vf_drawtext.c 34 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009
  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 Libav.
  7. *
  8. * Libav 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. * Libav 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 Libav; 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 "config.h"
  28. #include <sys/types.h>
  29. #include <sys/time.h>
  30. #include <sys/stat.h>
  31. #include <time.h>
  32. #include <unistd.h>
  33. #if CONFIG_LIBFONTCONFIG
  34. #include <fontconfig/fontconfig.h>
  35. #endif
  36. #include "libavutil/colorspace.h"
  37. #include "libavutil/common.h"
  38. #include "libavutil/file.h"
  39. #include "libavutil/eval.h"
  40. #include "libavutil/opt.h"
  41. #include "libavutil/mathematics.h"
  42. #include "libavutil/random_seed.h"
  43. #include "libavutil/parseutils.h"
  44. #include "libavutil/pixdesc.h"
  45. #include "libavutil/time_internal.h"
  46. #include "libavutil/tree.h"
  47. #include "libavutil/lfg.h"
  48. #include "avfilter.h"
  49. #include "drawutils.h"
  50. #include "formats.h"
  51. #include "internal.h"
  52. #include "video.h"
  53. #include <ft2build.h>
  54. #include FT_FREETYPE_H
  55. #include FT_GLYPH_H
  56. static const char *const var_names[] = {
  57. "E",
  58. "PHI",
  59. "PI",
  60. "main_w", "W", ///< width of the main video
  61. "main_h", "H", ///< height of the main video
  62. "text_w", "w", ///< width of the overlay text
  63. "text_h", "h", ///< height of the overlay text
  64. "x",
  65. "y",
  66. "n", ///< number of processed frames
  67. "t", ///< timestamp expressed in seconds
  68. NULL
  69. };
  70. static const char *const fun2_names[] = {
  71. "rand"
  72. };
  73. static double drand(void *opaque, double min, double max)
  74. {
  75. return min + (max-min) / UINT_MAX * av_lfg_get(opaque);
  76. }
  77. typedef double (*eval_func2)(void *, double a, double b);
  78. static const eval_func2 fun2[] = {
  79. drand,
  80. NULL
  81. };
  82. enum var_name {
  83. VAR_E,
  84. VAR_PHI,
  85. VAR_PI,
  86. VAR_MAIN_W, VAR_MW,
  87. VAR_MAIN_H, VAR_MH,
  88. VAR_TEXT_W, VAR_TW,
  89. VAR_TEXT_H, VAR_TH,
  90. VAR_X,
  91. VAR_Y,
  92. VAR_N,
  93. VAR_T,
  94. VAR_VARS_NB
  95. };
  96. typedef struct DrawTextContext {
  97. const AVClass *class;
  98. #if CONFIG_LIBFONTCONFIG
  99. uint8_t *font; ///< font to be used
  100. #endif
  101. uint8_t *fontfile; ///< font to be used
  102. uint8_t *text; ///< text to be drawn
  103. uint8_t *expanded_text; ///< used to contain the strftime()-expanded text
  104. size_t expanded_text_size; ///< size in bytes of the expanded_text buffer
  105. int ft_load_flags; ///< flags used for loading fonts, see FT_LOAD_*
  106. FT_Vector *positions; ///< positions for each element in the text
  107. size_t nb_positions; ///< number of elements of positions array
  108. char *textfile; ///< file with text to be drawn
  109. int x, y; ///< position to start drawing text
  110. int w, h; ///< dimension of the text block
  111. int shadowx, shadowy;
  112. unsigned int fontsize; ///< font size to use
  113. char *fontcolor_string; ///< font color as string
  114. char *boxcolor_string; ///< box color as string
  115. char *shadowcolor_string; ///< shadow color as string
  116. uint8_t fontcolor[4]; ///< foreground color
  117. uint8_t boxcolor[4]; ///< background color
  118. uint8_t shadowcolor[4]; ///< shadow color
  119. uint8_t fontcolor_rgba[4]; ///< foreground color in RGBA
  120. uint8_t boxcolor_rgba[4]; ///< background color in RGBA
  121. uint8_t shadowcolor_rgba[4]; ///< shadow color in RGBA
  122. short int draw_box; ///< draw box around text - true or false
  123. int use_kerning; ///< font kerning is used - true/false
  124. int tabsize; ///< tab size
  125. int fix_bounds; ///< do we let it go out of frame bounds - t/f
  126. FT_Library library; ///< freetype font library handle
  127. FT_Face face; ///< freetype font face handle
  128. struct AVTreeNode *glyphs; ///< rendered glyphs, stored using the UTF-32 char code
  129. int hsub, vsub; ///< chroma subsampling values
  130. int is_packed_rgb;
  131. int pixel_step[4]; ///< distance in bytes between the component of each pixel
  132. uint8_t rgba_map[4]; ///< map RGBA offsets to the positions in the packed RGBA format
  133. uint8_t *box_line[4]; ///< line used for filling the box background
  134. char *x_expr, *y_expr;
  135. AVExpr *x_pexpr, *y_pexpr; ///< parsed expressions for x and y
  136. double var_values[VAR_VARS_NB];
  137. char *d_expr;
  138. AVExpr *d_pexpr;
  139. int draw; ///< set to zero to prevent drawing
  140. char *a_expr;
  141. AVExpr *a_pexpr;
  142. int alpha;
  143. AVLFG prng; ///< random
  144. } DrawTextContext;
  145. #define OFFSET(x) offsetof(DrawTextContext, x)
  146. #define FLAGS AV_OPT_FLAG_VIDEO_PARAM
  147. static const AVOption drawtext_options[]= {
  148. #if CONFIG_LIBFONTCONFIG
  149. { "font", "Font name", OFFSET(font), AV_OPT_TYPE_STRING, { .str = "Sans" }, .flags = FLAGS },
