FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
vf_drawtext.c
Go to the documentation of this file.
1 /*
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>
5  *
6  * This file is part of FFmpeg.
7  *
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.
12  *
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.
17  *
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
21  */
22 
23 /**
24  * @file
25  * drawtext filter, based on the original vhook/drawtext.c
26  * filter by Gustavo Sverzut Barbieri
27  */
28 
29 #include <sys/time.h>
30 #include <time.h>
31 
32 #include "config.h"
33 #include "libavutil/avstring.h"
34 #include "libavutil/bprint.h"
35 #include "libavutil/common.h"
36 #include "libavutil/file.h"
37 #include "libavutil/eval.h"
38 #include "libavutil/opt.h"
39 #include "libavutil/random_seed.h"
40 #include "libavutil/parseutils.h"
41 #include "libavutil/timecode.h"
42 #include "libavutil/tree.h"
43 #include "libavutil/lfg.h"
44 #include "avfilter.h"
45 #include "drawutils.h"
46 #include "formats.h"
47 #include "internal.h"
48 #include "video.h"
49 
50 #include <ft2build.h>
51 #include FT_FREETYPE_H
52 #include FT_GLYPH_H
53 #if CONFIG_FONTCONFIG
54 #include <fontconfig/fontconfig.h>
55 #endif
56 
57 static const char *const var_names[] = {
58  "dar",
59  "hsub", "vsub",
60  "line_h", "lh", ///< line height, same as max_glyph_h
61  "main_h", "h", "H", ///< height of the input video
62  "main_w", "w", "W", ///< width of the input video
63  "max_glyph_a", "ascent", ///< max glyph ascent
64  "max_glyph_d", "descent", ///< min glyph descent
65  "max_glyph_h", ///< max glyph height
66  "max_glyph_w", ///< max glyph width
67  "n", ///< number of frame
68  "sar",
69  "t", ///< timestamp expressed in seconds
70  "text_h", "th", ///< height of the rendered text
71  "text_w", "tw", ///< width of the rendered text
72  "x",
73  "y",
74  "pict_type",
75  NULL
76 };
77 
78 static const char *const fun2_names[] = {
79  "rand"
80 };
81 
82 static double drand(void *opaque, double min, double max)
83 {
84  return min + (max-min) / UINT_MAX * av_lfg_get(opaque);
85 }
86 
87 typedef double (*eval_func2)(void *, double a, double b);
88 
89 static const eval_func2 fun2[] = {
90  drand,
91  NULL
92 };
93 
94 enum var_name {
113 };
114 
119 };
120 
121 typedef struct {
122  const AVClass *class;
123  enum expansion_mode exp_mode; ///< expansion mode to use for the text
124  int reinit; ///< tells if the filter is being reinited
125  uint8_t *fontfile; ///< font to be used
126  uint8_t *text; ///< text to be drawn
127  AVBPrint expanded_text; ///< used to contain the expanded text
128  int ft_load_flags; ///< flags used for loading fonts, see FT_LOAD_*
129  FT_Vector *positions; ///< positions for each element in the text
130  size_t nb_positions; ///< number of elements of positions array
131  char *textfile; ///< file with text to be drawn
132  int x; ///< x position to start drawing text
133  int y; ///< y position to start drawing text
134  int max_glyph_w; ///< max glyph width
135  int max_glyph_h; ///< max glyph height
136  int shadowx, shadowy;
137  unsigned int fontsize; ///< font size to use
138 
139  short int draw_box; ///< draw box around text - true or false
140  int use_kerning; ///< font kerning is used - true/false
141  int tabsize; ///< tab size
142  int fix_bounds; ///< do we let it go out of frame bounds - t/f
143 
145  FFDrawColor fontcolor; ///< foreground color
146  FFDrawColor shadowcolor; ///< shadow color
147  FFDrawColor boxcolor; ///< background color
148 
149  FT_Library library; ///< freetype font library handle
150  FT_Face face; ///< freetype font face handle
151  struct AVTreeNode *glyphs; ///< rendered glyphs, stored using the UTF-32 char code
152  char *x_expr; ///< expression for x position
153  char *y_expr; ///< expression for y position
154  AVExpr *x_pexpr, *y_pexpr; ///< parsed expressions for x and y
155  int64_t basetime; ///< base pts time in the real world for display
156  double var_values[VAR_VARS_NB];
157 #if FF_API_DRAWTEXT_OLD_TIMELINE
158  char *draw_expr; ///< expression for draw
159  AVExpr *draw_pexpr; ///< parsed expression for draw
160  int draw; ///< set to zero to prevent drawing
161 #endif
162  AVLFG prng; ///< random
163  char *tc_opt_string; ///< specified timecode option string
164  AVRational tc_rate; ///< frame rate for timecode
165  AVTimecode tc; ///< timecode context
166  int tc24hmax; ///< 1 if timecode is wrapped to 24 hours, 0 otherwise
167  int reload; ///< reload text file for each frame
168  int start_number; ///< starting frame number for n/frame_num var
171 
172 #define OFFSET(x) offsetof(DrawTextContext, x)
173 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
174 
175 static const AVOption drawtext_options[]= {
176  {"fontfile", "set font file", OFFSET(fontfile), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
177  {"text", "set text", OFFSET(text), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
178  {"textfile", "set text file", OFFSET(textfile), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
179  {"fontcolor", "set foreground color", OFFSET(fontcolor.rgba), AV_OPT_TYPE_COLOR, {.str="black"}, CHAR_MIN, CHAR_MAX, FLAGS},
180  {"boxcolor", "set box color", OFFSET(boxcolor.rgba), AV_OPT_TYPE_COLOR, {.str="white"}, CHAR_MIN, CHAR_MAX, FLAGS},
181  {"shadowcolor", "set shadow color", OFFSET(shadowcolor.rgba), AV_OPT_TYPE_COLOR, {.str="black"}, CHAR_MIN, CHAR_MAX, FLAGS},
