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.

906 lines
35KB

  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/colorspace.h"
  32. #include "libavutil/file.h"
  33. #include "libavutil/eval.h"
  34. #include "libavutil/opt.h"
  35. #include "libavutil/random_seed.h"
  36. #include "libavutil/parseutils.h"
  37. #include "libavutil/pixdesc.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. #undef time
  44. #include <ft2build.h>
  45. #include <freetype/config/ftheader.h>
  46. #include FT_FREETYPE_H
  47. #include FT_GLYPH_H
  48. static const char *const var_names[] = {
  49. "main_w", "w", "W", ///< width of the input video
  50. "main_h", "h", "H", ///< height of the input video
  51. "tw", "text_w", ///< width of the rendered text
  52. "th", "text_h", ///< height of the rendered text
  53. "max_glyph_w", ///< max glyph width
  54. "max_glyph_h", ///< max glyph height
  55. "max_glyph_a", "ascent", ///< max glyph ascent
  56. "max_glyph_d", "descent", ///< min glyph descent
  57. "line_h", "lh", ///< line height, same as max_glyph_h
  58. "sar",
  59. "dar",
  60. "hsub",
  61. "vsub",
  62. "x",
  63. "y",
  64. "n", ///< number of frame
  65. "t", ///< timestamp expressed in seconds
  66. NULL
  67. };
  68. static const char *const fun2_names[] = {
  69. "rand"
  70. };
  71. static double drand(void *opaque, double min, double max)
  72. {
  73. return min + (max-min) / UINT_MAX * av_lfg_get(opaque);
  74. }
  75. typedef double (*eval_func2)(void *, double a, double b);
  76. static const eval_func2 fun2[] = {
  77. drand,
  78. NULL
  79. };
  80. enum var_name {
  81. VAR_MAIN_W, VAR_w, VAR_W,
  82. VAR_MAIN_H, VAR_h, VAR_H,
  83. VAR_TW, VAR_TEXT_W,
  84. VAR_TH, VAR_TEXT_H,
  85. VAR_MAX_GLYPH_W,
  86. VAR_MAX_GLYPH_H,
  87. VAR_MAX_GLYPH_A, VAR_ASCENT,
  88. VAR_MAX_GLYPH_D, VAR_DESCENT,
  89. VAR_LINE_H, VAR_LH,
  90. VAR_SAR,
  91. VAR_DAR,
  92. VAR_HSUB,
  93. VAR_VSUB,
  94. VAR_X,
  95. VAR_Y,
  96. VAR_N,
  97. VAR_T,
  98. VAR_VARS_NB
  99. };
  100. typedef struct {
  101. const AVClass *class;
  102. int reinit; ///< tells if the filter is being reinited
  103. uint8_t *fontfile; ///< font to be used
  104. uint8_t *text; ///< text to be drawn
  105. uint8_t *expanded_text; ///< used to contain the strftime()-expanded text
  106. size_t expanded_text_size; ///< size in bytes of the expanded_text buffer
  107. int ft_load_flags; ///< flags used for loading fonts, see FT_LOAD_*
  108. FT_Vector *positions; ///< positions for each element in the text
  109. size_t nb_positions; ///< number of elements of positions array
  110. char *textfile; ///< file with text to be drawn
  111. int x; ///< x position to start drawing text
  112. int y; ///< y position to start drawing text
  113. int max_glyph_w; ///< max glyph width
  114. int max_glyph_h; ///< max glyph height
  115. int shadowx, shadowy;
  116. unsigned int fontsize; ///< font size to use
  117. char *fontcolor_string; ///< font color as string
  118. char *boxcolor_string; ///< box color as string
  119. char *shadowcolor_string; ///< shadow color as string
  120. uint8_t fontcolor[4]; ///< foreground color
  121. uint8_t boxcolor[4]; ///< background color
  122. uint8_t shadowcolor[4]; ///< shadow color
  123. uint8_t fontcolor_rgba[4]; ///< foreground color in RGBA
  124. uint8_t boxcolor_rgba[4]; ///< background color in RGBA
  125. uint8_t shadowcolor_rgba[4]; ///< shadow color in RGBA
  126. short int draw_box; ///< draw box around text - true or false
  127. int use_kerning; ///< font kerning is used - true/false
  128. int tabsize; ///< tab size
  129. int fix_bounds; ///< do we let it go out of frame bounds - t/f
  130. FT_Library library; ///< freetype font library handle
  131. FT_Face face; ///< freetype font face handle
  132. struct AVTreeNode *glyphs; ///< rendered glyphs, stored using the UTF-32 char code
  133. int hsub, vsub; ///< chroma subsampling values
  134. int is_packed_rgb;
  135. int pixel_step[4]; ///< distance in bytes between the component of each pixel
  136. uint8_t rgba_map[4]; ///< map RGBA offsets to the positions in the packed RGBA format
  137. uint8_t *box_line[4]; ///< line used for filling the box background
  138. char *x_expr; ///< expression for x position
  139. char *y_expr; ///< expression for y position
  140. AVExpr *x_pexpr, *y_pexpr; ///< parsed expressions for x and y
  141. int64_t basetime; ///< base pts time in the real world for display
  142. double var_values[VAR_VARS_NB];
  143. char *d_expr;
  144. AVExpr *d_pexpr;
  145. int draw; ///< set to zero to prevent drawing
  146. AVLFG prng; ///< random
  147. char *tc_opt_string; ///< specified timecode option string
  148. AVRational tc_rate; ///< frame rate for timecode
  149. AVTimecode tc; ///< timecode context
  150. int tc24hmax; ///< 1 if timecode is wrapped to 24 hours, 0 otherwise
  151. int frame_id;
  152. } DrawTextContext;
  153. #define OFFSET(x) offsetof(DrawTextContext, x)
  154. static const AVOption drawtext_options[]= {
