You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1047 lines
37KB

  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. * @file
  24. * drawtext filter, based on the original vhook/drawtext.c
  25. * filter by Gustavo Sverzut Barbieri
  26. */
  27. #include <sys/time.h>
  28. #include <time.h>
  29. #include "config.h"
  30. #include "libavutil/avstring.h"
  31. #include "libavutil/bprint.h"
  32. #include "libavutil/common.h"
  33. #include "libavutil/file.h"
  34. #include "libavutil/eval.h"
  35. #include "libavutil/opt.h"
  36. #include "libavutil/random_seed.h"
  37. #include "libavutil/parseutils.h"
  38. #include "libavutil/timecode.h"
  39. #include "libavutil/tree.h"
  40. #include "libavutil/lfg.h"
  41. #include "avfilter.h"
  42. #include "drawutils.h"
  43. #include "formats.h"
  44. #include "internal.h"
  45. #include "video.h"
  46. #include <ft2build.h>
  47. #include <freetype/config/ftheader.h>
  48. #include FT_FREETYPE_H
  49. #include FT_GLYPH_H
  50. #if CONFIG_FONTCONFIG
  51. #include <fontconfig/fontconfig.h>
  52. #endif
  53. static const char *const var_names[] = {
  54. "dar",
  55. "hsub", "vsub",
  56. "line_h", "lh", ///< line height, same as max_glyph_h
  57. "main_h", "h", "H", ///< height of the input video
  58. "main_w", "w", "W", ///< width of the input video
  59. "max_glyph_a", "ascent", ///< max glyph ascent
  60. "max_glyph_d", "descent", ///< min glyph descent
  61. "max_glyph_h", ///< max glyph height
  62. "max_glyph_w", ///< max glyph width
  63. "n", ///< number of frame
  64. "sar",
  65. "t", ///< timestamp expressed in seconds
  66. "text_h", "th", ///< height of the rendered text
  67. "text_w", "tw", ///< width of the rendered text
  68. "x",
  69. "y",
  70. "pict_type",
  71. NULL
  72. };
  73. static const char *const fun2_names[] = {
  74. "rand"
  75. };
  76. static double drand(void *opaque, double min, double max)
  77. {
  78. return min + (max-min) / UINT_MAX * av_lfg_get(opaque);
  79. }
  80. typedef double (*eval_func2)(void *, double a, double b);
  81. static const eval_func2 fun2[] = {
  82. drand,
  83. NULL
  84. };
  85. enum var_name {
  86. VAR_DAR,
  87. VAR_HSUB, VAR_VSUB,
  88. VAR_LINE_H, VAR_LH,
  89. VAR_MAIN_H, VAR_h, VAR_H,
  90. VAR_MAIN_W, VAR_w, VAR_W,
  91. VAR_MAX_GLYPH_A, VAR_ASCENT,
  92. VAR_MAX_GLYPH_D, VAR_DESCENT,
  93. VAR_MAX_GLYPH_H,
  94. VAR_MAX_GLYPH_W,
  95. VAR_N,
  96. VAR_SAR,
  97. VAR_T,
  98. VAR_TEXT_H, VAR_TH,
  99. VAR_TEXT_W, VAR_TW,
  100. VAR_X,
  101. VAR_Y,
  102. VAR_PICT_TYPE,
  103. VAR_VARS_NB
  104. };
  105. enum expansion_mode {
  106. EXP_NONE,
  107. EXP_NORMAL,
  108. EXP_STRFTIME,
  109. };
  110. typedef struct {
  111. const AVClass *class;
  112. enum expansion_mode exp_mode; ///< expansion mode to use for the text
  113. int reinit; ///< tells if the filter is being reinited
  114. uint8_t *fontfile; ///< font to be used
  115. uint8_t *text; ///< text to be drawn
  116. AVBPrint expanded_text; ///< used to contain the expanded text
  117. int ft_load_flags; ///< flags used for loading fonts, see FT_LOAD_*
  118. FT_Vector *positions; ///< positions for each element in the text
  119. size_t nb_positions; ///< number of elements of positions array
  120. char *textfile; ///< file with text to be drawn
  121. int x; ///< x position to start drawing text
  122. int y; ///< y position to start drawing text
  123. int max_glyph_w; ///< max glyph width
  124. int max_glyph_h; ///< max glyph height
  125. int shadowx, shadowy;
  126. unsigned int fontsize; ///< font size to use
  127. char *fontcolor_string; ///< font color as string
  128. char *boxcolor_string; ///< box color as string
  129. char *shadowcolor_string; ///< shadow color as string
  130. short int draw_box; ///< draw box around text - true or false
  131. int use_kerning; ///< font kerning is used - true/false
  132. int tabsize; ///< tab size
  133. int fix_bounds; ///< do we let it go out of frame bounds - t/f
  134. FFDrawContext dc;
  135. FFDrawColor fontcolor; ///< foreground color
  136. FFDrawColor shadowcolor; ///< shadow color
  137. FFDrawColor boxcolor; ///< background color
  138. FT_Library library; ///< freetype font library handle
  139. FT_Face face; ///< freetype font face handle
  140. struct AVTreeNode *glyphs; ///< rendered glyphs, stored using the UTF-32 char code
  141. char *x_expr; ///< expression for x position
  142. char *y_expr; ///< expression for y position
  143. AVExpr *x_pexpr, *y_pexpr; ///< parsed expressions for x and y
  144. int64_t basetime; ///< base pts time in the real world for display
  145. double var_values[VAR_VARS_NB];
  146. char *draw_expr; ///< expression for draw
  147. AVExpr *draw_pexpr; ///< parsed expression for draw
  148. int draw; ///< set to zero to prevent drawing
  149. AVLFG prng; ///< random
  150. char *tc_opt_string; ///< specified timecode option string
  151. AVRational tc_rate; ///< frame rate for timecode
  152. AVTimecode tc; ///< timecode context
  153. int tc24hmax; ///< 1 if timecode is wrapped to 24 hours, 0 otherwise
  154. int reload; ///< reload text file for each frame
  155. } DrawTextContext;
  156. #define OFFSET(x) offsetof(DrawTextContext, x)
  157. #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
  158. static const AVOption drawtext_options[]= {
  159. {"fontfile", "set font file", OFFSET(fontfile), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
  160. {"text", "set text", OFFSET(text), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
  161. {"textfile", "set text file", OFFSET(textfile), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