182  {"box", "set box", OFFSET(draw_box), AV_OPT_TYPE_INT, {.i64=0}, 0, 1 , FLAGS},
183  {"fontsize", "set font size", OFFSET(fontsize), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX , FLAGS},
184  {"x", "set x expression", OFFSET(x_expr), AV_OPT_TYPE_STRING, {.str="0"}, CHAR_MIN, CHAR_MAX, FLAGS},
185  {"y", "set y expression", OFFSET(y_expr), AV_OPT_TYPE_STRING, {.str="0"}, CHAR_MIN, CHAR_MAX, FLAGS},
186  {"shadowx", "set x", OFFSET(shadowx), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX , FLAGS},
187  {"shadowy", "set y", OFFSET(shadowy), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX , FLAGS},
188  {"tabsize", "set tab size", OFFSET(tabsize), AV_OPT_TYPE_INT, {.i64=4}, 0, INT_MAX , FLAGS},
189  {"basetime", "set base time", OFFSET(basetime), AV_OPT_TYPE_INT64, {.i64=AV_NOPTS_VALUE}, INT64_MIN, INT64_MAX , FLAGS},
190 #if FF_API_DRAWTEXT_OLD_TIMELINE
191  {"draw", "if false do not draw (deprecated)", OFFSET(draw_expr), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
192 #endif
193 
194  {"expansion", "set the expansion mode", OFFSET(exp_mode), AV_OPT_TYPE_INT, {.i64=EXP_NORMAL}, 0, 2, FLAGS, "expansion"},
195  {"none", "set no expansion", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_NONE}, 0, 0, FLAGS, "expansion"},
196  {"normal", "set normal expansion", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_NORMAL}, 0, 0, FLAGS, "expansion"},
197  {"strftime", "set strftime expansion (deprecated)", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_STRFTIME}, 0, 0, FLAGS, "expansion"},
198 
199  {"timecode", "set initial timecode", OFFSET(tc_opt_string), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
200  {"tc24hmax", "set 24 hours max (timecode only)", OFFSET(tc24hmax), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS},
201  {"timecode_rate", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
202  {"r", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
203  {"rate", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
204  {"reload", "reload text file for each frame", OFFSET(reload), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS},
205  {"fix_bounds", "if true, check and fix text coords to avoid clipping", OFFSET(fix_bounds), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, FLAGS},
206  {"start_number", "start frame number for n/frame_num variable", OFFSET(start_number), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS},
207 
208  /* FT_LOAD_* flags */
209  { "ft_load_flags", "set font loading flags for libfreetype", OFFSET(ft_load_flags), AV_OPT_TYPE_FLAGS, { .i64 = FT_LOAD_DEFAULT | FT_LOAD_RENDER}, 0, INT_MAX, FLAGS, "ft_load_flags" },
210  { "default", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_DEFAULT }, .flags = FLAGS, .unit = "ft_load_flags" },
211  { "no_scale", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_SCALE }, .flags = FLAGS, .unit = "ft_load_flags" },
212  { "no_hinting", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_HINTING }, .flags = FLAGS, .unit = "ft_load_flags" },
213  { "render", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_RENDER }, .flags = FLAGS, .unit = "ft_load_flags" },
214  { "no_bitmap", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_BITMAP }, .flags = FLAGS, .unit = "ft_load_flags" },
215  { "vertical_layout", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_VERTICAL_LAYOUT }, .flags = FLAGS, .unit = "ft_load_flags" },
216  { "force_autohint", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_FORCE_AUTOHINT }, .flags = FLAGS, .unit = "ft_load_flags" },
217  { "crop_bitmap", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_CROP_BITMAP }, .flags = FLAGS, .unit = "ft_load_flags" },
218  { "pedantic", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_PEDANTIC }, .flags = FLAGS, .unit = "ft_load_flags" },
219  { "ignore_global_advance_width", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_IGNORE_GLOBAL_ADVANCE_WIDTH }, .flags = FLAGS, .unit = "ft_load_flags" },
220  { "no_recurse", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_RECURSE }, .flags = FLAGS, .unit = "ft_load_flags" },
221  { "ignore_transform", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_IGNORE_TRANSFORM }, .flags = FLAGS, .unit = "ft_load_flags" },
222  { "monochrome", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_MONOCHROME }, .flags = FLAGS, .unit = "ft_load_flags" },
223  { "linear_design", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_LINEAR_DESIGN }, .flags = FLAGS, .unit = "ft_load_flags" },
224  { "no_autohint", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_AUTOHINT }, .flags = FLAGS, .unit = "ft_load_flags" },
225  { NULL }
226 };
227 
229 
230 #undef __FTERRORS_H__
231 #define FT_ERROR_START_LIST {
232 #define FT_ERRORDEF(e, v, s) { (e), (s) },
233 #define FT_ERROR_END_LIST { 0, NULL } };
234 
235 struct ft_error
236 {
237  int err;
238  const char *err_msg;
239 } static ft_errors[] =
240 #include FT_ERRORS_H
241 
242 #define FT_ERRMSG(e) ft_errors[e].err_msg
243 
244 typedef struct {
245  FT_Glyph *glyph;
246  uint32_t code;
247  FT_Bitmap bitmap; ///< array holding bitmaps of font
248  FT_BBox bbox;
249  int advance;
250  int bitmap_left;
251  int bitmap_top;
252 } Glyph;
253 
254 static int glyph_cmp(void *key, const void *b)
255 {
256  const Glyph *a = key, *bb = b;
257  int64_t diff = (int64_t)a->code - (int64_t)bb->code;
258  return diff > 0 ? 1 : diff < 0 ? -1 : 0;
259 }
260 
261 /**
262  * Load glyphs corresponding to the UTF-32 codepoint code.
