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_FORCE_AUTOHINT }, .flags = FLAGS, .unit = "ft_load_flags" },
245 { "crop_bitmap", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_CROP_BITMAP }, .flags = FLAGS, .unit = "ft_load_flags" },
246 { "pedantic", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_PEDANTIC }, .flags = FLAGS, .unit = "ft_load_flags" },
247 { "ignore_global_advance_width", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_IGNORE_GLOBAL_ADVANCE_WIDTH }, .flags = FLAGS, .unit = "ft_load_flags" },
248 { "no_recurse", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_RECURSE }, .flags = FLAGS, .unit = "ft_load_flags" },
249 { "ignore_transform", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_IGNORE_TRANSFORM }, .flags = FLAGS, .unit = "ft_load_flags" },
250 { "monochrome", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_MONOCHROME }, .flags = FLAGS, .unit = "ft_load_flags" },
251 { "linear_design", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_LINEAR_DESIGN }, .flags = FLAGS, .unit = "ft_load_flags" },
252 { "no_autohint", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_AUTOHINT }, .flags = FLAGS, .unit = "ft_load_flags" },
256 AVFILTER_DEFINE_CLASS(drawtext);
258 #undef __FTERRORS_H__
259 #define FT_ERROR_START_LIST {
260 #define FT_ERRORDEF(e, v, s) { (e), (s) },
261 #define FT_ERROR_END_LIST { 0, NULL } };
263 static const struct ft_error
270 #define FT_ERRMSG(e) ft_errors[e].err_msg
272 typedef struct Glyph {
274 FT_Glyph border_glyph;
276 FT_Bitmap bitmap; ///< array holding bitmaps of font
277 FT_Bitmap border_bitmap; ///< array holding bitmaps of font border
284 static int glyph_cmp(void *key, const void *b)
286 const Glyph *a = key, *bb = b;
287 int64_t diff = (int64_t)a->code - (int64_t)bb->code;
288 return diff > 0 ? 1 : diff < 0 ? -1 : 0;
292 * Load glyphs corresponding to the UTF-32 codepoint code.
294 static int load_glyph(AVFilterContext *ctx, Glyph **glyph_ptr, uint32_t code)
296 DrawTextContext *s = ctx->priv;
297 FT_BitmapGlyph bitmapglyph;
299 struct AVTreeNode *node = NULL;
302 /* load glyph into s->face->glyph */
303 if (FT_Load_Char(s->face, code, s->ft_load_flags))
304 return AVERROR(EINVAL);
306 glyph = av_mallocz(sizeof(*glyph));
308 ret = AVERROR(ENOMEM);
313 if (FT_Get_Glyph(s->face->glyph, &glyph->glyph)) {
314 ret = AVERROR(EINVAL);
318 glyph->border_glyph = glyph->glyph;
319 if (FT_Glyph_StrokeBorder(&glyph->border_glyph, s->stroker, 0, 0) ||
320 FT_Glyph_To_Bitmap(&glyph->border_glyph, FT_RENDER_MODE_NORMAL, 0, 1)) {
321 ret = AVERROR_EXTERNAL;
324 bitmapglyph = (FT_BitmapGlyph) glyph->border_glyph;
325 glyph->border_bitmap = bitmapglyph->bitmap;
327 if (FT_Glyph_To_Bitmap(&glyph->glyph, FT_RENDER_MODE_NORMAL, 0, 1)) {
328 ret = AVERROR_EXTERNAL;
331 bitmapglyph = (FT_BitmapGlyph) glyph->glyph;
333 glyph->bitmap = bitmapglyph->bitmap;
334 glyph->bitmap_left = bitmapglyph->left;
335 glyph->bitmap_top = bitmapglyph->top;
336 glyph->advance = s->face->glyph->advance.x >> 6;
338 /* measure text height to calculate text_height (or the maximum text height) */
339 FT_Glyph_Get_CBox(glyph->glyph, ft_glyph_bbox_pixels, &glyph->bbox);
341 /* cache the newly created glyph */
342 if (!(node = av_tree_node_alloc())) {