  162. {"fontcolor", "set foreground color", OFFSET(fontcolor_string), AV_OPT_TYPE_STRING, {.str="black"}, CHAR_MIN, CHAR_MAX, FLAGS},
  163. {"boxcolor", "set box color", OFFSET(boxcolor_string), AV_OPT_TYPE_STRING, {.str="white"}, CHAR_MIN, CHAR_MAX, FLAGS},
  164. {"shadowcolor", "set shadow color", OFFSET(shadowcolor_string), AV_OPT_TYPE_STRING, {.str="black"}, CHAR_MIN, CHAR_MAX, FLAGS},
  165. {"box", "set box", OFFSET(draw_box), AV_OPT_TYPE_INT, {.i64=0}, 0, 1 , FLAGS},
  166. {"fontsize", "set font size", OFFSET(fontsize), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX , FLAGS},
  167. {"x", "set x expression", OFFSET(x_expr), AV_OPT_TYPE_STRING, {.str="0"}, CHAR_MIN, CHAR_MAX, FLAGS},
  168. {"y", "set y expression", OFFSET(y_expr), AV_OPT_TYPE_STRING, {.str="0"}, CHAR_MIN, CHAR_MAX, FLAGS},
  169. {"shadowx", "set x", OFFSET(shadowx), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX , FLAGS},
  170. {"shadowy", "set y", OFFSET(shadowy), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX , FLAGS},
  171. {"tabsize", "set tab size", OFFSET(tabsize), AV_OPT_TYPE_INT, {.i64=4}, 0, INT_MAX , FLAGS},
  172. {"basetime", "set base time", OFFSET(basetime), AV_OPT_TYPE_INT64, {.i64=AV_NOPTS_VALUE}, INT64_MIN, INT64_MAX , FLAGS},
  173. {"draw", "if false do not draw", OFFSET(draw_expr), AV_OPT_TYPE_STRING, {.str="1"}, CHAR_MIN, CHAR_MAX, FLAGS},
  174. {"expansion", "set the expansion mode", OFFSET(exp_mode), AV_OPT_TYPE_INT, {.i64=EXP_NORMAL}, 0, 2, FLAGS, "expansion"},
  175. {"none", "set no expansion", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_NONE}, 0, 0, FLAGS, "expansion"},
  176. {"normal", "set normal expansion", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_NORMAL}, 0, 0, FLAGS, "expansion"},
  177. {"strftime", "set strftime expansion (deprecated)", OFFSET(exp_mode), AV_OPT_TYPE_CONST, {.i64=EXP_STRFTIME}, 0, 0, FLAGS, "expansion"},
  178. {"timecode", "set initial timecode", OFFSET(tc_opt_string), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS},
  179. {"tc24hmax", "set 24 hours max (timecode only)", OFFSET(tc24hmax), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS},
  180. {"timecode_rate", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
  181. {"r", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
  182. {"rate", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX, FLAGS},
  183. {"reload", "reload text file for each frame", OFFSET(reload), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS},
  184. {"fix_bounds", "if true, check and fix text coords to avoid clipping", OFFSET(fix_bounds), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, FLAGS},
  185. /* FT_LOAD_* flags */
  186. { "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" },
  187. { "default", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_DEFAULT }, .flags = FLAGS, .unit = "ft_load_flags" },
  188. { "no_scale", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_SCALE }, .flags = FLAGS, .unit = "ft_load_flags" },
  189. { "no_hinting", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_HINTING }, .flags = FLAGS, .unit = "ft_load_flags" },
  190. { "render", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_RENDER }, .flags = FLAGS, .unit = "ft_load_flags" },
  191. { "no_bitmap", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_BITMAP }, .flags = FLAGS, .unit = "ft_load_flags" },
  192. { "vertical_layout", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_VERTICAL_LAYOUT }, .flags = FLAGS, .unit = "ft_load_flags" },
  193. { "force_autohint", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_FORCE_AUTOHINT }, .flags = FLAGS, .unit = "ft_load_flags" },
  194. { "crop_bitmap", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_CROP_BITMAP }, .flags = FLAGS, .unit = "ft_load_flags" },
  195. { "pedantic", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_PEDANTIC }, .flags = FLAGS, .unit = "ft_load_flags" },
  196. { "ignore_global_advance_width", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_IGNORE_GLOBAL_ADVANCE_WIDTH }, .flags = FLAGS, .unit = "ft_load_flags" },
  197. { "no_recurse", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_RECURSE }, .flags = FLAGS, .unit = "ft_load_flags" },
  198. { "ignore_transform", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_IGNORE_TRANSFORM }, .flags = FLAGS, .unit = "ft_load_flags" },
  199. { "monochrome", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_MONOCHROME }, .flags = FLAGS, .unit = "ft_load_flags" },
  200. { "linear_design", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_LINEAR_DESIGN }, .flags = FLAGS, .unit = "ft_load_flags" },
  201. { "no_autohint", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_AUTOHINT }, .flags = FLAGS, .unit = "ft_load_flags" },
  202. { NULL},
  203. };
  204. AVFILTER_DEFINE_CLASS(drawtext);
  205. #undef __FTERRORS_H__
  206. #define FT_ERROR_START_LIST {
  207. #define FT_ERRORDEF(e, v, s) { (e), (s) },
  208. #define FT_ERROR_END_LIST { 0, NULL } };
  209. struct ft_error
  210. {
  211. int err;
  212. const char *err_msg;
  213. } static ft_errors[] =
  214. #include FT_ERRORS_H
  215. #define FT_ERRMSG(e) ft_errors[e].err_msg
  216. typedef struct {
  217. FT_Glyph *glyph;
  218. uint32_t code;
  219. FT_Bitmap bitmap; ///< array holding bitmaps of font
  220. FT_BBox bbox;
  221. int advance;
  222. int bitmap_left;
  223. int bitmap_top;
  224. } Glyph;
  225. static int glyph_cmp(void *key, const void *b)
  226. {
  227. const Glyph *a = key, *bb = b;
  228. int64_t diff = (int64_t)a->code - (int64_t)bb->code;
  229. return diff > 0 ? 1 : diff < 0 ? -1 : 0;
  230. }
  231. /**
  232. * Load glyphs corresponding to the UTF-32 codepoint code.