263  */
264 static int load_glyph(AVFilterContext *ctx, Glyph **glyph_ptr, uint32_t code)
265 {
266  DrawTextContext *s = ctx->priv;
267  Glyph *glyph;
268  struct AVTreeNode *node = NULL;
269  int ret;
270 
271  /* load glyph into s->face->glyph */
272  if (FT_Load_Char(s->face, code, s->ft_load_flags))
273  return AVERROR(EINVAL);
274 
275  /* save glyph */
276  if (!(glyph = av_mallocz(sizeof(*glyph))) ||
277  !(glyph->glyph = av_mallocz(sizeof(*glyph->glyph)))) {
278  ret = AVERROR(ENOMEM);
279  goto error;
280  }
281  glyph->code = code;
282 
283  if (FT_Get_Glyph(s->face->glyph, glyph->glyph)) {
284  ret = AVERROR(EINVAL);
285  goto error;
286  }
287 
288  glyph->bitmap = s->face->glyph->bitmap;
289  glyph->bitmap_left = s->face->glyph->bitmap_left;
290  glyph->bitmap_top = s->face->glyph->bitmap_top;
291  glyph->advance = s->face->glyph->advance.x >> 6;
292 
293  /* measure text height to calculate text_height (or the maximum text height) */
294  FT_Glyph_Get_CBox(*glyph->glyph, ft_glyph_bbox_pixels, &glyph->bbox);
295 
296  /* cache the newly created glyph */
297  if (!(node = av_tree_node_alloc())) {
298  ret = AVERROR(ENOMEM);
299  goto error;
300  }
301  av_tree_insert(&s->glyphs, glyph, glyph_cmp, &node);
302 
303  if (glyph_ptr)
304  *glyph_ptr = glyph;
305  return 0;
306 
307 error:
308  if (glyph)
309  av_freep(&glyph->glyph);
310  av_freep(&glyph);
311  av_freep(&node);
312  return ret;
313 }
314 
315 static int load_font_file(AVFilterContext *ctx, const char *path, int index,
316  const char **error)
317 {
318  DrawTextContext *s = ctx->priv;
319  int err;
320 
321  err = FT_New_Face(s->library, path, index, &s->face);
322  if (err) {
323  *error = FT_ERRMSG(err);
324  return AVERROR(EINVAL);
325  }
326  return 0;
327 }
328 
329 #if CONFIG_FONTCONFIG
330 static int load_font_fontconfig(AVFilterContext *ctx, const char **error)
331 {
332  DrawTextContext *s = ctx->priv;
333  FcConfig *fontconfig;
334  FcPattern *pattern, *fpat;
335  FcResult result = FcResultMatch;
336  FcChar8 *filename;
337  int err, index;
338  double size;
339 
340  fontconfig = FcInitLoadConfigAndFonts();
341  if (!fontconfig) {
342  *error = "impossible to init fontconfig\n";
343  return AVERROR(EINVAL);
344  }
345  pattern = FcNameParse(s->fontfile ? s->fontfile :
346  (uint8_t *)(intptr_t)"default");
347  if (!pattern) {
348  *error = "could not parse fontconfig pattern";
349  return AVERROR(EINVAL);
350  }
351  if (!FcConfigSubstitute(fontconfig, pattern, FcMatchPattern)) {
352  *error = "could not substitue fontconfig options"; /* very unlikely */
353  return AVERROR(EINVAL);
354  }
355  FcDefaultSubstitute(pattern);
356  fpat = FcFontMatch(fontconfig, pattern, &result);
357  if (!fpat || result != FcResultMatch) {
358  *error = "impossible to find a matching font";
359  return AVERROR(EINVAL);
360  }
361  if (FcPatternGetString (fpat, FC_FILE, 0, &filename) != FcResultMatch ||
362  FcPatternGetInteger(fpat, FC_INDEX, 0, &index ) != FcResultMatch ||
363  FcPatternGetDouble (fpat, FC_SIZE, 0, &size ) != FcResultMatch) {
364  *error = "impossible to find font information";
365  return AVERROR(EINVAL);
366  }
367  av_log(ctx, AV_LOG_INFO, "Using \"%s\"\n", filename);
368  if (!s->fontsize)
369  s->fontsize = size + 0.5;
370  err = load_font_file(ctx, filename, index, error);
371  if (err)
372  return err;
373  FcPatternDestroy(fpat);
374  FcPatternDestroy(pattern);
375  FcConfigDestroy(fontconfig);
376  return 0;
377 }
378 #endif
379 
380 static int load_font(AVFilterContext *ctx)
381 {
382  DrawTextContext *s = ctx->priv;
383  int err;