343 ret = AVERROR(ENOMEM);
346 av_tree_insert(&s->glyphs, glyph, glyph_cmp, &node);
354 av_freep(&glyph->glyph);
361 static int load_font_file(AVFilterContext *ctx, const char *path, int index)
363 DrawTextContext *s = ctx->priv;
366 err = FT_New_Face(s->library, path, index, &s->face);
368 av_log(ctx, AV_LOG_ERROR, "Could not load font \"%s\": %s\n",
369 s->fontfile, FT_ERRMSG(err));
370 return AVERROR(EINVAL);
375 #if CONFIG_LIBFONTCONFIG
376 static int load_font_fontconfig(AVFilterContext *ctx)
378 DrawTextContext *s = ctx->priv;
379 FcConfig *fontconfig;
380 FcPattern *pat, *best;
381 FcResult result = FcResultMatch;
385 int err = AVERROR(ENOENT);
387 fontconfig = FcInitLoadConfigAndFonts();
389 av_log(ctx, AV_LOG_ERROR, "impossible to init fontconfig\n");
390 return AVERROR_UNKNOWN;
392 pat = FcNameParse(s->fontfile ? s->fontfile :
393 (uint8_t *)(intptr_t)"default");
395 av_log(ctx, AV_LOG_ERROR, "could not parse fontconfig pat");
396 return AVERROR(EINVAL);
399 FcPatternAddString(pat, FC_FAMILY, s->font);
401 FcPatternAddDouble(pat, FC_SIZE, (double)s->fontsize);
403 FcDefaultSubstitute(pat);
405 if (!FcConfigSubstitute(fontconfig, pat, FcMatchPattern)) {
406 av_log(ctx, AV_LOG_ERROR, "could not substitue fontconfig options"); /* very unlikely */
407 FcPatternDestroy(pat);
408 return AVERROR(ENOMEM);
411 best = FcFontMatch(fontconfig, pat, &result);
412 FcPatternDestroy(pat);
414 if (!best || result != FcResultMatch) {
415 av_log(ctx, AV_LOG_ERROR,
416 "Cannot find a valid font for the family %s\n",
422 FcPatternGetInteger(best, FC_INDEX, 0, &index ) != FcResultMatch ||
423 FcPatternGetDouble (best, FC_SIZE, 0, &size ) != FcResultMatch) {
424 av_log(ctx, AV_LOG_ERROR, "impossible to find font information");
425 return AVERROR(EINVAL);
428 if (FcPatternGetString(best, FC_FILE, 0, &filename) != FcResultMatch) {
429 av_log(ctx, AV_LOG_ERROR, "No file path for %s\n",
434 av_log(ctx, AV_LOG_INFO, "Using \"%s\"\n", filename);
436 s->fontsize = size + 0.5;
438 err = load_font_file(ctx, filename, index);
441 FcConfigDestroy(fontconfig);
443 FcPatternDestroy(best);
448 static int load_font(AVFilterContext *ctx)
450 DrawTextContext *s = ctx->priv;
453 /* load the face, and set up the encoding, which is by default UTF-8 */
454 err = load_font_file(ctx, s->fontfile, 0);
457 #if CONFIG_LIBFONTCONFIG
458 err = load_font_fontconfig(ctx);
465 static int load_textfile(AVFilterContext *ctx)
467 DrawTextContext *s = ctx->priv;
473 if ((err = av_file_map(s->textfile, &textbuf, &textbuf_size, 0, ctx)) < 0) {
474 av_log(ctx, AV_LOG_ERROR,
475 "The text file '%s' could not be read or is empty\n",
480 if (!(tmp = av_realloc(s->text, textbuf_size + 1))) {
481 av_file_unmap(textbuf, textbuf_size);
482 return AVERROR(ENOMEM);
485 memcpy(s->text, textbuf, textbuf_size);
486 s->text[textbuf_size] = 0;
487 av_file_unmap(textbuf, textbuf_size);
492 static inline int is_newline(uint32_t c)
494 return c == '\n' || c == '\r' || c == '\f' || c == '\v';
497 #if CONFIG_LIBFRIBIDI
498 static int shape_text(AVFilterContext *ctx)
500 DrawTextContext *s = ctx->priv;
502 int ret = AVERROR(ENOMEM);
503 static const FriBidiFlags flags = FRIBIDI_FLAGS_DEFAULT |
504 FRIBIDI_FLAGS_ARABIC;
505 FriBidiChar *unicodestr = NULL;
507 FriBidiParType direction = FRIBIDI_PAR_LTR;