  233. */
  234. static int load_glyph(AVFilterContext *ctx, Glyph **glyph_ptr, uint32_t code)
  235. {
  236. DrawTextContext *dtext = ctx->priv;
  237. Glyph *glyph;
  238. struct AVTreeNode *node = NULL;
  239. int ret;
  240. /* load glyph into dtext->face->glyph */
  241. if (FT_Load_Char(dtext->face, code, dtext->ft_load_flags))
  242. return AVERROR(EINVAL);
  243. /* save glyph */
  244. if (!(glyph = av_mallocz(sizeof(*glyph))) ||
  245. !(glyph->glyph = av_mallocz(sizeof(*glyph->glyph)))) {
  246. ret = AVERROR(ENOMEM);
  247. goto error;
  248. }
  249. glyph->code = code;
  250. if (FT_Get_Glyph(dtext->face->glyph, glyph->glyph)) {
  251. ret = AVERROR(EINVAL);
  252. goto error;
  253. }
  254. glyph->bitmap = dtext->face->glyph->bitmap;
  255. glyph->bitmap_left = dtext->face->glyph->bitmap_left;
  256. glyph->bitmap_top = dtext->face->glyph->bitmap_top;
  257. glyph->advance = dtext->face->glyph->advance.x >> 6;
  258. /* measure text height to calculate text_height (or the maximum text height) */
  259. FT_Glyph_Get_CBox(*glyph->glyph, ft_glyph_bbox_pixels, &glyph->bbox);
  260. /* cache the newly created glyph */
  261. if (!(node = av_tree_node_alloc())) {
  262. ret = AVERROR(ENOMEM);
  263. goto error;
  264. }
  265. av_tree_insert(&dtext->glyphs, glyph, glyph_cmp, &node);
  266. if (glyph_ptr)
  267. *glyph_ptr = glyph;
  268. return 0;
  269. error:
  270. if (glyph)
  271. av_freep(&glyph->glyph);
  272. av_freep(&glyph);
  273. av_freep(&node);
  274. return ret;
  275. }
  276. static int load_font_file(AVFilterContext *ctx, const char *path, int index,
  277. const char **error)
  278. {
  279. DrawTextContext *dtext = ctx->priv;
  280. int err;
  281. err = FT_New_Face(dtext->library, path, index, &dtext->face);
  282. if (err) {
  283. *error = FT_ERRMSG(err);
  284. return AVERROR(EINVAL);
  285. }
  286. return 0;
  287. }
  288. #if CONFIG_FONTCONFIG
  289. static int load_font_fontconfig(AVFilterContext *ctx, const char **error)
  290. {
  291. DrawTextContext *dtext = ctx->priv;
  292. FcConfig *fontconfig;
  293. FcPattern *pattern, *fpat;
  294. FcResult result = FcResultMatch;
  295. FcChar8 *filename;
  296. int err, index;
  297. double size;
  298. fontconfig = FcInitLoadConfigAndFonts();
  299. if (!fontconfig) {
  300. *error = "impossible to init fontconfig\n";
  301. return AVERROR(EINVAL);
  302. }
  303. pattern = FcNameParse(dtext->fontfile ? dtext->fontfile :
  304. (uint8_t *)(intptr_t)"default");
  305. if (!pattern) {
  306. *error = "could not parse fontconfig pattern";
  307. return AVERROR(EINVAL);
  308. }
  309. if (!FcConfigSubstitute(fontconfig, pattern, FcMatchPattern)) {
  310. *error = "could not substitue fontconfig options"; /* very unlikely */
  311. return AVERROR(EINVAL);
  312. }
  313. FcDefaultSubstitute(pattern);
  314. fpat = FcFontMatch(fontconfig, pattern, &result);
  315. if (!fpat || result != FcResultMatch) {
  316. *error = "impossible to find a matching font";
  317. return AVERROR(EINVAL);
  318. }
  319. if (FcPatternGetString (fpat, FC_FILE, 0, &filename) != FcResultMatch ||
  320. FcPatternGetInteger(fpat, FC_INDEX, 0, &index ) != FcResultMatch ||
  321. FcPatternGetDouble (fpat, FC_SIZE, 0, &size ) != FcResultMatch) {
  322. *error = "impossible to find font information";
  323. return AVERROR(EINVAL);
  324. }
  325. av_log(ctx, AV_LOG_INFO, "Using \"%s\"\n", filename);
  326. if (!dtext->fontsize)
  327. dtext->fontsize = size + 0.5;
  328. err = load_font_file(ctx, filename, index, error);
  329. if (err)
  330. return err;
  331. FcPatternDestroy(fpat);
  332. FcPatternDestroy(pattern);
  333. FcConfigDestroy(fontconfig);
  334. return 0;
  335. }
  336. #endif
  337. static int load_font(AVFilterContext *ctx)
  338. {
  339. DrawTextContext *dtext = ctx->priv;
  340. int err;
  341. const char *error = "unknown error\n";
  342. /* load the face, and set up the encoding, which is by default UTF-8 */
  343. err = load_font_file(ctx, dtext->fontfile, 0, &error);
  344. if (!err)
  345. return 0;
  346. #if CONFIG_FONTCONFIG
  347. err = load_font_fontconfig(ctx, &error);