  150. #endif
  151. { "fontfile", NULL, OFFSET(fontfile), AV_OPT_TYPE_STRING, .flags = FLAGS },
  152. { "text", NULL, OFFSET(text), AV_OPT_TYPE_STRING, .flags = FLAGS },
  153. { "textfile", NULL, OFFSET(textfile), AV_OPT_TYPE_STRING, .flags = FLAGS },
  154. { "fontcolor", NULL, OFFSET(fontcolor_string), AV_OPT_TYPE_STRING, { .str = "black" }, .flags = FLAGS },
  155. { "boxcolor", NULL, OFFSET(boxcolor_string), AV_OPT_TYPE_STRING, { .str = "white" }, .flags = FLAGS },
  156. { "shadowcolor", NULL, OFFSET(shadowcolor_string), AV_OPT_TYPE_STRING, { .str = "black" }, .flags = FLAGS },
  157. { "box", NULL, OFFSET(draw_box), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, FLAGS },
  158. { "fontsize", NULL, OFFSET(fontsize), AV_OPT_TYPE_INT, { .i64 = 16 }, 1, 1024, FLAGS },
  159. { "x", NULL, OFFSET(x_expr), AV_OPT_TYPE_STRING, { .str = "0" }, .flags = FLAGS },
  160. { "y", NULL, OFFSET(y_expr), AV_OPT_TYPE_STRING, { .str = "0" }, .flags = FLAGS },
  161. { "shadowx", NULL, OFFSET(shadowx), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FLAGS },
  162. { "shadowy", NULL, OFFSET(shadowy), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, FLAGS },
  163. { "tabsize", NULL, OFFSET(tabsize), AV_OPT_TYPE_INT, { .i64 = 4 }, 0, INT_MAX, FLAGS },
  164. { "draw", "if false do not draw", OFFSET(d_expr), AV_OPT_TYPE_STRING, { .str = "1" }, .flags = FLAGS },
  165. { "alpha", "apply alpha while rendering", OFFSET(a_expr), AV_OPT_TYPE_STRING, { .str = "1" }, .flags = FLAGS },
  166. { "fix_bounds", "if true, check and fix text coords to avoid clipping",
  167. OFFSET(fix_bounds), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, FLAGS },
  168. /* FT_LOAD_* flags */
  169. { "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" },
  170. { "default", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_DEFAULT }, .flags = FLAGS, .unit = "ft_load_flags" },
  171. { "no_scale", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_SCALE }, .flags = FLAGS, .unit = "ft_load_flags" },
  172. { "no_hinting", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_HINTING }, .flags = FLAGS, .unit = "ft_load_flags" },
  173. { "render", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_RENDER }, .flags = FLAGS, .unit = "ft_load_flags" },
  174. { "no_bitmap", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_BITMAP }, .flags = FLAGS, .unit = "ft_load_flags" },
  175. { "vertical_layout", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_VERTICAL_LAYOUT }, .flags = FLAGS, .unit = "ft_load_flags" },
  176. { "force_autohint", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_FORCE_AUTOHINT }, .flags = FLAGS, .unit = "ft_load_flags" },
  177. { "crop_bitmap", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_CROP_BITMAP }, .flags = FLAGS, .unit = "ft_load_flags" },
  178. { "pedantic", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_PEDANTIC }, .flags = FLAGS, .unit = "ft_load_flags" },
  179. { "ignore_global_advance_width", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_IGNORE_GLOBAL_ADVANCE_WIDTH }, .flags = FLAGS, .unit = "ft_load_flags" },
  180. { "no_recurse", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_RECURSE }, .flags = FLAGS, .unit = "ft_load_flags" },
  181. { "ignore_transform", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_IGNORE_TRANSFORM }, .flags = FLAGS, .unit = "ft_load_flags" },
  182. { "monochrome", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_MONOCHROME }, .flags = FLAGS, .unit = "ft_load_flags" },
  183. { "linear_design", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_LINEAR_DESIGN }, .flags = FLAGS, .unit = "ft_load_flags" },
  184. { "no_autohint", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FT_LOAD_NO_AUTOHINT }, .flags = FLAGS, .unit = "ft_load_flags" },
  185. { NULL},
  186. };
  187. static const char *drawtext_get_name(void *ctx)
  188. {
  189. return "drawtext";
  190. }
  191. static const AVClass drawtext_class = {
  192. "DrawTextContext",
  193. drawtext_get_name,
  194. drawtext_options
  195. };
  196. #undef __FTERRORS_H__
  197. #define FT_ERROR_START_LIST {
  198. #define FT_ERRORDEF(e, v, s) { (e), (s) },
  199. #define FT_ERROR_END_LIST { 0, NULL } };
  200. struct ft_error
  201. {
  202. int err;
  203. const char *err_msg;
  204. } static ft_errors[] =
  205. #include FT_ERRORS_H
  206. #define FT_ERRMSG(e) ft_errors[e].err_msg
  207. typedef struct Glyph {
  208. FT_Glyph *glyph;
  209. uint32_t code;
  210. FT_Bitmap bitmap; ///< array holding bitmaps of font
  211. FT_BBox bbox;
  212. int advance;
  213. int bitmap_left;
  214. int bitmap_top;
  215. } Glyph;
  216. static int glyph_cmp(void *key, const void *b)
  217. {
  218. const Glyph *a = key, *bb = b;
  219. int64_t diff = (int64_t)a->code - (int64_t)bb->code;
  220. return diff > 0 ? 1 : diff < 0 ? -1 : 0;
  221. }
  222. /**
  223. * Load glyphs corresponding to the UTF-32 codepoint code.