508 FriBidiStrIndex line_start = 0;
509 FriBidiStrIndex line_end = 0;
510 FriBidiLevel *embedding_levels = NULL;
511 FriBidiArabicProp *ar_props = NULL;
512 FriBidiCharType *bidi_types = NULL;
515 len = strlen(s->text);
516 if (!(unicodestr = av_malloc_array(len, sizeof(*unicodestr)))) {
519 len = fribidi_charset_to_unicode(FRIBIDI_CHAR_SET_UTF8,
520 s->text, len, unicodestr);
522 bidi_types = av_malloc_array(len, sizeof(*bidi_types));
527 fribidi_get_bidi_types(unicodestr, len, bidi_types);
529 embedding_levels = av_malloc_array(len, sizeof(*embedding_levels));
530 if (!embedding_levels) {
534 if (!fribidi_get_par_embedding_levels(bidi_types, len, &direction,
539 ar_props = av_malloc_array(len, sizeof(*ar_props));
544 fribidi_get_joining_types(unicodestr, len, ar_props);
545 fribidi_join_arabic(bidi_types, len, embedding_levels, ar_props);
546 fribidi_shape(flags, embedding_levels, len, ar_props, unicodestr);
548 for (line_end = 0, line_start = 0; line_end < len; line_end++) {
549 if (is_newline(unicodestr[line_end]) || line_end == len - 1) {
550 if (!fribidi_reorder_line(flags, bidi_types,
551 line_end - line_start + 1, line_start,
552 direction, embedding_levels, unicodestr,
556 line_start = line_end + 1;
560 /* Remove zero-width fill chars put in by libfribidi */
561 for (i = 0, j = 0; i < len; i++)
562 if (unicodestr[i] != FRIBIDI_CHAR_FILL)
563 unicodestr[j++] = unicodestr[i];
566 if (!(tmp = av_realloc(s->text, (len * 4 + 1) * sizeof(*s->text)))) {
567 /* Use len * 4, as a unicode character can be up to 4 bytes in UTF-8 */
572 len = fribidi_unicode_to_charset(FRIBIDI_CHAR_SET_UTF8,
573 unicodestr, len, s->text);
578 av_free(embedding_levels);
585 static av_cold int init(AVFilterContext *ctx)
588 DrawTextContext *s = ctx->priv;
591 if (!s->fontfile && !CONFIG_LIBFONTCONFIG) {
592 av_log(ctx, AV_LOG_ERROR, "No font filename provided\n");
593 return AVERROR(EINVAL);
598 av_log(ctx, AV_LOG_ERROR,
599 "Both text and text file provided. Please provide only one\n");
600 return AVERROR(EINVAL);
602 if ((err = load_textfile(ctx)) < 0)
606 #if CONFIG_LIBFRIBIDI
608 if ((err = shape_text(ctx)) < 0)
612 if (s->reload && !s->textfile)
613 av_log(ctx, AV_LOG_WARNING, "No file to reload\n");
615 if (s->tc_opt_string) {
616 int ret = av_timecode_init_from_string(&s->tc, s->tc_rate,
617 s->tc_opt_string, ctx);
621 s->tc.flags |= AV_TIMECODE_FLAG_24HOURSMAX;
623 s->text = av_strdup("");
627 av_log(ctx, AV_LOG_ERROR,
628 "Either text, a valid file or a timecode must be provided\n");
629 return AVERROR(EINVAL);
632 if ((err = FT_Init_FreeType(&(s->library)))) {
633 av_log(ctx, AV_LOG_ERROR,
634 "Could not load FreeType: %s\n", FT_ERRMSG(err));
635 return AVERROR(EINVAL);
638 err = load_font(ctx);
643 if ((err = FT_Set_Pixel_Sizes(s->face, 0, s->fontsize))) {
644 av_log(ctx, AV_LOG_ERROR, "Could not set font size to %d pixels: %s\n",
645 s->fontsize, FT_ERRMSG(err));
646 return AVERROR(EINVAL);
650 if (FT_Stroker_New(s->library, &s->stroker)) {
651 av_log(ctx, AV_LOG_ERROR, "Coult not init FT stroker\n");
652 return AVERROR_EXTERNAL;
654 FT_Stroker_Set(s->stroker, s->borderw << 6, FT_STROKER_LINECAP_ROUND,
655 FT_STROKER_LINEJOIN_ROUND, 0);
658 s->use_kerning = FT_HAS_KERNING(s->face);