  348. if (!err)
  349. return 0;
  350. #endif
  351. av_log(ctx, AV_LOG_ERROR, "Could not load font \"%s\": %s\n",
  352. dtext->fontfile, error);
  353. return err;
  354. }
  355. static int load_textfile(AVFilterContext *ctx)
  356. {
  357. DrawTextContext *dtext = ctx->priv;
  358. int err;
  359. uint8_t *textbuf;
  360. size_t textbuf_size;
  361. if ((err = av_file_map(dtext->textfile, &textbuf, &textbuf_size, 0, ctx)) < 0) {
  362. av_log(ctx, AV_LOG_ERROR,
  363. "The text file '%s' could not be read or is empty\n",
  364. dtext->textfile);
  365. return err;
  366. }
  367. if (!(dtext->text = av_realloc(dtext->text, textbuf_size + 1)))
  368. return AVERROR(ENOMEM);
  369. memcpy(dtext->text, textbuf, textbuf_size);
  370. dtext->text[textbuf_size] = 0;
  371. av_file_unmap(textbuf, textbuf_size);
  372. return 0;
  373. }
  374. static av_cold int init(AVFilterContext *ctx)
  375. {
  376. int err;
  377. DrawTextContext *dtext = ctx->priv;
  378. Glyph *glyph;
  379. if (!dtext->fontfile && !CONFIG_FONTCONFIG) {
  380. av_log(ctx, AV_LOG_ERROR, "No font filename provided\n");
  381. return AVERROR(EINVAL);
  382. }
  383. if (dtext->textfile) {
  384. if (dtext->text) {
  385. av_log(ctx, AV_LOG_ERROR,
  386. "Both text and text file provided. Please provide only one\n");
  387. return AVERROR(EINVAL);
  388. }
  389. if ((err = load_textfile(ctx)) < 0)
  390. return err;
  391. }
  392. if (dtext->reload && !dtext->textfile)
  393. av_log(ctx, AV_LOG_WARNING, "No file to reload\n");
  394. if (dtext->tc_opt_string) {
  395. int ret = av_timecode_init_from_string(&dtext->tc, dtext->tc_rate,
  396. dtext->tc_opt_string, ctx);
  397. if (ret < 0)
  398. return ret;
  399. if (dtext->tc24hmax)
  400. dtext->tc.flags |= AV_TIMECODE_FLAG_24HOURSMAX;
  401. if (!dtext->text)
  402. dtext->text = av_strdup("");
  403. }
  404. if (!dtext->text) {
  405. av_log(ctx, AV_LOG_ERROR,
  406. "Either text, a valid file or a timecode must be provided\n");
  407. return AVERROR(EINVAL);
  408. }
  409. if ((err = av_parse_color(dtext->fontcolor.rgba, dtext->fontcolor_string, -1, ctx))) {
  410. av_log(ctx, AV_LOG_ERROR,
  411. "Invalid font color '%s'\n", dtext->fontcolor_string);
  412. return err;
  413. }
  414. if ((err = av_parse_color(dtext->boxcolor.rgba, dtext->boxcolor_string, -1, ctx))) {
  415. av_log(ctx, AV_LOG_ERROR,
  416. "Invalid box color '%s'\n", dtext->boxcolor_string);
  417. return err;
  418. }
  419. if ((err = av_parse_color(dtext->shadowcolor.rgba, dtext->shadowcolor_string, -1, ctx))) {
  420. av_log(ctx, AV_LOG_ERROR,
  421. "Invalid shadow color '%s'\n", dtext->shadowcolor_string);
  422. return err;
  423. }
  424. if ((err = FT_Init_FreeType(&(dtext->library)))) {
  425. av_log(ctx, AV_LOG_ERROR,
  426. "Could not load FreeType: %s\n", FT_ERRMSG(err));
  427. return AVERROR(EINVAL);
  428. }
  429. err = load_font(ctx);
  430. if (err)
  431. return err;
  432. if (!dtext->fontsize)
  433. dtext->fontsize = 16;
  434. if ((err = FT_Set_Pixel_Sizes(dtext->face, 0, dtext->fontsize))) {
  435. av_log(ctx, AV_LOG_ERROR, "Could not set font size to %d pixels: %s\n",
  436. dtext->fontsize, FT_ERRMSG(err));
  437. return AVERROR(EINVAL);
  438. }
  439. dtext->use_kerning = FT_HAS_KERNING(dtext->face);
  440. /* load the fallback glyph with code 0 */
  441. load_glyph(ctx, NULL, 0);
  442. /* set the tabsize in pixels */
  443. if ((err = load_glyph(ctx, &glyph, ' ')) < 0) {
  444. av_log(ctx, AV_LOG_ERROR, "Could not set tabsize.\n");
  445. return err;
  446. }
  447. dtext->tabsize *= glyph->advance;
  448. if (dtext->exp_mode == EXP_STRFTIME &&
  449. (strchr(dtext->text, '%') || strchr(dtext->text, '\\')))
  450. av_log(ctx, AV_LOG_WARNING, "expansion=strftime is deprecated.\n");
  451. av_bprint_init(&dtext->expanded_text, 0, AV_BPRINT_SIZE_UNLIMITED);
  452. return 0;
  453. }
  454. static int query_formats(AVFilterContext *ctx)
  455. {
  456. ff_set_common_formats(ctx, ff_draw_supported_pixel_formats(0));
  457. return 0;
  458. }
  459. static int glyph_enu_free(void *opaque, void *elem)
  460. {