384  const char *error = "unknown error\n";
385 
386  /* load the face, and set up the encoding, which is by default UTF-8 */
387  err = load_font_file(ctx, s->fontfile, 0, &error);
388  if (!err)
389  return 0;
390 #if CONFIG_FONTCONFIG
391  err = load_font_fontconfig(ctx, &error);
392  if (!err)
393  return 0;
394 #endif
395  av_log(ctx, AV_LOG_ERROR, "Could not load font \"%s\": %s\n",
396  s->fontfile, error);
397  return err;
398 }
399 
401 {
402  DrawTextContext *s = ctx->priv;
403  int err;
404  uint8_t *textbuf;
405  size_t textbuf_size;
406 
407  if ((err = av_file_map(s->textfile, &textbuf, &textbuf_size, 0, ctx)) < 0) {
408  av_log(ctx, AV_LOG_ERROR,
409  "The text file '%s' could not be read or is empty\n",
410  s->textfile);
411  return err;
412  }
413 
414  if (!(s->text = av_realloc(s->text, textbuf_size + 1)))
415  return AVERROR(ENOMEM);
416  memcpy(s->text, textbuf, textbuf_size);
417  s->text[textbuf_size] = 0;
418  av_file_unmap(textbuf, textbuf_size);
419 
420  return 0;
421 }
422 
423 static av_cold int init(AVFilterContext *ctx)
424 {
425  int err;
426  DrawTextContext *s = ctx->priv;
427  Glyph *glyph;
428 
429 #if FF_API_DRAWTEXT_OLD_TIMELINE
430  if (s->draw_expr)
431  av_log(ctx, AV_LOG_WARNING, "'draw' option is deprecated and will be removed soon, "
432  "you are encouraged to use the generic timeline support through the 'enable' option\n");
433 #endif
434 
435  if (!s->fontfile && !CONFIG_FONTCONFIG) {
436  av_log(ctx, AV_LOG_ERROR, "No font filename provided\n");
437  return AVERROR(EINVAL);
438  }
439 
440  if (s->textfile) {
441  if (s->text) {
442  av_log(ctx, AV_LOG_ERROR,
443  "Both text and text file provided. Please provide only one\n");
444  return AVERROR(EINVAL);
445  }
446  if ((err = load_textfile(ctx)) < 0)
447  return err;
448  }
449 
450  if (s->reload && !s->textfile)
451  av_log(ctx, AV_LOG_WARNING, "No file to reload\n");
452 
453  if (s->tc_opt_string) {
455  s->tc_opt_string, ctx);
456  if (ret < 0)
457  return ret;
458  if (s->tc24hmax)
460  if (!s->text)
461  s->text = av_strdup("");
462  }
463 
464  if (!s->text) {
465  av_log(ctx, AV_LOG_ERROR,
466  "Either text, a valid file or a timecode must be provided\n");
467  return AVERROR(EINVAL);
468  }
469 
470  if ((err = FT_Init_FreeType(&(s->library)))) {
471  av_log(ctx, AV_LOG_ERROR,
472  "Could not load FreeType: %s\n", FT_ERRMSG(err));
473  return AVERROR(EINVAL);
474  }
475 
476  err = load_font(ctx);
477  if (err)
478  return err;
479  if (!s->fontsize)
480  s->fontsize = 16;
481  if ((err = FT_Set_Pixel_Sizes(s->face, 0, s->fontsize))) {
482  av_log(ctx, AV_LOG_ERROR, "Could not set font size to %d pixels: %s\n",
483  s->fontsize, FT_ERRMSG(err));
484  return AVERROR(EINVAL);
485  }
486 
487  s->use_kerning = FT_HAS_KERNING(s->face);
488 
489  /* load the fallback glyph with code 0 */
490  load_glyph(ctx, NULL, 0);
491 
492  /* set the tabsize in pixels */
493  if ((err = load_glyph(ctx, &glyph, ' ')) < 0) {
494  av_log(ctx, AV_LOG_ERROR, "Could not set tabsize.\n");
495  return err;
496  }
497  s->tabsize *= glyph->advance;
498 
499  if (s->exp_mode == EXP_STRFTIME &&
500  (strchr(s->text, '%') || strchr(s->text, '\\')))
501  av_log(ctx, AV_LOG_WARNING, "expansion=strftime is deprecated.\n");
502 
504 
505  return 0;
506 }
507 
509 {
511  return 0;
512 }
513 
514 static int glyph_enu_free(void *opaque, void *elem)
515 {
516  Glyph *glyph = elem;
517 
518  FT_Done_Glyph(*glyph->glyph);
519  av_freep(&glyph->glyph);
520  av_free(elem);
521  return 0;
522 }
523 
524 static av_cold void uninit(AVFilterContext *ctx)
525 {