  224. */
  225. static int load_glyph(AVFilterContext *ctx, Glyph **glyph_ptr, uint32_t code)
  226. {
  227. DrawTextContext *s = ctx->priv;
  228. Glyph *glyph;
  229. struct AVTreeNode *node = NULL;
  230. int ret;
  231. /* load glyph into s->face->glyph */
  232. if (FT_Load_Char(s->face, code, s->ft_load_flags))
  233. return AVERROR(EINVAL);
  234. /* save glyph */
  235. if (!(glyph = av_mallocz(sizeof(*glyph))) ||
  236. !(glyph->glyph = av_mallocz(sizeof(*glyph->glyph)))) {
  237. ret = AVERROR(ENOMEM);
  238. goto error;
  239. }
  240. glyph->code = code;
  241. if (FT_Get_Glyph(s->face->glyph, glyph->glyph)) {
  242. ret = AVERROR(EINVAL);
  243. goto error;
  244. }
  245. glyph->bitmap = s->face->glyph->bitmap;
  246. glyph->bitmap_left = s->face->glyph->bitmap_left;
  247. glyph->bitmap_top = s->face->glyph->bitmap_top;
  248. glyph->advance = s->face->glyph->advance.x >> 6;
  249. /* measure text height to calculate text_height (or the maximum text height) */
  250. FT_Glyph_Get_CBox(*glyph->glyph, ft_glyph_bbox_pixels, &glyph->bbox);
  251. /* cache the newly created glyph */
  252. if (!(node = av_tree_node_alloc())) {
  253. ret = AVERROR(ENOMEM);
  254. goto error;
  255. }
  256. av_tree_insert(&s->glyphs, glyph, glyph_cmp, &node);
  257. if (glyph_ptr)
  258. *glyph_ptr = glyph;
  259. return 0;
  260. error:
  261. if (glyph)
  262. av_freep(&glyph->glyph);
  263. av_freep(&glyph);
  264. av_freep(&node);
  265. return ret;
  266. }
  267. static int parse_font(AVFilterContext *ctx)
  268. {
  269. DrawTextContext *s = ctx->priv;
  270. #if !CONFIG_LIBFONTCONFIG
  271. if (!s->fontfile) {
  272. av_log(ctx, AV_LOG_ERROR, "No font filename provided\n");
  273. return AVERROR(EINVAL);
  274. }
  275. return 0;
  276. #else
  277. FcPattern *pat, *best;
  278. FcResult result = FcResultMatch;
  279. FcBool fc_bool;
  280. FcChar8* fc_string;
  281. int err = AVERROR(ENOENT);
  282. if (s->fontfile)
  283. return 0;
  284. if (!FcInit())
  285. return AVERROR_UNKNOWN;
  286. if (!(pat = FcPatternCreate()))
  287. return AVERROR(ENOMEM);
  288. FcPatternAddString(pat, FC_FAMILY, s->font);
  289. FcPatternAddBool(pat, FC_OUTLINE, FcTrue);
  290. FcPatternAddDouble(pat, FC_SIZE, (double)s->fontsize);
  291. FcDefaultSubstitute(pat);
  292. if (!FcConfigSubstitute(NULL, pat, FcMatchPattern)) {
  293. FcPatternDestroy(pat);
  294. return AVERROR(ENOMEM);
  295. }
  296. best = FcFontMatch(NULL, pat, &result);
  297. FcPatternDestroy(pat);
  298. if (!best || result == FcResultNoMatch) {
  299. av_log(ctx, AV_LOG_ERROR,
  300. "Cannot find a valid font for the family %s\n",
  301. s->font);
  302. goto fail;
  303. }
  304. if (FcPatternGetBool(best, FC_OUTLINE, 0, &fc_bool) != FcResultMatch ||
  305. !fc_bool) {
  306. av_log(ctx, AV_LOG_ERROR, "Outline not available for %s\n",
  307. s->font);
  308. goto fail;
  309. }
  310. if (FcPatternGetString(best, FC_FAMILY, 0, &fc_string) != FcResultMatch) {
  311. av_log(ctx, AV_LOG_ERROR, "No matches for %s\n",
  312. s->font);
  313. goto fail;
  314. }
  315. if (FcPatternGetString(best, FC_FILE, 0, &fc_string) != FcResultMatch) {
  316. av_log(ctx, AV_LOG_ERROR, "No file path for %s\n",
  317. s->font);
  318. goto fail;
  319. }
  320. s->fontfile = av_strdup(fc_string);
  321. if (!s->fontfile)
  322. err = AVERROR(ENOMEM);
  323. else
  324. err = 0;
  325. fail:
  326. FcPatternDestroy(best);
  327. return err;
  328. #endif
  329. }
  330. static av_cold int init(AVFilterContext *ctx)
  331. {
  332. int err;
  333. DrawTextContext *s = ctx->priv;
  334. Glyph *glyph;
  335. if ((err = parse_font(ctx)) < 0)
  336. return err;
  337. if (s->textfile) {
  338. uint8_t *textbuf;
  339. size_t textbuf_size;
  340. if (s->text) {
  341. av_log(ctx, AV_LOG_ERROR,