  461. Glyph *glyph = elem;
  462. FT_Done_Glyph(*glyph->glyph);
  463. av_freep(&glyph->glyph);
  464. av_free(elem);
  465. return 0;
  466. }
  467. static av_cold void uninit(AVFilterContext *ctx)
  468. {
  469. DrawTextContext *dtext = ctx->priv;
  470. av_expr_free(dtext->x_pexpr); dtext->x_pexpr = NULL;
  471. av_expr_free(dtext->y_pexpr); dtext->y_pexpr = NULL;
  472. av_expr_free(dtext->draw_pexpr); dtext->draw_pexpr = NULL;
  473. av_freep(&dtext->positions);
  474. dtext->nb_positions = 0;
  475. av_tree_enumerate(dtext->glyphs, NULL, NULL, glyph_enu_free);
  476. av_tree_destroy(dtext->glyphs);
  477. dtext->glyphs = NULL;
  478. FT_Done_Face(dtext->face);
  479. FT_Done_FreeType(dtext->library);
  480. av_bprint_finalize(&dtext->expanded_text, NULL);
  481. }
  482. static inline int is_newline(uint32_t c)
  483. {
  484. return c == '\n' || c == '\r' || c == '\f' || c == '\v';
  485. }
  486. static int config_input(AVFilterLink *inlink)
  487. {
  488. AVFilterContext *ctx = inlink->dst;
  489. DrawTextContext *dtext = ctx->priv;
  490. int ret;
  491. ff_draw_init(&dtext->dc, inlink->format, 0);
  492. ff_draw_color(&dtext->dc, &dtext->fontcolor, dtext->fontcolor.rgba);
  493. ff_draw_color(&dtext->dc, &dtext->shadowcolor, dtext->shadowcolor.rgba);
  494. ff_draw_color(&dtext->dc, &dtext->boxcolor, dtext->boxcolor.rgba);
  495. dtext->var_values[VAR_w] = dtext->var_values[VAR_W] = dtext->var_values[VAR_MAIN_W] = inlink->w;
  496. dtext->var_values[VAR_h] = dtext->var_values[VAR_H] = dtext->var_values[VAR_MAIN_H] = inlink->h;
  497. dtext->var_values[VAR_SAR] = inlink->sample_aspect_ratio.num ? av_q2d(inlink->sample_aspect_ratio) : 1;
  498. dtext->var_values[VAR_DAR] = (double)inlink->w / inlink->h * dtext->var_values[VAR_SAR];
  499. dtext->var_values[VAR_HSUB] = 1 << dtext->dc.hsub_max;
  500. dtext->var_values[VAR_VSUB] = 1 << dtext->dc.vsub_max;
  501. dtext->var_values[VAR_X] = NAN;
  502. dtext->var_values[VAR_Y] = NAN;
  503. if (!dtext->reinit)
  504. dtext->var_values[VAR_N] = 0;
  505. dtext->var_values[VAR_T] = NAN;
  506. av_lfg_init(&dtext->prng, av_get_random_seed());
  507. if ((ret = av_expr_parse(&dtext->x_pexpr, dtext->x_expr, var_names,
  508. NULL, NULL, fun2_names, fun2, 0, ctx)) < 0 ||
  509. (ret = av_expr_parse(&dtext->y_pexpr, dtext->y_expr, var_names,
  510. NULL, NULL, fun2_names, fun2, 0, ctx)) < 0 ||
  511. (ret = av_expr_parse(&dtext->draw_pexpr, dtext->draw_expr, var_names,
  512. NULL, NULL, fun2_names, fun2, 0, ctx)) < 0)
  513. return AVERROR(EINVAL);
  514. return 0;
  515. }
  516. static int command(AVFilterContext *ctx, const char *cmd, const char *arg, char *res, int res_len, int flags)
  517. {
  518. DrawTextContext *dtext = ctx->priv;
  519. if (!strcmp(cmd, "reinit")) {
  520. int ret;
  521. uninit(ctx);
  522. dtext->reinit = 1;
  523. if ((ret = init(ctx)) < 0)
  524. return ret;
  525. return config_input(ctx->inputs[0]);
  526. }
  527. return AVERROR(ENOSYS);
  528. }
  529. static int func_pict_type(AVFilterContext *ctx, AVBPrint *bp,
  530. char *fct, unsigned argc, char **argv, int tag)
  531. {
  532. DrawTextContext *dtext = ctx->priv;
  533. av_bprintf(bp, "%c", av_get_picture_type_char(dtext->var_values[VAR_PICT_TYPE]));
  534. return 0;
  535. }
  536. static int func_pts(AVFilterContext *ctx, AVBPrint *bp,
  537. char *fct, unsigned argc, char **argv, int tag)
  538. {
  539. DrawTextContext *dtext = ctx->priv;
  540. av_bprintf(bp, "%.6f", dtext->var_values[VAR_T]);
  541. return 0;
  542. }
  543. static int func_frame_num(AVFilterContext *ctx, AVBPrint *bp,
  544. char *fct, unsigned argc, char **argv, int tag)
  545. {
  546. DrawTextContext *dtext = ctx->priv;
  547. av_bprintf(bp, "%d", (int)dtext->var_values[VAR_N]);
  548. return 0;
  549. }
  550. #if !HAVE_LOCALTIME_R
  551. static void localtime_r(const time_t *t, struct tm *tm)
  552. {
  553. *tm = *localtime(t);
  554. }
  555. #endif
  556. static int func_strftime(AVFilterContext *ctx, AVBPrint *bp,
  557. char *fct, unsigned argc, char **argv, int tag)
  558. {
  559. const char *fmt = argc ? argv[0] : "%Y-%m-%d %H:%M:%S";