526  DrawTextContext *s = ctx->priv;
527 
528  av_expr_free(s->x_pexpr);
529  av_expr_free(s->y_pexpr);
530 #if FF_API_DRAWTEXT_OLD_TIMELINE
532  s->x_pexpr = s->y_pexpr = s->draw_pexpr = NULL;
533 #endif
534  av_freep(&s->positions);
535  s->nb_positions = 0;
536 
537 
538  av_tree_enumerate(s->glyphs, NULL, NULL, glyph_enu_free);
540  s->glyphs = NULL;
541 
542  FT_Done_Face(s->face);
543  FT_Done_FreeType(s->library);
544 
546 }
547 
548 static inline int is_newline(uint32_t c)
549 {
550  return c == '\n' || c == '\r' || c == '\f' || c == '\v';
551 }
552 
553 static int config_input(AVFilterLink *inlink)
554 {
555  AVFilterContext *ctx = inlink->dst;
556  DrawTextContext *s = ctx->priv;
557  int ret;
558 
559  ff_draw_init(&s->dc, inlink->format, 0);
560  ff_draw_color(&s->dc, &s->fontcolor, s->fontcolor.rgba);
562  ff_draw_color(&s->dc, &s->boxcolor, s->boxcolor.rgba);
563 
564  s->var_values[VAR_w] = s->var_values[VAR_W] = s->var_values[VAR_MAIN_W] = inlink->w;
565  s->var_values[VAR_h] = s->var_values[VAR_H] = s->var_values[VAR_MAIN_H] = inlink->h;
567  s->var_values[VAR_DAR] = (double)inlink->w / inlink->h * s->var_values[VAR_SAR];
568  s->var_values[VAR_HSUB] = 1 << s->dc.hsub_max;
569  s->var_values[VAR_VSUB] = 1 << s->dc.vsub_max;
570  s->var_values[VAR_X] = NAN;
571  s->var_values[VAR_Y] = NAN;
572  s->var_values[VAR_T] = NAN;
573 
575 
576  av_expr_free(s->x_pexpr);
577  av_expr_free(s->y_pexpr);
578 #if FF_API_DRAWTEXT_OLD_TIMELINE
580  s->x_pexpr = s->y_pexpr = s->draw_pexpr = NULL;
581 #else
582  s->x_pexpr = s->y_pexpr = NULL;
583 #endif
584 
585  if ((ret = av_expr_parse(&s->x_pexpr, s->x_expr, var_names,
586  NULL, NULL, fun2_names, fun2, 0, ctx)) < 0 ||
587  (ret = av_expr_parse(&s->y_pexpr, s->y_expr, var_names,
588  NULL, NULL, fun2_names, fun2, 0, ctx)) < 0)
589 
590  return AVERROR(EINVAL);
591 #if FF_API_DRAWTEXT_OLD_TIMELINE
592  if (s->draw_expr &&
594  NULL, NULL, fun2_names, fun2, 0, ctx)) < 0)
595  return ret;
596 #endif
597 
598  return 0;
599 }
600 
601 static int command(AVFilterContext *ctx, const char *cmd, const char *arg, char *res, int res_len, int flags)
602 {
603  DrawTextContext *s = ctx->priv;
604 
605  if (!strcmp(cmd, "reinit")) {
606  int ret;
607  uninit(ctx);
608  s->reinit = 1;
609  if ((ret = init(ctx)) < 0)
610  return ret;
611  return config_input(ctx->inputs[0]);
612  }
613 
614  return AVERROR(ENOSYS);
615 }
616 
618  char *fct, unsigned argc, char **argv, int tag)
619 {
620  DrawTextContext *s = ctx->priv;
621 
623  return 0;
624 }
625 
626 static int func_pts(AVFilterContext *ctx, AVBPrint *bp,
627  char *fct, unsigned argc, char **argv, int tag)
628 {
629  DrawTextContext *s = ctx->priv;
630 
631  av_bprintf(bp, "%.6f", s->var_values[VAR_T]);
632  return 0;
633 }
634 
636  char *fct, unsigned argc, char **argv, int tag)
637 {
638  DrawTextContext *s = ctx->priv;
639 
640  av_bprintf(bp, "%d", (int)s->var_values[VAR_N]);
641  return 0;
642 }
643 
645  char *fct, unsigned argc, char **argv, int tag)
646 {
647  DrawTextContext *s = ctx->priv;
648  AVDictionaryEntry *e = av_dict_get(s->metadata, argv[0], NULL, 0);
649 
650  if (e && e->value)
651  av_bprintf(bp, "%s", e->value);
652  return 0;
653 }
654 
655 #if !HAVE_LOCALTIME_R
656 static void localtime_r(const time_t *t, struct tm *tm)
657 {
658  *tm = *localtime(t);
659 }
660 #endif
661 
663  char *fct, unsigned argc, char **argv, int tag)
664 {
665  const char *fmt = argc ? argv[0] : "%Y-%m-%d %H:%M:%S";
666  time_t now;
667  struct tm tm;
668 
669  time(&now);
670  if (tag == 'L')
671  localtime_r(&now, &tm);
672  else
673  tm = *gmtime(&now);