660 /* load the fallback glyph with code 0 */
661 load_glyph(ctx, NULL, 0);
663 /* set the tabsize in pixels */
664 if ((err = load_glyph(ctx, &glyph, ' ')) < 0) {
665 av_log(ctx, AV_LOG_ERROR, "Could not set tabsize.\n");
668 s->tabsize *= glyph->advance;
670 if (s->exp_mode == EXP_STRFTIME &&
671 (strchr(s->text, '%') || strchr(s->text, '\\')))
672 av_log(ctx, AV_LOG_WARNING, "expansion=strftime is deprecated.\n");
674 av_bprint_init(&s->expanded_text, 0, AV_BPRINT_SIZE_UNLIMITED);
675 av_bprint_init(&s->expanded_fontcolor, 0, AV_BPRINT_SIZE_UNLIMITED);
680 static int query_formats(AVFilterContext *ctx)
682 ff_set_common_formats(ctx, ff_draw_supported_pixel_formats(0));
686 static int glyph_enu_free(void *opaque, void *elem)
690 FT_Done_Glyph(glyph->glyph);
691 FT_Done_Glyph(glyph->border_glyph);
696 static av_cold void uninit(AVFilterContext *ctx)
698 DrawTextContext *s = ctx->priv;
700 av_expr_free(s->x_pexpr);
701 av_expr_free(s->y_pexpr);
702 s->x_pexpr = s->y_pexpr = NULL;
703 av_freep(&s->positions);
707 av_tree_enumerate(s->glyphs, NULL, NULL, glyph_enu_free);
708 av_tree_destroy(s->glyphs);
711 FT_Done_Face(s->face);
712 FT_Stroker_Done(s->stroker);
713 FT_Done_FreeType(s->library);
715 av_bprint_finalize(&s->expanded_text, NULL);
716 av_bprint_finalize(&s->expanded_fontcolor, NULL);
719 static int config_input(AVFilterLink *inlink)
721 AVFilterContext *ctx = inlink->dst;
722 DrawTextContext *s = ctx->priv;
725 ff_draw_init(&s->dc, inlink->format, 0);
726 ff_draw_color(&s->dc, &s->fontcolor, s->fontcolor.rgba);
727 ff_draw_color(&s->dc, &s->shadowcolor, s->shadowcolor.rgba);
728 ff_draw_color(&s->dc, &s->bordercolor, s->bordercolor.rgba);
729 ff_draw_color(&s->dc, &s->boxcolor, s->boxcolor.rgba);
731 s->var_values[VAR_w] = s->var_values[VAR_W] = s->var_values[VAR_MAIN_W] = inlink->w;
732 s->var_values[VAR_h] = s->var_values[VAR_H] = s->var_values[VAR_MAIN_H] = inlink->h;
733 s->var_values[VAR_SAR] = inlink->sample_aspect_ratio.num ? av_q2d(inlink->sample_aspect_ratio) : 1;
734 s->var_values[VAR_DAR] = (double)inlink->w / inlink->h * s->var_values[VAR_SAR];
735 s->var_values[VAR_HSUB] = 1 << s->dc.hsub_max;
736 s->var_values[VAR_VSUB] = 1 << s->dc.vsub_max;
737 s->var_values[VAR_X] = NAN;
738 s->var_values[VAR_Y] = NAN;
739 s->var_values[VAR_T] = NAN;
741 av_lfg_init(&s->prng, av_get_random_seed());
743 av_expr_free(s->x_pexpr);
744 av_expr_free(s->y_pexpr);
745 s->x_pexpr = s->y_pexpr = NULL;
747 if ((ret = av_expr_parse(&s->x_pexpr, s->x_expr, var_names,
748 NULL, NULL, fun2_names, fun2, 0, ctx)) < 0 ||
749 (ret = av_expr_parse(&s->y_pexpr, s->y_expr, var_names,
750 NULL, NULL, fun2_names, fun2, 0, ctx)) < 0)
752 return AVERROR(EINVAL);
757 static int command(AVFilterContext *ctx, const char *cmd, const char *arg, char *res, int res_len, int flags)
759 DrawTextContext *s = ctx->priv;
761 if (!strcmp(cmd, "reinit")) {
765 if ((ret = av_set_options_string(ctx, arg, "=", ":")) < 0)
767 if ((ret = init(ctx)) < 0)
769 return config_input(ctx->inputs[0]);
772 return AVERROR(ENOSYS);
775 static int func_pict_type(AVFilterContext *ctx, AVBPrint *bp,
776 char *fct, unsigned argc, char **argv, int tag)
778 DrawTextContext *s = ctx->priv;
780 av_bprintf(bp, "%c", av_get_picture_type_char(s->var_values[VAR_PICT_TYPE]));
