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.

974 lines
33KB

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