674  av_bprint_strftime(bp, fmt, &tm);
675  return 0;
676 }
677 
679  char *fct, unsigned argc, char **argv, int tag)
680 {
681  DrawTextContext *s = ctx->priv;
682  double res;
683  int ret;
684 
685  ret = av_expr_parse_and_eval(&res, argv[0], var_names, s->var_values,
686  NULL, NULL, fun2_names, fun2,
687  &s->prng, 0, ctx);
688  if (ret < 0)
689  av_log(ctx, AV_LOG_ERROR,
690  "Expression '%s' for the expr text expansion function is not valid\n",
691  argv[0]);
692  else
693  av_bprintf(bp, "%f", res);
694 
695  return ret;
696 }
697 
698 static const struct drawtext_function {
699  const char *name;
700  unsigned argc_min, argc_max;
701  int tag; /**< opaque argument to func */
702  int (*func)(AVFilterContext *, AVBPrint *, char *, unsigned, char **, int);
703 } functions[] = {
704  { "expr", 1, 1, 0, func_eval_expr },
705  { "e", 1, 1, 0, func_eval_expr },
706  { "pict_type", 0, 0, 0, func_pict_type },
707  { "pts", 0, 0, 0, func_pts },
708  { "gmtime", 0, 1, 'G', func_strftime },
709  { "localtime", 0, 1, 'L', func_strftime },
710  { "frame_num", 0, 0, 0, func_frame_num },
711  { "n", 0, 0, 0, func_frame_num },
712  { "metadata", 1, 1, 0, func_metadata },
713 };
714 
715 static int eval_function(AVFilterContext *ctx, AVBPrint *bp, char *fct,
716  unsigned argc, char **argv)
717 {
718  unsigned i;
719 
720  for (i = 0; i < FF_ARRAY_ELEMS(functions); i++) {
721  if (strcmp(fct, functions[i].name))
722  continue;
723  if (argc < functions[i].argc_min) {
724  av_log(ctx, AV_LOG_ERROR, "%%{%s} requires at least %d arguments\n",
725  fct, functions[i].argc_min);
726  return AVERROR(EINVAL);
727  }
728  if (argc > functions[i].argc_max) {
729  av_log(ctx, AV_LOG_ERROR, "%%{%s} requires at most %d arguments\n",
730  fct, functions[i].argc_max);
731  return AVERROR(EINVAL);
732  }
733  break;
734  }
735  if (i >= FF_ARRAY_ELEMS(functions)) {
736  av_log(ctx, AV_LOG_ERROR, "%%{%s} is not known\n", fct);
737  return AVERROR(EINVAL);
738  }
739  return functions[i].func(ctx, bp, fct, argc, argv, functions[i].tag);
740 }
741 
742 static int expand_function(AVFilterContext *ctx, AVBPrint *bp, char **rtext)
743 {
744  const char *text = *rtext;
745  char *argv[16] = { NULL };
746  unsigned argc = 0, i;
747  int ret;
748 
749  if (*text != '{') {
750  av_log(ctx, AV_LOG_ERROR, "Stray %% near '%s'\n", text);
751  return AVERROR(EINVAL);
752  }
753  text++;
754  while (1) {
755  if (!(argv[argc++] = av_get_token(&text, ":}"))) {
756  ret = AVERROR(ENOMEM);
757  goto end;
758  }
759  if (!*text) {
760  av_log(ctx, AV_LOG_ERROR, "Unterminated %%{} near '%s'\n", *rtext);
761  ret = AVERROR(EINVAL);
762  goto end;
763  }
764  if (argc == FF_ARRAY_ELEMS(argv))
765  av_freep(&argv[--argc]); /* error will be caught later */
766  if (*text == '}')
767  break;
768  text++;
769  }
770 
771  if ((ret = eval_function(ctx, bp, argv[0], argc - 1, argv + 1)) < 0)
772  goto end;
773  ret = 0;
774  *rtext = (char *)text + 1;
775 
776 end:
777  for (i = 0; i < argc; i++)
778  av_freep(&argv[i]);
779  return ret;
780 }
781 
782 static int expand_text(AVFilterContext *ctx)
783 {
784  DrawTextContext *s = ctx->priv;
785  char *text = s->text;
786  AVBPrint *bp = &s->expanded_text;
787  int ret;
788 
789  av_bprint_clear(bp);
790  while (*text) {
791  if (*text == '\\' && text[1]) {
792  av_bprint_chars(bp, text[1], 1);
793  text += 2;
794  } else if (*text == '%') {
795  text++;
796  if ((ret = expand_function(ctx, bp, &text)) < 0)
797  return ret;
798  } else {
799  av_bprint_chars(bp, *text, 1);
800  text++;
801  }
802  }
803  if (!av_bprint_is_complete(bp))