  342. "Both text and text file provided. Please provide only one\n");
  343. return AVERROR(EINVAL);
  344. }
  345. if ((err = av_file_map(s->textfile, &textbuf, &textbuf_size, 0, ctx)) < 0) {
  346. av_log(ctx, AV_LOG_ERROR,
  347. "The text file '%s' could not be read or is empty\n",
  348. s->textfile);
  349. return err;
  350. }
  351. if (textbuf_size > SIZE_MAX - 1 ||
  352. !(s->text = av_malloc(textbuf_size + 1))) {
  353. av_file_unmap(textbuf, textbuf_size);
  354. return AVERROR(ENOMEM);
  355. }
  356. memcpy(s->text, textbuf, textbuf_size);
  357. s->text[textbuf_size] = 0;
  358. av_file_unmap(textbuf, textbuf_size);
  359. }
  360. if (!s->text) {
  361. av_log(ctx, AV_LOG_ERROR,
  362. "Either text or a valid file must be provided\n");
  363. return AVERROR(EINVAL);
  364. }
  365. if ((err = av_parse_color(s->fontcolor_rgba, s->fontcolor_string, -1, ctx))) {
  366. av_log(ctx, AV_LOG_ERROR,
  367. "Invalid font color '%s'\n", s->fontcolor_string);
  368. return err;
  369. }
  370. if ((err = av_parse_color(s->boxcolor_rgba, s->boxcolor_string, -1, ctx))) {
  371. av_log(ctx, AV_LOG_ERROR,
  372. "Invalid box color '%s'\n", s->boxcolor_string);
  373. return err;
  374. }
  375. if ((err = av_parse_color(s->shadowcolor_rgba, s->shadowcolor_string, -1, ctx))) {
  376. av_log(ctx, AV_LOG_ERROR,
  377. "Invalid shadow color '%s'\n", s->shadowcolor_string);
  378. return err;
  379. }
  380. if ((err = FT_Init_FreeType(&(s->library)))) {
  381. av_log(ctx, AV_LOG_ERROR,
  382. "Could not load FreeType: %s\n", FT_ERRMSG(err));
  383. return AVERROR(EINVAL);
  384. }
  385. /* load the face, and set up the encoding, which is by default UTF-8 */
  386. if ((err = FT_New_Face(s->library, s->fontfile, 0, &s->face))) {
  387. av_log(ctx, AV_LOG_ERROR, "Could not load fontface from file '%s': %s\n",
  388. s->fontfile, FT_ERRMSG(err));
  389. return AVERROR(EINVAL);
  390. }
  391. if ((err = FT_Set_Pixel_Sizes(s->face, 0, s->fontsize))) {
  392. av_log(ctx, AV_LOG_ERROR, "Could not set font size to %d pixels: %s\n",
  393. s->fontsize, FT_ERRMSG(err));
  394. return AVERROR(EINVAL);
  395. }
  396. s->use_kerning = FT_HAS_KERNING(s->face);
  397. /* load the fallback glyph with code 0 */
  398. load_glyph(ctx, NULL, 0);
  399. /* set the tabsize in pixels */
  400. if ((err = load_glyph(ctx, &glyph, ' ') < 0)) {
  401. av_log(ctx, AV_LOG_ERROR, "Could not set tabsize.\n");
  402. return err;
  403. }
  404. s->tabsize *= glyph->advance;
  405. return 0;
  406. }
  407. static int query_formats(AVFilterContext *ctx)
  408. {
  409. static const enum AVPixelFormat pix_fmts[] = {
  410. AV_PIX_FMT_ARGB, AV_PIX_FMT_RGBA,
  411. AV_PIX_FMT_ABGR, AV_PIX_FMT_BGRA,
  412. AV_PIX_FMT_RGB24, AV_PIX_FMT_BGR24,
  413. AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV444P,
  414. AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV411P,
  415. AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV440P,
  416. AV_PIX_FMT_NONE
  417. };
  418. ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
  419. return 0;
  420. }
  421. static int glyph_enu_free(void *opaque, void *elem)
  422. {
  423. av_free(elem);
  424. return 0;
  425. }
  426. static av_cold void uninit(AVFilterContext *ctx)
  427. {
  428. DrawTextContext *s = ctx->priv;
  429. int i;
  430. av_expr_free(s->x_pexpr);
  431. av_expr_free(s->y_pexpr);
  432. av_expr_free(s->d_pexpr);
  433. s->x_pexpr = s->y_pexpr = s->d_pexpr = NULL;
  434. av_freep(&s->expanded_text);
  435. av_freep(&s->positions);
  436. av_tree_enumerate(s->glyphs, NULL, NULL, glyph_enu_free);
  437. av_tree_destroy(s->glyphs);
  438. s->glyphs = 0;
  439. FT_Done_Face(s->face);
  440. FT_Done_FreeType(s->library);
  441. for (i = 0; i < 4; i++) {
  442. av_freep(&s->box_line[i]);
  443. s->pixel_step[i] = 0;