784 static int func_pts(AVFilterContext *ctx, AVBPrint *bp,
785 char *fct, unsigned argc, char **argv, int tag)
787 DrawTextContext *s = ctx->priv;
789 double pts = s->var_values[VAR_T];
792 fmt = argc >= 1 ? argv[0] : "flt";
795 if ((ret = av_parse_time(&delta, argv[1], 1)) < 0) {
796 av_log(ctx, AV_LOG_ERROR, "Invalid delta '%s'\n", argv[1]);
799 pts += (double)delta / AV_TIME_BASE;
801 if (!strcmp(fmt, "flt")) {
802 av_bprintf(bp, "%.6f", s->var_values[VAR_T]);
803 } else if (!strcmp(fmt, "hms")) {
805 av_bprintf(bp, " ??:??:??.???");
807 int64_t ms = round(pts * 1000);
813 av_bprintf(bp, "%c%02d:%02d:%02d.%03d", sign,
814 (int)(ms / (60 * 60 * 1000)),
815 (int)(ms / (60 * 1000)) % 60,
816 (int)(ms / 1000) % 60,
820 av_log(ctx, AV_LOG_ERROR, "Invalid format '%s'\n", fmt);
821 return AVERROR(EINVAL);
826 static int func_frame_num(AVFilterContext *ctx, AVBPrint *bp,
827 char *fct, unsigned argc, char **argv, int tag)
829 DrawTextContext *s = ctx->priv;
831 av_bprintf(bp, "%d", (int)s->var_values[VAR_N]);
835 static int func_metadata(AVFilterContext *ctx, AVBPrint *bp,
836 char *fct, unsigned argc, char **argv, int tag)
838 DrawTextContext *s = ctx->priv;
839 AVDictionaryEntry *e = av_dict_get(s->metadata, argv[0], NULL, 0);
842 av_bprintf(bp, "%s", e->value);
846 #if !HAVE_LOCALTIME_R
847 static void localtime_r(const time_t *t, struct tm *tm)
853 static int func_strftime(AVFilterContext *ctx, AVBPrint *bp,
854 char *fct, unsigned argc, char **argv, int tag)
856 const char *fmt = argc ? argv[0] : "%Y-%m-%d %H:%M:%S";
862 localtime_r(&now, &tm);
865 av_bprint_strftime(bp, fmt, &tm);
869 static int func_eval_expr(AVFilterContext *ctx, AVBPrint *bp,
870 char *fct, unsigned argc, char **argv, int tag)
872 DrawTextContext *s = ctx->priv;
876 ret = av_expr_parse_and_eval(&res, argv[0], var_names, s->var_values,
877 NULL, NULL, fun2_names, fun2,
880 av_log(ctx, AV_LOG_ERROR,
881 "Expression '%s' for the expr text expansion function is not valid\n",
884 av_bprintf(bp, "%f", res);
889 static int func_eval_expr_int_format(AVFilterContext *ctx, AVBPrint *bp,
890 char *fct, unsigned argc, char **argv, int tag)
892 DrawTextContext *s = ctx->priv;
896 unsigned int positions = 0;
897 char fmt_str[30] = "%";
900 * argv[0] expression to be converted to `int`
901 * argv[1] format: 'x', 'X', 'd' or 'u'
902 * argv[2] positions printed (optional)
905 ret = av_expr_parse_and_eval(&res, argv[0], var_names, s->var_values,
906 NULL, NULL, fun2_names, fun2,
909 av_log(ctx, AV_LOG_ERROR,
910 "Expression '%s' for the expr text expansion function is not valid\n",
915 if (!strchr("xXdu", argv[1][0])) {
916 av_log(ctx, AV_LOG_ERROR, "Invalid format '%c' specified,"
917 " allowed values: 'x', 'X', 'd', 'u'\n", argv[1][0]);
918 return AVERROR(EINVAL);
922 ret = sscanf(argv[2], "%u", &positions);
924 av_log(ctx, AV_LOG_ERROR, "expr_int_format(): Invalid number of positions"
925 " to print: '%s'\n", argv[2]);
926 return AVERROR(EINVAL);
930 feclearexcept(FE_ALL_EXCEPT);
932 if ((ret = fetestexcept(FE_INVALID|FE_OVERFLOW|FE_UNDERFLOW))) {
933 av_log(ctx, AV_LOG_ERROR, "Conversion of floating-point result to int failed. Control register: 0x%08x. Conversion result: %d\n", ret, intval);
934 return AVERROR(EINVAL);
