FFmpeg
 All Data Structures 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 <freetype/config/ftheader.h>
52 #include FT_FREETYPE_H
53 #include FT_GLYPH_H
54 #if CONFIG_FONTCONFIG
55 #include <fontconfig/fontconfig.h>
56 #endif
57 
58 static const char *const var_names[] = {
59  "dar",
60  "hsub", "vsub",
61  "line_h", "lh", ///< line height, same as max_glyph_h
62  "main_h", "h", "H", ///< height of the input video
63  "main_w", "w", "W", ///< width of the input video
64  "max_glyph_a", "ascent", ///< max glyph ascent
65  "max_glyph_d", "descent", ///< min glyph descent
66  "max_glyph_h", ///< max glyph height
67  "max_glyph_w", ///< max glyph width
68  "n", ///< number of frame
69  "sar",
70  "t", ///< timestamp expressed in seconds
71  "text_h", "th", ///< height of the rendered text
72  "text_w", "tw", ///< width of the rendered text
73  "x",
74  "y",
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 {
95  VAR_DAR,
96  VAR_HSUB, VAR_VSUB,
97  VAR_LINE_H, VAR_LH,
98  VAR_MAIN_H, VAR_h, VAR_H,
99  VAR_MAIN_W, VAR_w, VAR_W,
100  VAR_MAX_GLYPH_A, VAR_ASCENT,
101  VAR_MAX_GLYPH_D, VAR_DESCENT,
102  VAR_MAX_GLYPH_H,
103  VAR_MAX_GLYPH_W,
104  VAR_N,
105  VAR_SAR,
106  VAR_T,
107  VAR_TEXT_H, VAR_TH,
108  VAR_TEXT_W, VAR_TW,
109  VAR_X,
110  VAR_Y,
111  VAR_VARS_NB
112 };
113 
114 enum expansion_mode {
115  EXP_NONE,
116  EXP_NORMAL,
117  EXP_STRFTIME,
118 };
119 
120 typedef struct {
121  const AVClass *class;
122  enum expansion_mode exp_mode; ///< expansion mode to use for the text
123  int reinit; ///< tells if the filter is being reinited
124  uint8_t *fontfile; ///< font to be used
125  uint8_t *text; ///< text to be drawn
126  AVBPrint expanded_text; ///< used to contain the expanded text
127  int ft_load_flags; ///< flags used for loading fonts, see FT_LOAD_*
128  FT_Vector *positions; ///< positions for each element in the text
129  size_t nb_positions; ///< number of elements of positions array
130  char *textfile; ///< file with text to be drawn
131  int x; ///< x position to start drawing text
132  int y; ///< y position to start drawing text
133  int max_glyph_w; ///< max glyph width
134  int max_glyph_h; ///< max glyph height
135  int shadowx, shadowy;
136  unsigned int fontsize; ///< font size to use
137  char *fontcolor_string; ///< font color as string
138  char *boxcolor_string; ///< box color as string
139  char *shadowcolor_string; ///< shadow color as string
140 
141  short int draw_box; ///< draw box around text - true or false
142  int use_kerning; ///< font kerning is used - true/false
143  int tabsize; ///< tab size
144  int fix_bounds; ///< do we let it go out of frame bounds - t/f
145 
146  FFDrawContext dc;
147  FFDrawColor fontcolor; ///< foreground color
148  FFDrawColor shadowcolor; ///< shadow color
149  FFDrawColor boxcolor; ///< background color
150 
151  FT_Library library; ///< freetype font library handle
152  FT_Face face; ///< freetype font face handle
153  struct AVTreeNode *glyphs; ///< rendered glyphs, stored using the UTF-32 char code
154  char *x_expr; ///< expression for x position
155  char *y_expr; ///< expression for y position
156  AVExpr *x_pexpr, *y_pexpr; ///< parsed expressions for x and y
157  int64_t basetime; ///< base pts time in the real world for display
158  double var_values[VAR_VARS_NB];
159  char *draw_expr; ///< expression for draw
160  AVExpr *draw_pexpr; ///< parsed expression for draw
161  int draw; ///< set to zero to prevent drawing
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 frame_id;
168  int reload; ///< reload text file for each frame
169 } DrawTextContext;
170 
171 #define OFFSET(x) offsetof(DrawTextContext, x)
172 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
173 
174 static const AVOption drawtext_options[]= {
175 {"fontfile", "set font file", OFFSET(fontfile), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
176 {"text", "set text", OFFSET(text), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
177 {"textfile", "set text file", OFFSET(textfile), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
178 {"fontcolor", "set foreground color", OFFSET(fontcolor_string), AV_OPT_TYPE_STRING, {.str="black"}, CHAR_MIN, CHAR_MAX, FLAGS},
179 {"boxcolor", "set box color", OFFSET(boxcolor_string), AV_OPT_TYPE_STRING, {.str="white"}, CHAR_MIN, CHAR_MAX, FLAGS},
180 {"shadowcolor", "set shadow color", OFFSET(shadowcolor_string), AV_OPT_TYPE_STRING, {.str="black"}, CHAR_MIN, CHAR_MAX, FLAGS},
181 {"box", "set box", OFFSET(draw_box), AV_OPT_TYPE_INT, {.i64=0}, 0, 1 , FLAGS},
182 {"fontsize", "set font size", OFFSET(fontsize), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX , FLAGS},
183 {"x", "set x expression", OFFSET(x_expr), AV_OPT_TYPE_STRING, {.str="0"}, CHAR_MIN, CHAR_MAX, FLAGS},
184 {"y", "set y expression", OFFSET(y_expr), AV_OPT_TYPE_STRING, {.str="0"}, CHAR_MIN, CHAR_MAX, FLAGS},
185 {"shadowx", "set x", OFFSET(shadowx), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX , FLAGS},