  444. }
  445. }
  446. static inline int is_newline(uint32_t c)
  447. {
  448. return c == '\n' || c == '\r' || c == '\f' || c == '\v';
  449. }
  450. static int expand_strftime(DrawTextContext *s)
  451. {
  452. struct tm ltime;
  453. time_t now = time(0);
  454. uint8_t *buf = s->expanded_text;
  455. int buf_size = s->expanded_text_size;
  456. if (!buf)
  457. buf_size = 2 * strlen(s->text) + 1;
  458. localtime_r(&now, &ltime);
  459. while ((buf = av_realloc(buf, buf_size))) {
  460. *buf = 1;
  461. if (strftime(buf, buf_size, s->text, &ltime) != 0 || *buf == 0)
  462. break;
  463. buf_size *= 2;
  464. }
  465. if (!buf)
  466. return AVERROR(ENOMEM);
  467. s->expanded_text = buf;
  468. s->expanded_text_size = buf_size;
  469. return 0;
  470. }
  471. static int dtext_prepare_text(AVFilterContext *ctx)
  472. {
  473. DrawTextContext *s = ctx->priv;
  474. uint32_t code = 0, prev_code = 0;
  475. int x = 0, y = 0, i = 0, ret;
  476. int text_height, baseline;
  477. char *text;
  478. uint8_t *p;
  479. int str_w = 0, len;
  480. int y_min = 32000, y_max = -32000;
  481. FT_Vector delta;
  482. Glyph *glyph = NULL, *prev_glyph = NULL;
  483. Glyph dummy = { 0 };
  484. int width = ctx->inputs[0]->w;
  485. int height = ctx->inputs[0]->h;
  486. ret = expand_strftime(s);
  487. if (ret < 0)
  488. return ret;
  489. text = s->expanded_text ? s->expanded_text : s->text;
  490. if ((len = strlen(text)) > s->nb_positions) {
  491. FT_Vector *p = av_realloc(s->positions,
  492. len * sizeof(*s->positions));
  493. if (!p) {
  494. av_freep(s->positions);
  495. s->nb_positions = 0;
  496. return AVERROR(ENOMEM);
  497. } else {
  498. s->positions = p;
  499. s->nb_positions = len;
  500. }
  501. }
  502. /* load and cache glyphs */
  503. for (i = 0, p = text; *p; i++) {
  504. GET_UTF8(code, *p++, continue;);
  505. /* get glyph */
  506. dummy.code = code;
  507. glyph = av_tree_find(s->glyphs, &dummy, glyph_cmp, NULL);
  508. if (!glyph) {
  509. ret = load_glyph(ctx, &glyph, code);
  510. if (ret)
  511. return ret;
  512. }
  513. y_min = FFMIN(glyph->bbox.yMin, y_min);
  514. y_max = FFMAX(glyph->bbox.yMax, y_max);
  515. }
  516. text_height = y_max - y_min;
  517. baseline = y_max;
  518. /* compute and save position for each glyph */
  519. glyph = NULL;
  520. for (i = 0, p = text; *p; i++) {
  521. GET_UTF8(code, *p++, continue;);
  522. /* skip the \n in the sequence \r\n */
  523. if (prev_code == '\r' && code == '\n')
  524. continue;
  525. prev_code = code;
  526. if (is_newline(code)) {
  527. str_w = FFMAX(str_w, x - s->x);
  528. y += text_height;
  529. x = 0;
  530. continue;
  531. }
  532. /* get glyph */
  533. prev_glyph = glyph;
  534. dummy.code = code;
  535. glyph = av_tree_find(s->glyphs, &dummy, glyph_cmp, NULL);
  536. /* kerning */
  537. if (s->use_kerning && prev_glyph && glyph->code) {
  538. FT_Get_Kerning(s->face, prev_glyph->code, glyph->code,
  539. ft_kerning_default, &delta);
  540. x += delta.x >> 6;
  541. }
  542. if (x + glyph->bbox.xMax >= width) {
  543. str_w = FFMAX(str_w, x);
  544. y += text_height;
  545. x = 0;
  546. }
  547. /* save position */
  548. s->positions[i].x = x + glyph->bitmap_left;
  549. s->positions[i].y = y - glyph->bitmap_top + baseline;
  550. if (code == '\t') x = (x / s->tabsize + 1)*s->tabsize;
  551. else x += glyph->advance;
  552. }
  553. str_w = FFMIN(width - 1, FFMAX(str_w, x));
  554. y = FFMIN(y + text_height, height - 1);
  555. s->w = str_w;
  556. s->var_values[VAR_TEXT_W] = s->var_values[VAR_TW] = s->w;
  557. s->h = y;
  558. s->var_values[VAR_TEXT_H] = s->var_values[VAR_TH] = s->h;
  559. return 0;
  560. }
  561. static int config_input(AVFilterLink *inlink)
  562. {
  563. AVFilterContext *ctx = inlink->dst;
  564. DrawTextContext *s = ctx->priv;