938 av_strlcatf(fmt_str, sizeof(fmt_str), "0%u", positions);
939 av_strlcatf(fmt_str, sizeof(fmt_str), "%c", argv[1][0]);
941 av_log(ctx, AV_LOG_DEBUG, "Formatting value %f (expr '%s') with spec '%s'\n",
942 res, argv[0], fmt_str);
944 av_bprintf(bp, fmt_str, intval);
949 static const struct drawtext_function {
951 unsigned argc_min, argc_max;
952 int tag; /**< opaque argument to func */
953 int (*func)(AVFilterContext *, AVBPrint *, char *, unsigned, char **, int);
955 { "expr", 1, 1, 0, func_eval_expr },
956 { "e", 1, 1, 0, func_eval_expr },
957 { "expr_int_format", 2, 3, 0, func_eval_expr_int_format },
958 { "eif", 2, 3, 0, func_eval_expr_int_format },
959 { "pict_type", 0, 0, 0, func_pict_type },
960 { "pts", 0, 2, 0, func_pts },
961 { "gmtime", 0, 1, 'G', func_strftime },
962 { "localtime", 0, 1, 'L', func_strftime },
963 { "frame_num", 0, 0, 0, func_frame_num },
964 { "n", 0, 0, 0, func_frame_num },
965 { "metadata", 1, 1, 0, func_metadata },
968 static int eval_function(AVFilterContext *ctx, AVBPrint *bp, char *fct,
969 unsigned argc, char **argv)
973 for (i = 0; i < FF_ARRAY_ELEMS(functions); i++) {
974 if (strcmp(fct, functions[i].name))
976 if (argc < functions[i].argc_min) {
977 av_log(ctx, AV_LOG_ERROR, "%%{%s} requires at least %d arguments\n",
978 fct, functions[i].argc_min);
979 return AVERROR(EINVAL);
981 if (argc > functions[i].argc_max) {
982 av_log(ctx, AV_LOG_ERROR, "%%{%s} requires at most %d arguments\n",
983 fct, functions[i].argc_max);
984 return AVERROR(EINVAL);
988 if (i >= FF_ARRAY_ELEMS(functions)) {
989 av_log(ctx, AV_LOG_ERROR, "%%{%s} is not known\n", fct);
990 return AVERROR(EINVAL);
992 return functions[i].func(ctx, bp, fct, argc, argv, functions[i].tag);
995 static int expand_function(AVFilterContext *ctx, AVBPrint *bp, char **rtext)
997 const char *text = *rtext;
998 char *argv[16] = { NULL };
999 unsigned argc = 0, i;
1003 av_log(ctx, AV_LOG_ERROR, "Stray %% near '%s'\n", text);
1004 return AVERROR(EINVAL);
1008 if (!(argv[argc++] = av_get_token(&text, ":}"))) {
1009 ret = AVERROR(ENOMEM);
1013 av_log(ctx, AV_LOG_ERROR, "Unterminated %%{} near '%s'\n", *rtext);
1014 ret = AVERROR(EINVAL);
1017 if (argc == FF_ARRAY_ELEMS(argv))
1018 av_freep(&argv[--argc]); /* error will be caught later */
1024 if ((ret = eval_function(ctx, bp, argv[0], argc - 1, argv + 1)) < 0)
1027 *rtext = (char *)text + 1;
1030 for (i = 0; i < argc; i++)
1035 static int expand_text(AVFilterContext *ctx, char *text, AVBPrint *bp)
1039 av_bprint_clear(bp);
1041 if (*text == '\\' && text[1]) {
1042 av_bprint_chars(bp, text[1], 1);
1044 } else if (*text == '%') {
1046 if ((ret = expand_function(ctx, bp, &text)) < 0)
1049 av_bprint_chars(bp, *text, 1);
1053 if (!av_bprint_is_complete(bp))
1054 return AVERROR(ENOMEM);
1058 static int draw_glyphs(DrawTextContext *s, AVFrame *frame,
1059 int width, int height,
1060 FFDrawColor *color, int x, int y, int borderw)
1062 char *text = s->expanded_text.str;
1066 Glyph *glyph = NULL;
1068 for (i = 0, p = text; *p; i++) {
1070 Glyph dummy = { 0 };
1071 GET_UTF8(code, *p++, continue;);
1073 /* skip new line chars, just go to new line */
1074 if (code == '\n' || code == '\r' || code == '\t')
1078 glyph = av_tree_find(s->glyphs, &dummy, (void *)glyph_cmp, NULL);
1080 bitmap = borderw ? glyph->border_bitmap : glyph->bitmap;