186 {"shadowy", "set y", OFFSET(shadowy), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX , FLAGS},
187 {"tabsize", "set tab size", OFFSET(tabsize), AV_OPT_TYPE_INT, {.i64=4}, 0, INT_MAX , FLAGS},
188 {"basetime", "set base time", OFFSET(basetime), AV_OPT_TYPE_INT64, {.i64=AV_NOPTS_VALUE}, INT64_MIN, INT64_MAX , FLAGS},
189 {"draw", "if false do not draw", OFFSET(draw_expr), AV_OPT_TYPE_STRING, {.str="1"}, CHAR_MIN, CHAR_MAX, FLAGS},
190 
191 {"expansion","set the expansion mode", OFFSET(exp_mode), AV_OPT_TYPE_INT, {.i64=EXP_STRFTIME}, 0, 2, FLAGS, "expansion"},
192 {"none", "set no expansion", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_NONE}, 0, 0, FLAGS, "expansion"},
193 {"normal", "set normal expansion", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_NORMAL}, 0, 0, FLAGS, "expansion"},
194 {"strftime", "set strftime expansion (deprecated)", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_STRFTIME}, 0, 0, FLAGS, "expansion"},
195 
196 {"timecode", "set initial timecode", OFFSET(tc_opt_string), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
197 {"tc24hmax", "set 24 hours max (timecode only)", OFFSET(tc24hmax), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS},
198 {"timecode_rate", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
199 {"r", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
200 {"rate", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
201 {"reload", "reload text file for each frame", OFFSET(reload), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS},
202 {"fix_bounds", "if true, check and fix text coords to avoid clipping", OFFSET(fix_bounds), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, FLAGS},
203 
204 /* FT_LOAD_* flags */
205 {"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"},
206 {"default", "set default", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_DEFAULT}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
207 {"no_scale", "set no_scale", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_NO_SCALE}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
208 {"no_hinting", "set no_hinting", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_NO_HINTING}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
209 {"render", "set render", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_RENDER}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
210 {"no_bitmap", "set no_bitmap", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_NO_BITMAP}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
211 {"vertical_layout", "set vertical_layout", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_VERTICAL_LAYOUT}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
212 {"force_autohint", "set force_autohint", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_FORCE_AUTOHINT}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
213 {"crop_bitmap", "set crop_bitmap", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_CROP_BITMAP}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
214 {"pedantic", "set pedantic", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_PEDANTIC}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
215 {"ignore_global_advance_width", "set ignore_global_advance_width", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_IGNORE_GLOBAL_ADVANCE_WIDTH}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
216 {"no_recurse", "set no_recurse", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_NO_RECURSE}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
217 {"ignore_transform", "set ignore_transform", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_IGNORE_TRANSFORM}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
218 {"monochrome", "set monochrome", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_MONOCHROME}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
219 {"linear_design", "set linear_design", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_LINEAR_DESIGN}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
220 {"no_autohint", "set no_autohint", 0, AV_OPT_TYPE_CONST, {.i64=FT_LOAD_NO_AUTOHINT}, INT_MIN, INT_MAX, FLAGS, "ft_load_flags"},
221 {NULL},
222 };
223 
224 AVFILTER_DEFINE_CLASS(drawtext);
225 
226 #undef __FTERRORS_H__
227 #define FT_ERROR_START_LIST {
228 #define FT_ERRORDEF(e, v, s) { (e), (s) },
229 #define FT_ERROR_END_LIST { 0, NULL } };
230 
231 struct ft_error
232 {
233  int err;
234  const char *err_msg;
235 } static ft_errors[] =
236 #include FT_ERRORS_H
237 
238 #define FT_ERRMSG(e) ft_errors[e].err_msg
239 
240 typedef struct {
241  FT_Glyph *glyph;
242  uint32_t code;
243  FT_Bitmap bitmap; ///< array holding bitmaps of font
244  FT_BBox bbox;
245  int advance;
246  int bitmap_left;
247  int bitmap_top;
248 } Glyph;
249 
250 static int glyph_cmp(void *key, const void *b)
251 {
252  const Glyph *a = key, *bb = b;
253  int64_t diff = (int64_t)a->code - (int64_t)bb->code;
254  return diff > 0 ? 1 : diff < 0 ? -1 : 0;
255 }
256 
257 /**
258  * Load glyphs corresponding to the UTF-32 codepoint code.