  560. time_t now;
  561. struct tm tm;
  562. time(&now);
  563. if (tag == 'L')
  564. localtime_r(&now, &tm);
  565. else
  566. tm = *gmtime(&now);
  567. av_bprint_strftime(bp, fmt, &tm);
  568. return 0;
  569. }
  570. static int func_eval_expr(AVFilterContext *ctx, AVBPrint *bp,
  571. char *fct, unsigned argc, char **argv, int tag)
  572. {
  573. DrawTextContext *dtext = ctx->priv;
  574. double res;
  575. int ret;
  576. ret = av_expr_parse_and_eval(&res, argv[0], var_names, dtext->var_values,
  577. NULL, NULL, fun2_names, fun2,
  578. &dtext->prng, 0, ctx);
  579. if (ret < 0)
  580. av_log(ctx, AV_LOG_ERROR,
  581. "Expression '%s' for the expr text expansion function is not valid\n",
  582. argv[0]);
  583. else
  584. av_bprintf(bp, "%f", res);
  585. return ret;
  586. }
  587. static const struct drawtext_function {
  588. const char *name;
  589. unsigned argc_min, argc_max;
  590. int tag; /** opaque argument to func */
  591. int (*func)(AVFilterContext *, AVBPrint *, char *, unsigned, char **, int);
  592. } functions[] = {
  593. { "expr", 1, 1, 0, func_eval_expr },
  594. { "e", 1, 1, 0, func_eval_expr },
  595. { "pict_type", 0, 0, 0, func_pict_type },
  596. { "pts", 0, 0, 0, func_pts },
  597. { "gmtime", 0, 1, 'G', func_strftime },
  598. { "localtime", 0, 1, 'L', func_strftime },
  599. { "frame_num", 0, 0, 0, func_frame_num },
  600. { "n", 0, 0, 0, func_frame_num },
  601. };
  602. static int eval_function(AVFilterContext *ctx, AVBPrint *bp, char *fct,
  603. unsigned argc, char **argv)
  604. {
  605. unsigned i;
  606. for (i = 0; i < FF_ARRAY_ELEMS(functions); i++) {
  607. if (strcmp(fct, functions[i].name))
  608. continue;
  609. if (argc < functions[i].argc_min) {
  610. av_log(ctx, AV_LOG_ERROR, "%%{%s} requires at least %d arguments\n",
  611. fct, functions[i].argc_min);
  612. return AVERROR(EINVAL);
  613. }
  614. if (argc > functions[i].argc_max) {
  615. av_log(ctx, AV_LOG_ERROR, "%%{%s} requires at most %d arguments\n",
  616. fct, functions[i].argc_max);
  617. return AVERROR(EINVAL);
  618. }
  619. break;
  620. }
  621. if (i >= FF_ARRAY_ELEMS(functions)) {
  622. av_log(ctx, AV_LOG_ERROR, "%%{%s} is not known\n", fct);
  623. return AVERROR(EINVAL);
  624. }
  625. return functions[i].func(ctx, bp, fct, argc, argv, functions[i].tag);
  626. }
  627. static int expand_function(AVFilterContext *ctx, AVBPrint *bp, char **rtext)
  628. {
  629. const char *text = *rtext;
  630. char *argv[16] = { NULL };
  631. unsigned argc = 0, i;
  632. int ret;
  633. if (*text != '{') {
  634. av_log(ctx, AV_LOG_ERROR, "Stray %% near '%s'\n", text);
  635. return AVERROR(EINVAL);
  636. }
  637. text++;
  638. while (1) {
  639. if (!(argv[argc++] = av_get_token(&text, ":}"))) {
  640. ret = AVERROR(ENOMEM);
  641. goto end;
  642. }
  643. if (!*text) {
  644. av_log(ctx, AV_LOG_ERROR, "Unterminated %%{} near '%s'\n", *rtext);
  645. ret = AVERROR(EINVAL);
  646. goto end;
  647. }
  648. if (argc == FF_ARRAY_ELEMS(argv))
  649. av_freep(&argv[--argc]); /* error will be caught later */
  650. if (*text == '}')
  651. break;
  652. text++;
  653. }
  654. if ((ret = eval_function(ctx, bp, argv[0], argc - 1, argv + 1)) < 0)
  655. goto end;
  656. ret = 0;
  657. *rtext = (char *)text + 1;
  658. end:
  659. for (i = 0; i < argc; i++)
  660. av_freep(&argv[i]);
  661. return ret;
  662. }
  663. static int expand_text(AVFilterContext *ctx)
  664. {
  665. DrawTextContext *dtext = ctx->priv;
  666. char *text = dtext->text;
  667. AVBPrint *bp = &dtext->expanded_text;
  668. int ret;
  669. av_bprint_clear(bp);
  670. while (*text) {
  671. if (*text == '\\' && text[1]) {
  672. av_bprint_chars(bp, text[1], 1);
  673. text += 2;
  674. } else if (*text == '%') {
  675. text++;
  676. if ((ret = expand_function(ctx, bp, &text)) < 0)
  677. return ret;
  678. } else {
  679. av_bprint_chars(bp, *text, 1);
  680. text++;
  681. }
  682. }
  683. if (!av_bprint_is_complete(bp))
  684. return AVERROR(ENOMEM);
  685. return 0;
  686. }
  687. static int draw_glyphs(DrawTextContext *dtext, AVFrame *frame,