  565. const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format);
  566. int ret;
  567. s->hsub = pix_desc->log2_chroma_w;
  568. s->vsub = pix_desc->log2_chroma_h;
  569. s->var_values[VAR_E ] = M_E;
  570. s->var_values[VAR_PHI] = M_PHI;
  571. s->var_values[VAR_PI ] = M_PI;
  572. s->var_values[VAR_MAIN_W] =
  573. s->var_values[VAR_MW] = ctx->inputs[0]->w;
  574. s->var_values[VAR_MAIN_H] =
  575. s->var_values[VAR_MH] = ctx->inputs[0]->h;
  576. s->var_values[VAR_X] = 0;
  577. s->var_values[VAR_Y] = 0;
  578. s->var_values[VAR_T] = NAN;
  579. av_lfg_init(&s->prng, av_get_random_seed());
  580. av_expr_free(s->x_pexpr);
  581. av_expr_free(s->y_pexpr);
  582. av_expr_free(s->d_pexpr);
  583. s->x_pexpr = s->y_pexpr = s->d_pexpr = NULL;
  584. if ((ret = av_expr_parse(&s->x_pexpr, s->x_expr, var_names,
  585. NULL, NULL, fun2_names, fun2, 0, ctx)) < 0 ||
  586. (ret = av_expr_parse(&s->y_pexpr, s->y_expr, var_names,
  587. NULL, NULL, fun2_names, fun2, 0, ctx)) < 0 ||
  588. (ret = av_expr_parse(&s->d_pexpr, s->d_expr, var_names,
  589. NULL, NULL, fun2_names, fun2, 0, ctx)) < 0 ||
  590. (ret = av_expr_parse(&s->a_pexpr, s->a_expr, var_names,
  591. NULL, NULL, fun2_names, fun2, 0, ctx)) < 0)
  592. return AVERROR(EINVAL);
  593. if ((ret =
  594. ff_fill_line_with_color(s->box_line, s->pixel_step,
  595. inlink->w, s->boxcolor,
  596. inlink->format, s->boxcolor_rgba,
  597. &s->is_packed_rgb, s->rgba_map)) < 0)
  598. return ret;
  599. if (!s->is_packed_rgb) {
  600. uint8_t *rgba = s->fontcolor_rgba;
  601. s->fontcolor[0] = RGB_TO_Y_CCIR(rgba[0], rgba[1], rgba[2]);
  602. s->fontcolor[1] = RGB_TO_U_CCIR(rgba[0], rgba[1], rgba[2], 0);
  603. s->fontcolor[2] = RGB_TO_V_CCIR(rgba[0], rgba[1], rgba[2], 0);
  604. s->fontcolor[3] = rgba[3];
  605. rgba = s->shadowcolor_rgba;
  606. s->shadowcolor[0] = RGB_TO_Y_CCIR(rgba[0], rgba[1], rgba[2]);
  607. s->shadowcolor[1] = RGB_TO_U_CCIR(rgba[0], rgba[1], rgba[2], 0);
  608. s->shadowcolor[2] = RGB_TO_V_CCIR(rgba[0], rgba[1], rgba[2], 0);
  609. s->shadowcolor[3] = rgba[3];
  610. }
  611. s->draw = 1;
  612. return dtext_prepare_text(ctx);
  613. }
  614. #define GET_BITMAP_VAL(r, c) \
  615. bitmap->pixel_mode == FT_PIXEL_MODE_MONO ? \
  616. (bitmap->buffer[(r) * bitmap->pitch + ((c)>>3)] & (0x80 >> ((c)&7))) * 255 : \
  617. bitmap->buffer[(r) * bitmap->pitch + (c)]
  618. #define SET_PIXEL_YUV(frame, yuva_color, val, x, y, hsub, vsub) { \
  619. luma_pos = ((x) ) + ((y) ) * frame->linesize[0]; \
  620. alpha = yuva_color[3] * alpha_mul * (val) * 129 / 255; \
  621. frame->data[0][luma_pos] = (alpha * yuva_color[0] + (255*255*129 - alpha) * frame->data[0][luma_pos] ) >> 23; \
  622. if (((x) & ((1<<(hsub)) - 1)) == 0 && ((y) & ((1<<(vsub)) - 1)) == 0) {\
  623. chroma_pos1 = ((x) >> (hsub)) + ((y) >> (vsub)) * frame->linesize[1]; \
  624. chroma_pos2 = ((x) >> (hsub)) + ((y) >> (vsub)) * frame->linesize[2]; \
  625. frame->data[1][chroma_pos1] = (alpha * yuva_color[1] + (255*255*129 - alpha) * frame->data[1][chroma_pos1]) >> 23; \
  626. frame->data[2][chroma_pos2] = (alpha * yuva_color[2] + (255*255*129 - alpha) * frame->data[2][chroma_pos2]) >> 23; \
  627. }\
  628. }
  629. static inline int draw_glyph_yuv(AVFrame *frame, FT_Bitmap *bitmap, unsigned int x,
  630. unsigned int y, unsigned int width, unsigned int height,
  631. const uint8_t yuva_color[4], int hsub, int vsub,
  632. int alpha_mul)
  633. {
  634. int r, c, alpha;
  635. unsigned int luma_pos, chroma_pos1, chroma_pos2;
  636. uint8_t src_val;
  637. for (r = 0; r < bitmap->rows && r+y < height; r++) {
  638. for (c = 0; c < bitmap->width && c+x < width; c++) {
  639. /* get intensity value in the glyph bitmap (source) */
  640. src_val = GET_BITMAP_VAL(r, c);
  641. if (!src_val)
  642. continue;