259  */
260 static int load_glyph(AVFilterContext *ctx, Glyph **glyph_ptr, uint32_t code)
261 {
262  DrawTextContext *dtext = ctx->priv;
263  Glyph *glyph;
264  struct AVTreeNode *node = NULL;
265  int ret;
266 
267  /* load glyph into dtext->face->glyph */
268  if (FT_Load_Char(dtext->face, code, dtext->ft_load_flags))
269  return AVERROR(EINVAL);
270 
271  /* save glyph */
272  if (!(glyph = av_mallocz(sizeof(*glyph))) ||
273  !(glyph->glyph = av_mallocz(sizeof(*glyph->glyph)))) {
274  ret = AVERROR(ENOMEM);
275  goto error;
276  }
277  glyph->code = code;
278 
279  if (FT_Get_Glyph(dtext->face->glyph, glyph->glyph)) {
280  ret = AVERROR(EINVAL);
281  goto error;
282  }
283 
284  glyph->bitmap = dtext->face->glyph->bitmap;
285  glyph->bitmap_left = dtext->face->glyph->bitmap_left;
286  glyph->bitmap_top = dtext->face->glyph->bitmap_top;
287  glyph->advance = dtext->face->glyph->advance.x >> 6;
288 
289  /* measure text height to calculate text_height (or the maximum text height) */
290  FT_Glyph_Get_CBox(*glyph->glyph, ft_glyph_bbox_pixels, &glyph->bbox);
291 
292  /* cache the newly created glyph */
293  if (!(node = av_tree_node_alloc())) {
294  ret = AVERROR(ENOMEM);
295  goto error;
296  }
297  av_tree_insert(&dtext->glyphs, glyph, glyph_cmp, &node);
298 
299  if (glyph_ptr)
300  *glyph_ptr = glyph;
301  return 0;
302 
303 error:
304  if (glyph)
305  av_freep(&glyph->glyph);
306  av_freep(&glyph);
307  av_freep(&node);
308  return ret;
309 }
310 
311 static int load_font_file(AVFilterContext *ctx, const char *path, int index,
312  const char **error)
313 {
314  DrawTextContext *dtext = ctx->priv;
315  int err;
316 
317  err = FT_New_Face(dtext->library, path, index, &dtext->face);
318  if (err) {
319  *error = FT_ERRMSG(err);
320  return AVERROR(EINVAL);
321  }
322  return 0;
323 }
324 
325 #if CONFIG_FONTCONFIG
326 static int load_font_fontconfig(AVFilterContext *ctx, const char **error)
327 {
328  DrawTextContext *dtext = ctx->priv;
329  FcConfig *fontconfig;
330  FcPattern *pattern, *fpat;
331  FcResult result = FcResultMatch;
332  FcChar8 *filename;
333  int err, index;
334  double size;
335 
336  fontconfig = FcInitLoadConfigAndFonts();
337  if (!fontconfig) {
338  *error = "impossible to init fontconfig\n";
339  return AVERROR(EINVAL);
340  }
341  pattern = FcNameParse(dtext->fontfile ? dtext->fontfile :
342  (uint8_t *)(intptr_t)"default");
343  if (!pattern) {
344  *error = "could not parse fontconfig pattern";
345  return AVERROR(EINVAL);
346  }
347  if (!FcConfigSubstitute(fontconfig, pattern, FcMatchPattern)) {
348  *error = "could not substitue fontconfig options"; /* very unlikely */
349  return AVERROR(EINVAL);
350  }
351  FcDefaultSubstitute(pattern);
352  fpat = FcFontMatch(fontconfig, pattern, &result);
353  if (!fpat || result != FcResultMatch) {
354  *error = "impossible to find a matching font";
355  return AVERROR(EINVAL);
356  }
357  if (FcPatternGetString (fpat, FC_FILE, 0, &filename) != FcResultMatch ||
358  FcPatternGetInteger(fpat, FC_INDEX, 0, &index ) != FcResultMatch ||
359  FcPatternGetDouble (fpat, FC_SIZE, 0, &size ) != FcResultMatch) {
360  *error = "impossible to find font information";
361  return AVERROR(EINVAL);
362  }
363  av_log(ctx, AV_LOG_INFO, "Using \"%s\"\n", filename);
364  if (!dtext->fontsize)
365  dtext->fontsize = size + 0.5;
366  err = load_font_file(ctx, filename, index, error);
367  if (err)
368  return err;
369  FcPatternDestroy(fpat);
370  FcPatternDestroy(pattern);
371  FcConfigDestroy(fontconfig);
372  return 0;
373 }
374 #endif
375 
376 static int load_font(AVFilterContext *ctx)
377 {
378  DrawTextContext *dtext = ctx->priv;
379  int err;
380  const char *error = "unknown error\n";
381 
382  /* load the face, and set up the encoding, which is by default UTF-8 */
383  err = load_font_file(ctx, dtext->fontfile, 0, &error);
384  if (!err)
385  return 0;
386 #if CONFIG_FONTCONFIG
387  err = load_font_fontconfig(ctx, &error);
388  if (!err)