  688. int width, int height, const uint8_t rgbcolor[4], FFDrawColor *color, int x, int y)
  689. {
  690. char *text = dtext->expanded_text.str;
  691. uint32_t code = 0;
  692. int i, x1, y1;
  693. uint8_t *p;
  694. Glyph *glyph = NULL;
  695. for (i = 0, p = text; *p; i++) {
  696. Glyph dummy = { 0 };
  697. GET_UTF8(code, *p++, continue;);
  698. /* skip new line chars, just go to new line */
  699. if (code == '\n' || code == '\r' || code == '\t')
  700. continue;
  701. dummy.code = code;
  702. glyph = av_tree_find(dtext->glyphs, &dummy, (void *)glyph_cmp, NULL);
  703. if (glyph->bitmap.pixel_mode != FT_PIXEL_MODE_MONO &&
  704. glyph->bitmap.pixel_mode != FT_PIXEL_MODE_GRAY)
  705. return AVERROR(EINVAL);
  706. x1 = dtext->positions[i].x+dtext->x+x;
  707. y1 = dtext->positions[i].y+dtext->y+y;
  708. ff_blend_mask(&dtext->dc, color,
  709. frame->data, frame->linesize, width, height,
  710. glyph->bitmap.buffer, glyph->bitmap.pitch,
  711. glyph->bitmap.width, glyph->bitmap.rows,
  712. glyph->bitmap.pixel_mode == FT_PIXEL_MODE_MONO ? 0 : 3,
  713. 0, x1, y1);
  714. }
  715. return 0;
  716. }
  717. static int draw_text(AVFilterContext *ctx, AVFrame *frame,
  718. int width, int height)
  719. {
  720. DrawTextContext *dtext = ctx->priv;
  721. AVFilterLink *inlink = ctx->inputs[0];
  722. uint32_t code = 0, prev_code = 0;
  723. int x = 0, y = 0, i = 0, ret;
  724. int max_text_line_w = 0, len;
  725. int box_w, box_h;
  726. char *text = dtext->text;
  727. uint8_t *p;
  728. int y_min = 32000, y_max = -32000;
  729. int x_min = 32000, x_max = -32000;
  730. FT_Vector delta;
  731. Glyph *glyph = NULL, *prev_glyph = NULL;
  732. Glyph dummy = { 0 };
  733. time_t now = time(0);
  734. struct tm ltime;
  735. AVBPrint *bp = &dtext->expanded_text;
  736. av_bprint_clear(bp);
  737. if(dtext->basetime != AV_NOPTS_VALUE)
  738. now= frame->pts*av_q2d(ctx->inputs[0]->time_base) + dtext->basetime/1000000;
  739. switch (dtext->exp_mode) {
  740. case EXP_NONE:
  741. av_bprintf(bp, "%s", dtext->text);
  742. break;
  743. case EXP_NORMAL:
  744. if ((ret = expand_text(ctx)) < 0)
  745. return ret;
  746. break;
  747. case EXP_STRFTIME:
  748. localtime_r(&now, &ltime);
  749. av_bprint_strftime(bp, dtext->text, &ltime);
  750. break;
  751. }
  752. if (dtext->tc_opt_string) {
  753. char tcbuf[AV_TIMECODE_STR_SIZE];
  754. av_timecode_make_string(&dtext->tc, tcbuf, inlink->frame_count);
  755. av_bprint_clear(bp);
  756. av_bprintf(bp, "%s%s", dtext->text, tcbuf);
  757. }
  758. if (!av_bprint_is_complete(bp))
  759. return AVERROR(ENOMEM);
  760. text = dtext->expanded_text.str;
  761. if ((len = dtext->expanded_text.len) > dtext->nb_positions) {
  762. if (!(dtext->positions =
  763. av_realloc(dtext->positions, len*sizeof(*dtext->positions))))
  764. return AVERROR(ENOMEM);
  765. dtext->nb_positions = len;
  766. }
  767. x = 0;
  768. y = 0;
  769. /* load and cache glyphs */
  770. for (i = 0, p = text; *p; i++) {
  771. GET_UTF8(code, *p++, continue;);
  772. /* get glyph */
  773. dummy.code = code;
  774. glyph = av_tree_find(dtext->glyphs, &dummy, glyph_cmp, NULL);
  775. if (!glyph) {
  776. load_glyph(ctx, &glyph, code);
  777. }
  778. y_min = FFMIN(glyph->bbox.yMin, y_min);
  779. y_max = FFMAX(glyph->bbox.yMax, y_max);
  780. x_min = FFMIN(glyph->bbox.xMin, x_min);
  781. x_max = FFMAX(glyph->bbox.xMax, x_max);
  782. }
  783. dtext->max_glyph_h = y_max - y_min;
  784. dtext->max_glyph_w = x_max - x_min;
  785. /* compute and save position for each glyph */
  786. glyph = NULL;
  787. for (i = 0, p = text; *p; i++) {
  788. GET_UTF8(code, *p++, continue;);
  789. /* skip the \n in the sequence \r\n */
  790. if (prev_code == '\r' && code == '\n')
  791. continue;
  792. prev_code = code;
  793. if (is_newline(code)) {
  794. max_text_line_w = FFMAX(max_text_line_w, x);
  795. y += dtext->max_glyph_h;
  796. x = 0;
  797. continue;
  798. }
  799. /* get glyph */
  800. prev_glyph = glyph;
  801. dummy.code = code;
  802. glyph = av_tree_find(dtext->glyphs, &dummy, glyph_cmp, NULL);
  803. /* kerning */