804  return AVERROR(ENOMEM);
805  return 0;
806 }
807 
809  int width, int height, const uint8_t rgbcolor[4], FFDrawColor *color, int x, int y)
810 {
811  char *text = s->expanded_text.str;
812  uint32_t code = 0;
813  int i, x1, y1;
814  uint8_t *p;
815  Glyph *glyph = NULL;
816 
817  for (i = 0, p = text; *p; i++) {
818  Glyph dummy = { 0 };
819  GET_UTF8(code, *p++, continue;);
820 
821  /* skip new line chars, just go to new line */
822  if (code == '\n' || code == '\r' || code == '\t')
823  continue;
824 
825  dummy.code = code;
826  glyph = av_tree_find(s->glyphs, &dummy, (void *)glyph_cmp, NULL);
827 
828  if (glyph->bitmap.pixel_mode != FT_PIXEL_MODE_MONO &&
829  glyph->bitmap.pixel_mode != FT_PIXEL_MODE_GRAY)
830  return AVERROR(EINVAL);
831 
832  x1 = s->positions[i].x+s->x+x;
833  y1 = s->positions[i].y+s->y+y;
834 
835  ff_blend_mask(&s->dc, color,
836  frame->data, frame->linesize, width, height,
837  glyph->bitmap.buffer, glyph->bitmap.pitch,
838  glyph->bitmap.width, glyph->bitmap.rows,
839  glyph->bitmap.pixel_mode == FT_PIXEL_MODE_MONO ? 0 : 3,
840  0, x1, y1);
841  }
842 
843  return 0;
844 }
845 
847  int width, int height)
848 {
849  DrawTextContext *s = ctx->priv;
850  AVFilterLink *inlink = ctx->inputs[0];
851 
852  uint32_t code = 0, prev_code = 0;
853  int x = 0, y = 0, i = 0, ret;
854  int max_text_line_w = 0, len;
855  int box_w, box_h;
856  char *text;
857  uint8_t *p;
858  int y_min = 32000, y_max = -32000;
859  int x_min = 32000, x_max = -32000;
860  FT_Vector delta;
861  Glyph *glyph = NULL, *prev_glyph = NULL;
862  Glyph dummy = { 0 };
863 
864  time_t now = time(0);
865  struct tm ltime;
866  AVBPrint *bp = &s->expanded_text;
867 
868  av_bprint_clear(bp);
869 
870  if(s->basetime != AV_NOPTS_VALUE)
871  now= frame->pts*av_q2d(ctx->inputs[0]->time_base) + s->basetime/1000000;
872 
873  switch (s->exp_mode) {
874  case EXP_NONE:
875  av_bprintf(bp, "%s", s->text);
876  break;
877  case EXP_NORMAL:
878  if ((ret = expand_text(ctx)) < 0)
879  return ret;
880  break;
881  case EXP_STRFTIME:
882  localtime_r(&now, &ltime);
883  av_bprint_strftime(bp, s->text, &ltime);
884  break;
885  }
886 
887  if (s->tc_opt_string) {
888  char tcbuf[AV_TIMECODE_STR_SIZE];
889  av_timecode_make_string(&s->tc, tcbuf, inlink->frame_count);
890  av_bprint_clear(bp);
891  av_bprintf(bp, "%s%s", s->text, tcbuf);
892  }
893 
894  if (!av_bprint_is_complete(bp))
895  return AVERROR(ENOMEM);
896  text = s->expanded_text.str;
897  if ((len = s->expanded_text.len) > s->nb_positions) {
898  if (!(s->positions =
899  av_realloc(s->positions, len*sizeof(*s->positions))))
900  return AVERROR(ENOMEM);
901  s->nb_positions = len;
902  }
903 
904  x = 0;
905  y = 0;
906 
907  /* load and cache glyphs */
908  for (i = 0, p = text; *p; i++) {
909  GET_UTF8(code, *p++, continue;);
910 
911  /* get glyph */
912  dummy.code = code;
913  glyph = av_tree_find(s->glyphs, &dummy, glyph_cmp, NULL);
914  if (!glyph) {
915  load_glyph(ctx, &glyph, code);
916  }
917 
918  y_min = FFMIN(glyph->bbox.yMin, y_min);
919  y_max = FFMAX(glyph->bbox.yMax, y_max);
920  x_min = FFMIN(glyph->bbox.xMin, x_min);
921  x_max = FFMAX(glyph->bbox.xMax, x_max);
922  }
923  s->max_glyph_h = y_max - y_min;
924  s->max_glyph_w = x_max - x_min;
925 
926  /* compute and save position for each glyph */
927  glyph = NULL;
928  for (i = 0, p = text; *p; i++) {
929  GET_UTF8(code, *p++, continue;);
930 
931  /* skip the \n in the sequence \r\n */
932  if (prev_code == '\r' && code == '\n')
933  continue;
934 
935  prev_code = code;
936  if (is_newline(code)) {
937 