389  return 0;
390 #endif
391  av_log(ctx, AV_LOG_ERROR, "Could not load font \"%s\": %s\n",
392  dtext->fontfile, error);
393  return err;
394 }
395 
396 static int load_textfile(AVFilterContext *ctx)
397 {
398  DrawTextContext *dtext = ctx->priv;
399  int err;
400  uint8_t *textbuf;
401  size_t textbuf_size;
402 
403  if ((err = av_file_map(dtext->textfile, &textbuf, &textbuf_size, 0, ctx)) < 0) {
404  av_log(ctx, AV_LOG_ERROR,
405  "The text file '%s' could not be read or is empty\n",
406  dtext->textfile);
407  return err;
408  }
409 
410  if (!(dtext->text = av_realloc(dtext->text, textbuf_size + 1)))
411  return AVERROR(ENOMEM);
412  memcpy(dtext->text, textbuf, textbuf_size);
413  dtext->text[textbuf_size] = 0;
414  av_file_unmap(textbuf, textbuf_size);
415 
416  return 0;
417 }
418 
419 static av_cold int init(AVFilterContext *ctx, const char *args)
420 {
421  int err;
422  DrawTextContext *dtext = ctx->priv;
423  Glyph *glyph;
424 
425  dtext->class = &drawtext_class;
426  av_opt_set_defaults(dtext);
427 
428  if ((err = av_set_options_string(dtext, args, "=", ":")) < 0)
429  return err;
430 
431  if (!dtext->fontfile && !CONFIG_FONTCONFIG) {
432  av_log(ctx, AV_LOG_ERROR, "No font filename provided\n");
433  return AVERROR(EINVAL);
434  }
435 
436  if (dtext->textfile) {
437  if (dtext->text) {
438  av_log(ctx, AV_LOG_ERROR,
439  "Both text and text file provided. Please provide only one\n");
440  return AVERROR(EINVAL);
441  }
442  if ((err = load_textfile(ctx)) < 0)
443  return err;
444  }
445 
446  if (dtext->reload && !dtext->textfile)
447  av_log(ctx, AV_LOG_WARNING, "No file to reload\n");
448 
449  if (dtext->tc_opt_string) {
450  int ret = av_timecode_init_from_string(&dtext->tc, dtext->tc_rate,
451  dtext->tc_opt_string, ctx);
452  if (ret < 0)
453  return ret;
454  if (dtext->tc24hmax)
455  dtext->tc.flags |= AV_TIMECODE_FLAG_24HOURSMAX;
456  if (!dtext->text)
457  dtext->text = av_strdup("");
458  }
459 
460  if (!dtext->text) {
461  av_log(ctx, AV_LOG_ERROR,
462  "Either text, a valid file or a timecode must be provided\n");
463  return AVERROR(EINVAL);
464  }
465 
466  if ((err = av_parse_color(dtext->fontcolor.rgba, dtext->fontcolor_string, -1, ctx))) {
467  av_log(ctx, AV_LOG_ERROR,
468  "Invalid font color '%s'\n", dtext->fontcolor_string);
469  return err;
470  }
471 
472  if ((err = av_parse_color(dtext->boxcolor.rgba, dtext->boxcolor_string, -1, ctx))) {
473  av_log(ctx, AV_LOG_ERROR,
474  "Invalid box color '%s'\n", dtext->boxcolor_string);
475  return err;
476  }
477 
478  if ((err = av_parse_color(dtext->shadowcolor.rgba, dtext->shadowcolor_string, -1, ctx))) {
479  av_log(ctx, AV_LOG_ERROR,
480  "Invalid shadow color '%s'\n", dtext->shadowcolor_string);
481  return err;
482  }
483 
484  if ((err = FT_Init_FreeType(&(dtext->library)))) {
485  av_log(ctx, AV_LOG_ERROR,
486  "Could not load FreeType: %s\n", FT_ERRMSG(err));
487  return AVERROR(EINVAL);
488  }
489 
490  err = load_font(ctx);
491  if (err)
492  return err;
493  if (!dtext->fontsize)
494  dtext->fontsize = 16;
495  if ((err = FT_Set_Pixel_Sizes(dtext->face, 0, dtext->fontsize))) {
496  av_log(ctx, AV_LOG_ERROR, "Could not set font size to %d pixels: %s\n",
497  dtext->fontsize, FT_ERRMSG(err));
498  return AVERROR(EINVAL);
499  }
500 
501  dtext->use_kerning = FT_HAS_KERNING(dtext->face);
502 
503  /* load the fallback glyph with code 0 */
504  load_glyph(ctx, NULL, 0);
505 
506  /* set the tabsize in pixels */
507  if ((err = load_glyph(ctx, &glyph, ' ')) < 0) {
508  av_log(ctx, AV_LOG_ERROR, "Could not set tabsize.\n");
509  return err;
510  }
511  dtext->tabsize *= glyph->advance;
512 
513  if (dtext->exp_mode == EXP_STRFTIME &&
514  (strchr(dtext->text, '%') || strchr(dtext->text, '\\')))
515  av_log(ctx, AV_LOG_WARNING, "expansion=strftime is deprecated.\n");
516 
517  av_bprint_init(&dtext->expanded_text, 0, AV_BPRINT_SIZE_UNLIMITED);
518 
519  return 0;
520 }
521 
522 static int query_formats(AVFilterContext *ctx)