1082 if (glyph->bitmap.pixel_mode != FT_PIXEL_MODE_MONO &&
1083 glyph->bitmap.pixel_mode != FT_PIXEL_MODE_GRAY)
1084 return AVERROR(EINVAL);
1086 x1 = s->positions[i].x+s->x+x - borderw;
1087 y1 = s->positions[i].y+s->y+y - borderw;
1089 ff_blend_mask(&s->dc, color,
1090 frame->data, frame->linesize, width, height,
1091 bitmap.buffer, bitmap.pitch,
1092 bitmap.width, bitmap.rows,
1093 bitmap.pixel_mode == FT_PIXEL_MODE_MONO ? 0 : 3,
1100 static int draw_text(AVFilterContext *ctx, AVFrame *frame,
1101 int width, int height)
1103 DrawTextContext *s = ctx->priv;
1104 AVFilterLink *inlink = ctx->inputs[0];
1106 uint32_t code = 0, prev_code = 0;
1107 int x = 0, y = 0, i = 0, ret;
1108 int max_text_line_w = 0, len;
1112 int y_min = 32000, y_max = -32000;
1113 int x_min = 32000, x_max = -32000;
1115 Glyph *glyph = NULL, *prev_glyph = NULL;
1116 Glyph dummy = { 0 };
1118 time_t now = time(0);
1120 AVBPrint *bp = &s->expanded_text;
1122 av_bprint_clear(bp);
1124 if(s->basetime != AV_NOPTS_VALUE)
1125 now= frame->pts*av_q2d(ctx->inputs[0]->time_base) + s->basetime/1000000;
1127 switch (s->exp_mode) {
1129 av_bprintf(bp, "%s", s->text);
1132 if ((ret = expand_text(ctx, s->text, &s->expanded_text)) < 0)
1136 localtime_r(&now, <ime);
1137 av_bprint_strftime(bp, s->text, <ime);
1141 if (s->tc_opt_string) {
1142 char tcbuf[AV_TIMECODE_STR_SIZE];
1143 av_timecode_make_string(&s->tc, tcbuf, inlink->frame_count);
1144 av_bprint_clear(bp);
1145 av_bprintf(bp, "%s%s", s->text, tcbuf);
1148 if (!av_bprint_is_complete(bp))
1149 return AVERROR(ENOMEM);
1150 text = s->expanded_text.str;
1151 if ((len = s->expanded_text.len) > s->nb_positions) {
1152 if (!(s->positions =
1153 av_realloc(s->positions, len*sizeof(*s->positions))))
1154 return AVERROR(ENOMEM);
1155 s->nb_positions = len;
1158 if (s->fontcolor_expr[0]) {
1159 /* If expression is set, evaluate and replace the static value */
1160 av_bprint_clear(&s->expanded_fontcolor);
1161 if ((ret = expand_text(ctx, s->fontcolor_expr, &s->expanded_fontcolor)) < 0)
1163 if (!av_bprint_is_complete(&s->expanded_fontcolor))
1164 return AVERROR(ENOMEM);
1165 av_log(s, AV_LOG_DEBUG, "Evaluated fontcolor is '%s'\n", s->expanded_fontcolor.str);
1166 ret = av_parse_color(s->fontcolor.rgba, s->expanded_fontcolor.str, -1, s);
1169 ff_draw_color(&s->dc, &s->fontcolor, s->fontcolor.rgba);
1175 /* load and cache glyphs */
1176 for (i = 0, p = text; *p; i++) {
1177 GET_UTF8(code, *p++, continue;);
1181 glyph = av_tree_find(s->glyphs, &dummy, glyph_cmp, NULL);
1183 load_glyph(ctx, &glyph, code);
1186 y_min = FFMIN(glyph->bbox.yMin, y_min);
1187 y_max = FFMAX(glyph->bbox.yMax, y_max);
1188 x_min = FFMIN(glyph->bbox.xMin, x_min);
1189 x_max = FFMAX(glyph->bbox.xMax, x_max);
1191 s->max_glyph_h = y_max - y_min;
1192 s->max_glyph_w = x_max - x_min;
1194 /* compute and save position for each glyph */
1196 for (i = 0, p = text; *p; i++) {
1197 GET_UTF8(code, *p++, continue;);
1199 /* skip the \n in the sequence \r\n */
1200 if (prev_code == '\r' && code == '\n')
1204 if (is_newline(code)) {
1206 max_text_line_w = FFMAX(max_text_line_w, x);
1207 y += s->max_glyph_h;
1215 glyph = av_tree_find(s->glyphs, &dummy, glyph_cmp, NULL);
1218 if (s->use_kerning && prev_glyph && glyph->code) {