  643. SET_PIXEL_YUV(frame, yuva_color, src_val, c+x, y+r, hsub, vsub);
  644. }
  645. }
  646. return 0;
  647. }
  648. #define SET_PIXEL_RGB(frame, rgba_color, val, x, y, pixel_step, r_off, g_off, b_off, a_off) { \
  649. p = frame->data[0] + (x) * pixel_step + ((y) * frame->linesize[0]); \
  650. alpha = rgba_color[3] * alpha_mul * (val) * 129 / 255; \
  651. *(p+r_off) = (alpha * rgba_color[0] + (255*255*129 - alpha) * *(p+r_off)) >> 23; \
  652. *(p+g_off) = (alpha * rgba_color[1] + (255*255*129 - alpha) * *(p+g_off)) >> 23; \
  653. *(p+b_off) = (alpha * rgba_color[2] + (255*255*129 - alpha) * *(p+b_off)) >> 23; \
  654. }
  655. static inline int draw_glyph_rgb(AVFrame *frame, FT_Bitmap *bitmap,
  656. unsigned int x, unsigned int y,
  657. unsigned int width, unsigned int height, int pixel_step,
  658. const uint8_t rgba_color[4], const uint8_t rgba_map[4],
  659. int alpha_mul)
  660. {
  661. int r, c, alpha;
  662. uint8_t *p;
  663. uint8_t src_val;
  664. for (r = 0; r < bitmap->rows && r+y < height; r++) {
  665. for (c = 0; c < bitmap->width && c+x < width; c++) {
  666. /* get intensity value in the glyph bitmap (source) */
  667. src_val = GET_BITMAP_VAL(r, c);
  668. if (!src_val)
  669. continue;
  670. SET_PIXEL_RGB(frame, rgba_color, src_val, c+x, y+r, pixel_step,
  671. rgba_map[0], rgba_map[1], rgba_map[2], rgba_map[3]);
  672. }
  673. }
  674. return 0;
  675. }
  676. static inline void drawbox(AVFrame *frame, unsigned int x, unsigned int y,
  677. unsigned int width, unsigned int height,
  678. uint8_t *line[4], int pixel_step[4], uint8_t color[4],
  679. int hsub, int vsub, int is_rgba_packed, uint8_t rgba_map[4],
  680. int alpha_mul)
  681. {
  682. int i, j, alpha;
  683. if (color[3] != 0xFF || alpha_mul != 0xFF) {
  684. if (is_rgba_packed) {
  685. uint8_t *p;
  686. for (j = 0; j < height; j++)
  687. for (i = 0; i < width; i++)
  688. SET_PIXEL_RGB(frame, color, 255, i+x, y+j, pixel_step[0],
  689. rgba_map[0], rgba_map[1], rgba_map[2], rgba_map[3]);
  690. } else {
  691. unsigned int luma_pos, chroma_pos1, chroma_pos2;
  692. for (j = 0; j < height; j++)
  693. for (i = 0; i < width; i++)
  694. SET_PIXEL_YUV(frame, color, 255, i+x, y+j, hsub, vsub);
  695. }
  696. } else {
  697. ff_draw_rectangle(frame->data, frame->linesize,
  698. line, pixel_step, hsub, vsub,
  699. x, y, width, height);
  700. }
  701. }
  702. static int draw_glyphs(DrawTextContext *s, AVFrame *frame,
  703. int width, int height,
  704. const uint8_t rgbcolor[4], const uint8_t yuvcolor[4],
  705. int x, int y)
  706. {
  707. char *text = s->expanded_text;
  708. uint32_t code = 0;
  709. int i;
  710. uint8_t *p;
  711. Glyph *glyph = NULL;
  712. for (i = 0, p = text; *p; i++) {
  713. Glyph dummy = { 0 };
  714. GET_UTF8(code, *p++, continue;);
  715. /* skip new line chars, just go to new line */
  716. if (code == '\n' || code == '\r' || code == '\t')
  717. continue;
  718. dummy.code = code;
  719. glyph = av_tree_find(s->glyphs, &dummy, (void *)glyph_cmp, NULL);
  720. if (glyph->bitmap.pixel_mode != FT_PIXEL_MODE_MONO &&
  721. glyph->bitmap.pixel_mode != FT_PIXEL_MODE_GRAY)
  722. return AVERROR(EINVAL);
  723. if (s->is_packed_rgb) {
  724. draw_glyph_rgb(frame, &glyph->bitmap,
  725. s->positions[i].x+x, s->positions[i].y+y, width, height,
  726. s->pixel_step[0], rgbcolor, s->rgba_map, s->alpha);
  727. } else {
  728. draw_glyph_yuv(frame, &glyph->bitmap,
  729. s->positions[i].x+x, s->positions[i].y+y, width, height,
  730. yuvcolor, s->hsub, s->vsub, s->alpha);
  731. }
  732. }
  733. return 0;
  734. }
  735. static int draw_text(AVFilterContext *ctx, AVFrame *frame,
  736. int width, int height)
  737. {
  738. DrawTextContext *s = ctx->priv;
  739. int ret;
  740. /* draw box */
  741. if (s->draw_box)