523 {
524  ff_set_common_formats(ctx, ff_draw_supported_pixel_formats(0));
525  return 0;
526 }
527 
528 static int glyph_enu_free(void *opaque, void *elem)
529 {
530  Glyph *glyph = elem;
531 
532  FT_Done_Glyph(*glyph->glyph);
533  av_freep(&glyph->glyph);
534  av_free(elem);
535  return 0;
536 }
537 
538 static av_cold void uninit(AVFilterContext *ctx)
539 {
540  DrawTextContext *dtext = ctx->priv;
541 
542  av_expr_free(dtext->x_pexpr); dtext->x_pexpr = NULL;
543  av_expr_free(dtext->y_pexpr); dtext->y_pexpr = NULL;
544  av_expr_free(dtext->draw_pexpr); dtext->draw_pexpr = NULL;
545  av_opt_free(dtext);
546 
547  av_freep(&dtext->positions);
548  dtext->nb_positions = 0;
549 
550  av_tree_enumerate(dtext->glyphs, NULL, NULL, glyph_enu_free);
551  av_tree_destroy(dtext->glyphs);
552  dtext->glyphs = NULL;
553 
554  FT_Done_Face(dtext->face);
555  FT_Done_FreeType(dtext->library);
556 
557  av_bprint_finalize(&dtext->expanded_text, NULL);
558 }
559 
560 static inline int is_newline(uint32_t c)
561 {
562  return c == '\n' || c == '\r' || c == '\f' || c == '\v';
563 }
564 
565 static int config_input(AVFilterLink *inlink)
566 {
567  AVFilterContext *ctx = inlink->dst;
568  DrawTextContext *dtext = ctx->priv;
569  int ret;
570 
571  ff_draw_init(&dtext->dc, inlink->format, 0);
572  ff_draw_color(&dtext->dc, &dtext->fontcolor, dtext->fontcolor.rgba);
573  ff_draw_color(&dtext->dc, &dtext->shadowcolor, dtext->shadowcolor.rgba);
574  ff_draw_color(&dtext->dc, &dtext->boxcolor, dtext->boxcolor.rgba);
575 
576  dtext->var_values[VAR_w] = dtext->var_values[VAR_W] = dtext->var_values[VAR_MAIN_W] = inlink->w;
577  dtext->var_values[VAR_h] = dtext->var_values[VAR_H] = dtext->var_values[VAR_MAIN_H] = inlink->h;
578  dtext->var_values[VAR_SAR] = inlink->sample_aspect_ratio.num ? av_q2d(inlink->sample_aspect_ratio) : 1;
579  dtext->var_values[VAR_DAR] = (double)inlink->w / inlink->h * dtext->var_values[VAR_SAR];
580  dtext->var_values[VAR_HSUB] = 1 << dtext->dc.hsub_max;
581  dtext->var_values[VAR_VSUB] = 1 << dtext->dc.vsub_max;
582  dtext->var_values[VAR_X] = NAN;
583  dtext->var_values[VAR_Y] = NAN;
584  if (!dtext->reinit)
585  dtext->var_values[VAR_N] = 0;
586  dtext->var_values[VAR_T] = NAN;
587 
588  av_lfg_init(&dtext->prng, av_get_random_seed());
589 
590  if ((ret = av_expr_parse(&dtext->x_pexpr, dtext->x_expr, var_names,
591  NULL, NULL, fun2_names, fun2, 0, ctx)) < 0 ||
592  (ret = av_expr_parse(&dtext->y_pexpr, dtext->y_expr, var_names,
593  NULL, NULL, fun2_names, fun2, 0, ctx)) < 0 ||
594  (ret = av_expr_parse(&dtext->draw_pexpr, dtext->draw_expr, var_names,
595  NULL, NULL, fun2_names, fun2, 0, ctx)) < 0)
596 
597  return AVERROR(EINVAL);
598 
599  return 0;
600 }
601 
602 static int command(AVFilterContext *ctx, const char *cmd, const char *arg, char *res, int res_len, int flags)
603 {
604  DrawTextContext *dtext = ctx->priv;
605 
606  if (!strcmp(cmd, "reinit")) {
607  int ret;
608  uninit(ctx);
609  dtext->reinit = 1;
610  if ((ret = init(ctx, arg)) < 0)
611  return ret;
612  return config_input(ctx->inputs[0]);
613  }
614 
615  return AVERROR(ENOSYS);
616 }
617 
618 static int func_pts(AVFilterContext *ctx, AVBPrint *bp,
619  char *fct, unsigned argc, char **argv, int tag)
620 {
621  DrawTextContext *dtext = ctx->priv;
622 
623  av_bprintf(bp, "%.6f", dtext->var_values[VAR_T]);
624  return 0;
625 }
626 
627 static int func_frame_num(AVFilterContext *ctx, AVBPrint *bp,
628  char *fct, unsigned argc, char **argv, int tag)
629 {
630  DrawTextContext *dtext = ctx->priv;
631 
632  av_bprintf(bp, "%d", (int)dtext->var_values[VAR_N]);
633  return 0;
634 }
635 
636 #if !HAVE_LOCALTIME_R
637 static void localtime_r(const time_t *t, struct tm *tm)
638 {
639  *tm = *localtime(t);
640 }
641 #endif
642 
643 static int func_strftime(AVFilterContext *ctx, AVBPrint *bp,
644  char *fct, unsigned argc, char **argv, int tag)
645 {
646  const char *fmt = argc ? argv[0] : "%Y-%m-%d %H:%M:%S";
647  time_t now;