1219 FT_Get_Kerning(s->face, prev_glyph->code, glyph->code,
1220 ft_kerning_default, &delta);
1225 s->positions[i].x = x + glyph->bitmap_left;
1226 s->positions[i].y = y - glyph->bitmap_top + y_max;
1227 if (code == '\t') x = (x / s->tabsize + 1)*s->tabsize;
1228 else x += glyph->advance;
1231 max_text_line_w = FFMAX(x, max_text_line_w);
1233 s->var_values[VAR_TW] = s->var_values[VAR_TEXT_W] = max_text_line_w;
1234 s->var_values[VAR_TH] = s->var_values[VAR_TEXT_H] = y + s->max_glyph_h;
1236 s->var_values[VAR_MAX_GLYPH_W] = s->max_glyph_w;
1237 s->var_values[VAR_MAX_GLYPH_H] = s->max_glyph_h;
1238 s->var_values[VAR_MAX_GLYPH_A] = s->var_values[VAR_ASCENT ] = y_max;
1239 s->var_values[VAR_MAX_GLYPH_D] = s->var_values[VAR_DESCENT] = y_min;
1241 s->var_values[VAR_LINE_H] = s->var_values[VAR_LH] = s->max_glyph_h;
1243 s->x = s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, &s->prng);
1244 s->y = s->var_values[VAR_Y] = av_expr_eval(s->y_pexpr, s->var_values, &s->prng);
1245 s->x = s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, &s->prng);
1247 box_w = FFMIN(width - 1 , max_text_line_w);
1248 box_h = FFMIN(height - 1, y + s->max_glyph_h);
1252 ff_blend_rectangle(&s->dc, &s->boxcolor,
1253 frame->data, frame->linesize, width, height,
1254 s->x, s->y, box_w, box_h);
1256 if (s->shadowx || s->shadowy) {
1257 if ((ret = draw_glyphs(s, frame, width, height,
1258 &s->shadowcolor, s->shadowx, s->shadowy, 0)) < 0)
1263 if ((ret = draw_glyphs(s, frame, width, height,
1264 &s->bordercolor, 0, 0, s->borderw)) < 0)
1267 if ((ret = draw_glyphs(s, frame, width, height,
1268 &s->fontcolor, 0, 0, 0)) < 0)
1274 static int filter_frame(AVFilterLink *inlink, AVFrame *frame)
1276 AVFilterContext *ctx = inlink->dst;
1277 AVFilterLink *outlink = ctx->outputs[0];
1278 DrawTextContext *s = ctx->priv;
1282 if ((ret = load_textfile(ctx)) < 0)
1284 #if CONFIG_LIBFRIBIDI
1285 if (s->text_shaping)
1286 if ((ret = shape_text(ctx)) < 0)
1291 s->var_values[VAR_N] = inlink->frame_count+s->start_number;
1292 s->var_values[VAR_T] = frame->pts == AV_NOPTS_VALUE ?
1293 NAN : frame->pts * av_q2d(inlink->time_base);
1295 s->var_values[VAR_PICT_TYPE] = frame->pict_type;
1296 s->metadata = av_frame_get_metadata(frame);
1298 draw_text(ctx, frame, frame->width, frame->height);
1300 av_log(ctx, AV_LOG_DEBUG, "n:%d t:%f text_w:%d text_h:%d x:%d y:%d\n",
1301 (int)s->var_values[VAR_N], s->var_values[VAR_T],
1302 (int)s->var_values[VAR_TEXT_W], (int)s->var_values[VAR_TEXT_H],
1305 return ff_filter_frame(outlink, frame);
1308 static const AVFilterPad avfilter_vf_drawtext_inputs[] = {
1311 .type = AVMEDIA_TYPE_VIDEO,
1312 .filter_frame = filter_frame,
1313 .config_props = config_input,
1314 .needs_writable = 1,
1319 static const AVFilterPad avfilter_vf_drawtext_outputs[] = {
1322 .type = AVMEDIA_TYPE_VIDEO,
1327 AVFilter ff_vf_drawtext = {
1329 .description = NULL_IF_CONFIG_SMALL("Draw text on top of video frames using libfreetype library."),
1330 .priv_size = sizeof(DrawTextContext),
1331 .priv_class = &drawtext_class,
1334 .query_formats = query_formats,
1335 .inputs = avfilter_vf_drawtext_inputs,
1336 .outputs = avfilter_vf_drawtext_outputs,
1337 .process_command = command,
1338 .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,