  804. if (dtext->use_kerning && prev_glyph && glyph->code) {
  805. FT_Get_Kerning(dtext->face, prev_glyph->code, glyph->code,
  806. ft_kerning_default, &delta);
  807. x += delta.x >> 6;
  808. }
  809. /* save position */
  810. dtext->positions[i].x = x + glyph->bitmap_left;
  811. dtext->positions[i].y = y - glyph->bitmap_top + y_max;
  812. if (code == '\t') x = (x / dtext->tabsize + 1)*dtext->tabsize;
  813. else x += glyph->advance;
  814. }
  815. max_text_line_w = FFMAX(x, max_text_line_w);
  816. dtext->var_values[VAR_TW] = dtext->var_values[VAR_TEXT_W] = max_text_line_w;
  817. dtext->var_values[VAR_TH] = dtext->var_values[VAR_TEXT_H] = y + dtext->max_glyph_h;
  818. dtext->var_values[VAR_MAX_GLYPH_W] = dtext->max_glyph_w;
  819. dtext->var_values[VAR_MAX_GLYPH_H] = dtext->max_glyph_h;
  820. dtext->var_values[VAR_MAX_GLYPH_A] = dtext->var_values[VAR_ASCENT ] = y_max;
  821. dtext->var_values[VAR_MAX_GLYPH_D] = dtext->var_values[VAR_DESCENT] = y_min;
  822. dtext->var_values[VAR_LINE_H] = dtext->var_values[VAR_LH] = dtext->max_glyph_h;
  823. dtext->x = dtext->var_values[VAR_X] = av_expr_eval(dtext->x_pexpr, dtext->var_values, &dtext->prng);
  824. dtext->y = dtext->var_values[VAR_Y] = av_expr_eval(dtext->y_pexpr, dtext->var_values, &dtext->prng);
  825. dtext->x = dtext->var_values[VAR_X] = av_expr_eval(dtext->x_pexpr, dtext->var_values, &dtext->prng);
  826. dtext->draw = av_expr_eval(dtext->draw_pexpr, dtext->var_values, &dtext->prng);
  827. if(!dtext->draw)
  828. return 0;
  829. box_w = FFMIN(width - 1 , max_text_line_w);
  830. box_h = FFMIN(height - 1, y + dtext->max_glyph_h);
  831. /* draw box */
  832. if (dtext->draw_box)
  833. ff_blend_rectangle(&dtext->dc, &dtext->boxcolor,
  834. frame->data, frame->linesize, width, height,
  835. dtext->x, dtext->y, box_w, box_h);
  836. if (dtext->shadowx || dtext->shadowy) {
  837. if ((ret = draw_glyphs(dtext, frame, width, height, dtext->shadowcolor.rgba,
  838. &dtext->shadowcolor, dtext->shadowx, dtext->shadowy)) < 0)
  839. return ret;
  840. }
  841. if ((ret = draw_glyphs(dtext, frame, width, height, dtext->fontcolor.rgba,
  842. &dtext->fontcolor, 0, 0)) < 0)
  843. return ret;
  844. return 0;
  845. }
  846. static int filter_frame(AVFilterLink *inlink, AVFrame *frame)
  847. {
  848. AVFilterContext *ctx = inlink->dst;
  849. AVFilterLink *outlink = ctx->outputs[0];
  850. DrawTextContext *dtext = ctx->priv;
  851. int ret;
  852. if (dtext->reload)
  853. if ((ret = load_textfile(ctx)) < 0)
  854. return ret;
  855. dtext->var_values[VAR_N] = inlink->frame_count;
  856. dtext->var_values[VAR_T] = frame->pts == AV_NOPTS_VALUE ?
  857. NAN : frame->pts * av_q2d(inlink->time_base);
  858. dtext->var_values[VAR_PICT_TYPE] = frame->pict_type;
  859. draw_text(ctx, frame, frame->width, frame->height);
  860. av_log(ctx, AV_LOG_DEBUG, "n:%d t:%f text_w:%d text_h:%d x:%d y:%d\n",
  861. (int)dtext->var_values[VAR_N], dtext->var_values[VAR_T],
  862. (int)dtext->var_values[VAR_TEXT_W], (int)dtext->var_values[VAR_TEXT_H],
  863. dtext->x, dtext->y);
  864. return ff_filter_frame(outlink, frame);
  865. }
  866. static const AVFilterPad avfilter_vf_drawtext_inputs[] = {
  867. {
  868. .name = "default",
  869. .type = AVMEDIA_TYPE_VIDEO,
  870. .get_video_buffer = ff_null_get_video_buffer,
  871. .filter_frame = filter_frame,
  872. .config_props = config_input,
  873. .needs_writable = 1,
  874. },
  875. { NULL }
  876. };
  877. static const AVFilterPad avfilter_vf_drawtext_outputs[] = {
  878. {
  879. .name = "default",
  880. .type = AVMEDIA_TYPE_VIDEO,
  881. },
  882. { NULL }
  883. };
  884. AVFilter avfilter_vf_drawtext = {
  885. .name = "drawtext",
  886. .description = NULL_IF_CONFIG_SMALL("Draw text on top of video frames using libfreetype library."),
  887. .priv_size = sizeof(DrawTextContext),
  888. .priv_class = &drawtext_class,
  889. .init = init,
  890. .uninit = uninit,
  891. .query_formats = query_formats,
  892. .inputs = avfilter_vf_drawtext_inputs,
  893. .outputs = avfilter_vf_drawtext_outputs,
  894. .process_command = command,
  895. };