648  struct tm tm;
649 
650  time(&now);
651  if (tag == 'L')
652  localtime_r(&now, &tm);
653  else
654  tm = *gmtime(&now);
655  av_bprint_strftime(bp, fmt, &tm);
656  return 0;
657 }
658 
659 static int func_eval_expr(AVFilterContext *ctx, AVBPrint *bp,
660  char *fct, unsigned argc, char **argv, int tag)
661 {
662  DrawTextContext *dtext = ctx->priv;
663  double res;
664  int ret;
665 
666  ret = av_expr_parse_and_eval(&res, argv[0], var_names, dtext->var_values,
667  NULL, NULL, fun2_names, fun2,
668  &dtext->prng, 0, ctx);
669  if (ret < 0)
670  av_log(ctx, AV_LOG_ERROR,
671  "Expression '%s' for the expr text expansion function is not valid\n",
672  argv[0]);
673  else
674  av_bprintf(bp, "%f", res);
675 
676  return ret;
677 }
678 
679 static const struct drawtext_function {
680  const char *name;
681  unsigned argc_min, argc_max;
682  int tag; /** opaque argument to func */
683  int (*func)(AVFilterContext *, AVBPrint *, char *, unsigned, char **, int);
684 } functions[] = {
685  { "expr", 1, 1, 0, func_eval_expr },
686  { "e", 1, 1, 0, func_eval_expr },
687  { "pts", 0, 0, 0, func_pts },
688  { "gmtime", 0, 1, 'G', func_strftime },
689  { "localtime", 0, 1, 'L', func_strftime },
690  { "frame_num", 0, 0, 0, func_frame_num },
691  { "n", 0, 0, 0, func_frame_num },
692 };
693 
694 static int eval_function(AVFilterContext *ctx, AVBPrint *bp, char *fct,
695  unsigned argc, char **argv)
696 {
697  unsigned i;
698 
699  for (i = 0; i < FF_ARRAY_ELEMS(functions); i++) {
700  if (strcmp(fct, functions[i].name))
701  continue;
702  if (argc < functions[i].argc_min) {
703  av_log(ctx, AV_LOG_ERROR, "%%{%s} requires at least %d arguments\n",
704  fct, functions[i].argc_min);
705  return AVERROR(EINVAL);
706  }
707  if (argc > functions[i].argc_max) {
708  av_log(ctx, AV_LOG_ERROR, "%%{%s} requires at most %d arguments\n",
709  fct, functions[i].argc_max);
710  return AVERROR(EINVAL);
711  }
712  break;
713  }
714  if (i >= FF_ARRAY_ELEMS(functions)) {
715  av_log(ctx, AV_LOG_ERROR, "%%{%s} is not known\n", fct);
716  return AVERROR(EINVAL);
717  }
718  return functions[i].func(ctx, bp, fct, argc, argv, functions[i].tag);
719 }
720 
721 static int expand_function(AVFilterContext *ctx, AVBPrint *bp, char **rtext)
722 {
723  const char *text = *rtext;
724  char *argv[16] = { NULL };
725  unsigned argc = 0, i;
726  int ret;
727 
728  if (*text != '{') {
729  av_log(ctx, AV_LOG_ERROR, "Stray %% near '%s'\n", text);
730  return AVERROR(EINVAL);
731  }
732  text++;
733  while (1) {
734  if (!(argv[argc++] = av_get_token(&text, ":}"))) {
735  ret = AVERROR(ENOMEM);
736  goto end;
737  }
738  if (!*text) {
739  av_log(ctx, AV_LOG_ERROR, "Unterminated %%{} near '%s'\n", *rtext);
740  ret = AVERROR(EINVAL);
741  goto end;
742  }
743  if (argc == FF_ARRAY_ELEMS(argv))
744  av_freep(&argv[--argc]); /* error will be caught later */
745  if (*text == '}')
746  break;
747  text++;
748  }
749 
750  if ((ret = eval_function(ctx, bp, argv[0], argc - 1, argv + 1)) < 0)
751  goto end;
752  ret = 0;
753  *rtext = (char *)text + 1;
754 
755 end:
756  for (i = 0; i < argc; i++)
757  av_freep(&argv[i]);
758  return ret;
759 }
760 
761 static int expand_text(AVFilterContext *ctx)
762 {
763  DrawTextContext *dtext = ctx->priv;
764  char *text = dtext->text;
765  AVBPrint *bp = &dtext->expanded_text;
766  int ret;
767 
768  av_bprint_clear(bp);
769  while (*text) {
770  if (*text == '\\' && text[1]) {
771  av_bprint_chars(bp, text[1], 1);
772  text += 2;
773  } else if (*text == '%') {
774  text++;
775  if ((ret = expand_function(ctx, bp, &text)) < 0)
776  return ret;
777  } else {
778  av_bprint_chars(bp, *text, 1);
779  text++;
780  }
781  }
782  if (!av_bprint_is_complete(bp))
783  return AVERROR(ENOMEM);
784  return 0;
785 }
786 
787 static int draw_glyphs(DrawTextContext *dtext, AVFilterBufferRef *picref,
788  int width, int height, const uint8_t rgbcolor[4], FFDrawColor *color, int x, int y)
789 {
790  char *text = dtext->expanded_text.str;