938  max_text_line_w = FFMAX(max_text_line_w, x);
939  y += s->max_glyph_h;
940  x = 0;
941  continue;
942  }
943 
944  /* get glyph */
945  prev_glyph = glyph;
946  dummy.code = code;
947  glyph = av_tree_find(s->glyphs, &dummy, glyph_cmp, NULL);
948 
949  /* kerning */
950  if (s->use_kerning && prev_glyph && glyph->code) {
951  FT_Get_Kerning(s->face, prev_glyph->code, glyph->code,
952  ft_kerning_default, &delta);
953  x += delta.x >> 6;
954  }
955 
956  /* save position */
957  s->positions[i].x = x + glyph->bitmap_left;
958  s->positions[i].y = y - glyph->bitmap_top + y_max;
959  if (code == '\t') x = (x / s->tabsize + 1)*s->tabsize;
960  else x += glyph->advance;
961  }
962 
963  max_text_line_w = FFMAX(x, max_text_line_w);
964 
965  s->var_values[VAR_TW] = s->var_values[VAR_TEXT_W] = max_text_line_w;
967 
970  s->var_values[VAR_MAX_GLYPH_A] = s->var_values[VAR_ASCENT ] = y_max;
972 
974 
975  s->x = s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, &s->prng);
976  s->y = s->var_values[VAR_Y] = av_expr_eval(s->y_pexpr, s->var_values, &s->prng);
977  s->x = s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, &s->prng);
978 #if FF_API_DRAWTEXT_OLD_TIMELINE
979  if (s->draw_pexpr){
980  s->draw = av_expr_eval(s->draw_pexpr, s->var_values, &s->prng);
981 
982  if(!s->draw)
983  return 0;
984  }
985  if (ctx->is_disabled)
986  return 0;
987 #endif
988 
989  box_w = FFMIN(width - 1 , max_text_line_w);
990  box_h = FFMIN(height - 1, y + s->max_glyph_h);
991 
992  /* draw box */
993  if (s->draw_box)
994  ff_blend_rectangle(&s->dc, &s->boxcolor,
995  frame->data, frame->linesize, width, height,
996  s->x, s->y, box_w, box_h);
997 
998  if (s->shadowx || s->shadowy) {
999  if ((ret = draw_glyphs(s, frame, width, height, s->shadowcolor.rgba,
1000  &s->shadowcolor, s->shadowx, s->shadowy)) < 0)
1001  return ret;
1002  }
1003 
1004  if ((ret = draw_glyphs(s, frame, width, height, s->fontcolor.rgba,
1005  &s->fontcolor, 0, 0)) < 0)
1006  return ret;
1007 
1008  return 0;
1009 }
1010 
1012 {
1013  AVFilterContext *ctx = inlink->dst;
1014  AVFilterLink *outlink = ctx->outputs[0];
1015  DrawTextContext *s = ctx->priv;
1016  int ret;
1017 
1018  if (s->reload)
1019  if ((ret = load_textfile(ctx)) < 0)
1020  return ret;
1021 
1022  s->var_values[VAR_N] = inlink->frame_count+s->start_number;
1023  s->var_values[VAR_T] = frame->pts == AV_NOPTS_VALUE ?
1024  NAN : frame->pts * av_q2d(inlink->time_base);
1025 
1026  s->var_values[VAR_PICT_TYPE] = frame->pict_type;
1027  s->metadata = av_frame_get_metadata(frame);
1028 
1029  draw_text(ctx, frame, frame->width, frame->height);
1030 
1031  av_log(ctx, AV_LOG_DEBUG, "n:%d t:%f text_w:%d text_h:%d x:%d y:%d\n",
1032  (int)s->var_values[VAR_N], s->var_values[VAR_T],
1033  (int)s->var_values[VAR_TEXT_W], (int)s->var_values[VAR_TEXT_H],
1034  s->x, s->y);
1035 
1036  return ff_filter_frame(outlink, frame);
1037 }
1038 
1040  {
1041  .name = "default",
1042  .type = AVMEDIA_TYPE_VIDEO,
1043  .filter_frame = filter_frame,
1044  .config_props = config_input,
1045  .needs_writable = 1,
1046  },
1047  { NULL }
1048 };
1049 
1051  {
1052  .name = "default",
1053  .type = AVMEDIA_TYPE_VIDEO,
1054  },
1055  { NULL }
1056 };
1057 
1059  .name = "drawtext",
1060  .description = NULL_IF_CONFIG_SMALL("Draw text on top of video frames using libfreetype library."),
1061  .priv_size = sizeof(DrawTextContext),
1062  .priv_class = &drawtext_class,
1063  .init = init,
1064  .uninit = uninit,
1066  .inputs = avfilter_vf_drawtext_inputs,
1067  .outputs = avfilter_vf_drawtext_outputs,
1071 #else
1073 #endif
1074 };