  742. drawbox(frame, s->x, s->y, s->w, s->h,
  743. s->box_line, s->pixel_step, s->boxcolor,
  744. s->hsub, s->vsub, s->is_packed_rgb,
  745. s->rgba_map, s->alpha);
  746. if (s->shadowx || s->shadowy) {
  747. if ((ret = draw_glyphs(s, frame, width, height,
  748. s->shadowcolor_rgba,
  749. s->shadowcolor,
  750. s->x + s->shadowx,
  751. s->y + s->shadowy)) < 0)
  752. return ret;
  753. }
  754. if ((ret = draw_glyphs(s, frame, width, height,
  755. s->fontcolor_rgba,
  756. s->fontcolor,
  757. s->x,
  758. s->y)) < 0)
  759. return ret;
  760. return 0;
  761. }
  762. static inline int normalize_double(int *n, double d)
  763. {
  764. int ret = 0;
  765. if (isnan(d)) {
  766. ret = AVERROR(EINVAL);
  767. } else if (d > INT_MAX || d < INT_MIN) {
  768. *n = d > INT_MAX ? INT_MAX : INT_MIN;
  769. ret = AVERROR(EINVAL);
  770. } else
  771. *n = round(d);
  772. return ret;
  773. }
  774. static void update_alpha(DrawTextContext *s)
  775. {
  776. double alpha = av_expr_eval(s->a_pexpr, s->var_values, &s->prng);
  777. if (isnan(alpha))
  778. return;
  779. if (alpha >= 1.0)
  780. s->alpha = 255;
  781. else if (alpha <= 0)
  782. s->alpha = 0;
  783. else
  784. s->alpha = 256 * alpha;
  785. }
  786. static int filter_frame(AVFilterLink *inlink, AVFrame *frame)
  787. {
  788. AVFilterContext *ctx = inlink->dst;
  789. DrawTextContext *s = ctx->priv;
  790. int ret = 0;
  791. if ((ret = dtext_prepare_text(ctx)) < 0) {
  792. av_log(ctx, AV_LOG_ERROR, "Can't draw text\n");
  793. av_frame_free(&frame);
  794. return ret;
  795. }
  796. s->var_values[VAR_T] = frame->pts == AV_NOPTS_VALUE ?
  797. NAN : frame->pts * av_q2d(inlink->time_base);
  798. s->var_values[VAR_X] =
  799. av_expr_eval(s->x_pexpr, s->var_values, &s->prng);
  800. s->var_values[VAR_Y] =
  801. av_expr_eval(s->y_pexpr, s->var_values, &s->prng);
  802. s->var_values[VAR_X] =
  803. av_expr_eval(s->x_pexpr, s->var_values, &s->prng);
  804. s->draw = av_expr_eval(s->d_pexpr, s->var_values, &s->prng);
  805. update_alpha(s);
  806. normalize_double(&s->x, s->var_values[VAR_X]);
  807. normalize_double(&s->y, s->var_values[VAR_Y]);
  808. if (s->fix_bounds) {
  809. if (s->x < 0) s->x = 0;
  810. if (s->y < 0) s->y = 0;
  811. if ((unsigned)s->x + (unsigned)s->w > inlink->w)
  812. s->x = inlink->w - s->w;
  813. if ((unsigned)s->y + (unsigned)s->h > inlink->h)
  814. s->y = inlink->h - s->h;
  815. }
  816. s->x &= ~((1 << s->hsub) - 1);
  817. s->y &= ~((1 << s->vsub) - 1);
  818. av_log(ctx, AV_LOG_TRACE, "n:%d t:%f x:%d y:%d x+w:%d y+h:%d\n",
  819. (int)s->var_values[VAR_N], s->var_values[VAR_T],
  820. s->x, s->y, s->x+s->w, s->y+s->h);
  821. if (s->draw)
  822. draw_text(inlink->dst, frame, frame->width, frame->height);
  823. s->var_values[VAR_N] += 1.0;
  824. return ff_filter_frame(inlink->dst->outputs[0], frame);
  825. }
  826. static const AVFilterPad avfilter_vf_drawtext_inputs[] = {
  827. {
  828. .name = "default",
  829. .type = AVMEDIA_TYPE_VIDEO,
  830. .get_video_buffer = ff_null_get_video_buffer,
  831. .filter_frame = filter_frame,
  832. .config_props = config_input,
  833. .needs_writable = 1,
  834. },
  835. { NULL }
  836. };
  837. static const AVFilterPad avfilter_vf_drawtext_outputs[] = {
  838. {
  839. .name = "default",
  840. .type = AVMEDIA_TYPE_VIDEO,
  841. },
  842. { NULL }
  843. };
  844. AVFilter ff_vf_drawtext = {
  845. .name = "drawtext",
  846. .description = NULL_IF_CONFIG_SMALL("Draw text on top of video frames using libfreetype library."),
  847. .priv_size = sizeof(DrawTextContext),
  848. .priv_class = &drawtext_class,
  849. .init = init,
  850. .uninit = uninit,
  851. .query_formats = query_formats,
  852. .inputs = avfilter_vf_drawtext_inputs,
  853. .outputs = avfilter_vf_drawtext_outputs,
  854. };