791  uint32_t code = 0;
792  int i, x1, y1;
793  uint8_t *p;
794  Glyph *glyph = NULL;
795 
796  for (i = 0, p = text; *p; i++) {
797  Glyph dummy = { 0 };
798  GET_UTF8(code, *p++, continue;);
799 
800  /* skip new line chars, just go to new line */
801  if (code == '\n' || code == '\r' || code == '\t')
802  continue;
803 
804  dummy.code = code;
805  glyph = av_tree_find(dtext->glyphs, &dummy, (void *)glyph_cmp, NULL);
806 
807  if (glyph->bitmap.pixel_mode != FT_PIXEL_MODE_MONO &&
808  glyph->bitmap.pixel_mode != FT_PIXEL_MODE_GRAY)
809  return AVERROR(EINVAL);
810 
811  x1 = dtext->positions[i].x+dtext->x+x;
812  y1 = dtext->positions[i].y+dtext->y+y;
813 
814  ff_blend_mask(&dtext->dc, color,
815  picref->data, picref->linesize, width, height,
816  glyph->bitmap.buffer, glyph->bitmap.pitch,
817  glyph->bitmap.width, glyph->bitmap.rows,
818  glyph->bitmap.pixel_mode == FT_PIXEL_MODE_MONO ? 0 : 3,
819  0, x1, y1);
820  }
821 
822  return 0;
823 }
824 
825 static int draw_text(AVFilterContext *ctx, AVFilterBufferRef *picref,
826  int width, int height)
827 {
828  DrawTextContext *dtext = ctx->priv;
829  uint32_t code = 0, prev_code = 0;
830  int x = 0, y = 0, i = 0, ret;
831  int max_text_line_w = 0, len;
832  int box_w, box_h;
833  char *text = dtext->text;
834  uint8_t *p;
835  int y_min = 32000, y_max = -32000;
836  int x_min = 32000, x_max = -32000;
837  FT_Vector delta;
838  Glyph *glyph = NULL, *prev_glyph = NULL;
839  Glyph dummy = { 0 };
840 
841  time_t now = time(0);
842  struct tm ltime;
843  AVBPrint *bp = &dtext->expanded_text;
844 
845  av_bprint_clear(bp);
846 
847  if(dtext->basetime != AV_NOPTS_VALUE)
848  now= picref->pts*av_q2d(ctx->inputs[0]->time_base) + dtext->basetime/1000000;
849 
850  switch (dtext->exp_mode) {
851  case EXP_NONE:
852  av_bprintf(bp, "%s", dtext->text);
853  break;
854  case EXP_NORMAL:
855  if ((ret = expand_text(ctx)) < 0)
856  return ret;
857  break;
858  case EXP_STRFTIME:
859  localtime_r(&now, &ltime);
860  av_bprint_strftime(bp, dtext->text, &ltime);
861  break;
862  }
863 
864  if (dtext->tc_opt_string) {
865  char tcbuf[AV_TIMECODE_STR_SIZE];
866  av_timecode_make_string(&dtext->tc, tcbuf, dtext->frame_id++);
867  av_bprint_clear(bp);
868  av_bprintf(bp, "%s%s", dtext->text, tcbuf);
869  }
870 
871  if (!av_bprint_is_complete(bp))
872  return AVERROR(ENOMEM);
873  text = dtext->expanded_text.str;
874  if ((len = dtext->expanded_text.len) > dtext->nb_positions) {
875  if (!(dtext->positions =
876  av_realloc(dtext->positions, len*sizeof(*dtext->positions))))
877  return AVERROR(ENOMEM);
878  dtext->nb_positions = len;
879  }
880 
881  x = 0;
882  y = 0;
883 
884  /* load and cache glyphs */
885  for (i = 0, p = text; *p; i++) {
886  GET_UTF8(code, *p++, continue;);
887 
888  /* get glyph */
889  dummy.code = code;
890  glyph = av_tree_find(dtext->glyphs, &dummy, glyph_cmp, NULL);
891  if (!glyph) {
892  load_glyph(ctx, &glyph, code);
893  }
894 
895  y_min = FFMIN(glyph->bbox.yMin, y_min);
896  y_max = FFMAX(glyph->bbox.yMax, y_max);
897  x_min = FFMIN(glyph->bbox.xMin, x_min);
898  x_max = FFMAX(glyph->bbox.xMax, x_max);
899  }
900  dtext->max_glyph_h = y_max - y_min;
901  dtext->max_glyph_w = x_max - x_min;
902 
903  /* compute and save position for each glyph */
904  glyph = NULL;
905  for (i = 0, p = text; *p; i++) {
906  GET_UTF8(code, *p++, continue;);
907 
908  /* skip the \n in the sequence \r\n */
909  if (prev_code == '\r' && code == '\n')
910  continue;
911 
912  prev_code = code;
913  if (is_newline(code)) {
914  max_text_line_w = FFMAX(max_text_line_w, x);
915  y += dtext->max_glyph_h;
916  x = 0;
917  continue;
918  }
919 
920  /* get glyph */
921  prev_glyph = glyph;
922  dummy.code = code;
923  glyph = av_tree_find(dtext->glyphs, &dummy, glyph_cmp, NULL);
924 
925  /* kerning */
926  if (dtext->use_kerning && prev_glyph && glyph->code) {
927  FT_Get_Kerning(dtext->face, prev_glyph->code, glyph->code,
928  ft_kerning_default, &delta);
929  x += delta.x >> 6;
930  }
931 
932  /* save position */
933  dtext->positions[i].x = x + glyph->bitmap_left;
934  dtext->positions[i].y = y - glyph->bitmap_top + y_max;
935  if (code == '\t') x = (x / dtext->tabsize + 1)*dtext->tabsize;
936  else x += glyph->advance;
937  }
938 
939  max_text_line_w = FFMAX(x, max_text_line_w);
940 
941  dtext->var_values[VAR_TW] = dtext->var_values[VAR_TEXT_W] = max_text_line_w;
942  dtext->var_values[VAR_TH] = dtext->var_values[VAR_TEXT_H] = y + dtext->max_glyph_h;
943 
944  dtext->var_values[VAR_MAX_GLYPH_W] = dtext->max_glyph_w;
945  dtext->var_values[VAR_MAX_GLYPH_H] = dtext->max_glyph_h;
946  dtext->var_values[VAR_MAX_GLYPH_A] = dtext->var_values[VAR_ASCENT ] = y_max;
947  dtext->var_values[VAR_MAX_GLYPH_D] = dtext->var_values[VAR_DESCENT] = y_min;
948 
949  dtext->var_values[VAR_LINE_H] = dtext->var_values[VAR_LH] = dtext->max_glyph_h;
950 
951  dtext->x = dtext->var_values[VAR_X] = av_expr_eval(dtext->x_pexpr, dtext->var_values, &dtext->prng);
952  dtext->y = dtext->var_values[VAR_Y] = av_expr_eval(dtext->y_pexpr, dtext->var_values, &dtext->prng);
953  dtext->x = dtext->var_values[VAR_X] = av_expr_eval(dtext->x_pexpr, dtext->var_values, &dtext->prng);
954  dtext->draw = av_expr_eval(dtext->draw_pexpr, dtext->var_values, &dtext->prng);
955 
956  if(!dtext->draw)
957  return 0;
958 
959  box_w = FFMIN(width - 1 , max_text_line_w);
960  box_h = FFMIN(height - 1, y + dtext->max_glyph_h);
961 
962  /* draw box */
963  if (dtext->draw_box)
964  ff_blend_rectangle(&dtext->dc, &dtext->boxcolor,
965  picref->data, picref->linesize, width, height,
966  dtext->x, dtext->y, box_w, box_h);
967 
968  if (dtext->shadowx || dtext->shadowy) {
969  if ((ret = draw_glyphs(dtext, picref, width, height, dtext->shadowcolor.rgba,
970  &dtext->shadowcolor, dtext->shadowx, dtext->shadowy)) < 0)
971  return ret;
972  }
973 
974  if ((ret = draw_glyphs(dtext, picref, width, height, dtext->fontcolor.rgba,
975  &dtext->fontcolor, 0, 0)) < 0)
976  return ret;
977 
978  return 0;
979 }
980 
981 static int filter_frame(AVFilterLink *inlink, AVFilterBufferRef *frame)
982 {
983  AVFilterContext *ctx = inlink->dst;
984  AVFilterLink *outlink = ctx->outputs[0];
985  DrawTextContext *dtext = ctx->priv;
986  int ret;
987 
988  if (dtext->reload)
989  if ((ret = load_textfile(ctx)) < 0)
990  return ret;
991 
992  dtext->var_values[VAR_T] = frame->pts == AV_NOPTS_VALUE ?
993  NAN : frame->pts * av_q2d(inlink->time_base);
994 
995  draw_text(ctx, frame, frame->video->w, frame->video->h);
996 
997  av_log(ctx, AV_LOG_DEBUG, "n:%d t:%f text_w:%d text_h:%d x:%d y:%d\n",
998  (int)dtext->var_values[VAR_N], dtext->var_values[VAR_T],
999  (int)dtext->var_values[VAR_TEXT_W], (int)dtext->var_values[VAR_TEXT_H],
1000  dtext->x, dtext->y);
1001 
1002  dtext->var_values[VAR_N] += 1.0;
1003 
1004  return ff_filter_frame(outlink, frame);
1005 }
1006 
1007 static const AVFilterPad avfilter_vf_drawtext_inputs[] = {
1008  {
1009  .name = "default",
1010  .type = AVMEDIA_TYPE_VIDEO,
1011  .get_video_buffer = ff_null_get_video_buffer,
1012  .filter_frame = filter_frame,
1013  .config_props = config_input,
1014  .min_perms = AV_PERM_WRITE |
1015  AV_PERM_READ,
1016  },
1017  { NULL }
1018 };
1019 
1020 static const AVFilterPad avfilter_vf_drawtext_outputs[] = {
1021  {
1022  .name = "default",
1023  .type = AVMEDIA_TYPE_VIDEO,
1024  },
1025  { NULL }
1026 };
1027 
1028 AVFilter avfilter_vf_drawtext = {
1029  .name = "drawtext",
1030  .description = NULL_IF_CONFIG_SMALL("Draw text on top of video frames using libfreetype library."),
1031  .priv_size = sizeof(DrawTextContext),
1032  .init = init,
1033  .uninit = uninit,
1034  .query_formats = query_formats,
1035 
1036  .inputs = avfilter_vf_drawtext_inputs,
1037  .outputs = avfilter_vf_drawtext_outputs,
1038  .process_command = command,
1039  .priv_class = &drawtext_class,
1040 };
Generated on Sat May 25 2013 03:58:45 for FFmpeg by   doxygen 1.8.2