  155. {"fontfile", "set font file", OFFSET(fontfile), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX },
  156. {"text", "set text", OFFSET(text), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX },
  157. {"textfile", "set text file", OFFSET(textfile), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX },
  158. {"fontcolor", "set foreground color", OFFSET(fontcolor_string), AV_OPT_TYPE_STRING, {.str="black"}, CHAR_MIN, CHAR_MAX },
  159. {"boxcolor", "set box color", OFFSET(boxcolor_string), AV_OPT_TYPE_STRING, {.str="white"}, CHAR_MIN, CHAR_MAX },
  160. {"shadowcolor", "set shadow color", OFFSET(shadowcolor_string), AV_OPT_TYPE_STRING, {.str="black"}, CHAR_MIN, CHAR_MAX },
  161. {"box", "set box", OFFSET(draw_box), AV_OPT_TYPE_INT, {.dbl=0}, 0, 1 },
  162. {"fontsize", "set font size", OFFSET(fontsize), AV_OPT_TYPE_INT, {.dbl=16}, 1, INT_MAX },
  163. {"x", "set x expression", OFFSET(x_expr), AV_OPT_TYPE_STRING, {.str="0"}, CHAR_MIN, CHAR_MAX },
  164. {"y", "set y expression", OFFSET(y_expr), AV_OPT_TYPE_STRING, {.str="0"}, CHAR_MIN, CHAR_MAX },
  165. {"shadowx", "set x", OFFSET(shadowx), AV_OPT_TYPE_INT, {.dbl=0}, INT_MIN, INT_MAX },
  166. {"shadowy", "set y", OFFSET(shadowy), AV_OPT_TYPE_INT, {.dbl=0}, INT_MIN, INT_MAX },
  167. {"tabsize", "set tab size", OFFSET(tabsize), AV_OPT_TYPE_INT, {.dbl=4}, 0, INT_MAX },
  168. {"basetime", "set base time", OFFSET(basetime), AV_OPT_TYPE_INT64, {.dbl=AV_NOPTS_VALUE}, INT64_MIN, INT64_MAX },
  169. {"draw", "if false do not draw", OFFSET(d_expr), AV_OPT_TYPE_STRING, {.str="1"}, CHAR_MIN, CHAR_MAX },
  170. {"timecode", "set initial timecode", OFFSET(tc_opt_string), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX },
  171. {"tc24hmax", "set 24 hours max (timecode only)", OFFSET(tc24hmax), AV_OPT_TYPE_INT, {.dbl=0}, 0, 1 },
  172. {"r", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX },
  173. {"rate", "set rate (timecode only)", OFFSET(tc_rate), AV_OPT_TYPE_RATIONAL, {.dbl=0}, 0, INT_MAX },
  174. {"fix_bounds", "if true, check and fix text coords to avoid clipping",
  175. OFFSET(fix_bounds), AV_OPT_TYPE_INT, {.dbl=1}, 0, 1 },
  176. /* FT_LOAD_* flags */
  177. {"ft_load_flags", "set font loading flags for libfreetype", OFFSET(ft_load_flags), AV_OPT_TYPE_FLAGS, {.dbl=FT_LOAD_DEFAULT|FT_LOAD_RENDER}, 0, INT_MAX, 0, "ft_load_flags" },
  178. {"default", "set default", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_DEFAULT}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  179. {"no_scale", "set no_scale", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_NO_SCALE}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  180. {"no_hinting", "set no_hinting", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_NO_HINTING}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  181. {"render", "set render", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_RENDER}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  182. {"no_bitmap", "set no_bitmap", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_NO_BITMAP}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  183. {"vertical_layout", "set vertical_layout", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_VERTICAL_LAYOUT}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  184. {"force_autohint", "set force_autohint", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_FORCE_AUTOHINT}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  185. {"crop_bitmap", "set crop_bitmap", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_CROP_BITMAP}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  186. {"pedantic", "set pedantic", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_PEDANTIC}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  187. {"ignore_global_advance_width", "set ignore_global_advance_width", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_IGNORE_GLOBAL_ADVANCE_WIDTH}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  188. {"no_recurse", "set no_recurse", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_NO_RECURSE}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  189. {"ignore_transform", "set ignore_transform", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_IGNORE_TRANSFORM}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  190. {"monochrome", "set monochrome", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_MONOCHROME}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  191. {"linear_design", "set linear_design", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_LINEAR_DESIGN}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  192. {"no_autohint", "set no_autohint", 0, AV_OPT_TYPE_CONST, {.dbl=FT_LOAD_NO_AUTOHINT}, INT_MIN, INT_MAX, 0, "ft_load_flags" },
  193. {NULL},
  194. };
  195. static const char *drawtext_get_name(void *ctx)
  196. {
  197. return "drawtext";
  198. }
  199. static const AVClass drawtext_class = {
  200. "DrawTextContext",
  201. drawtext_get_name,
  202. drawtext_options
  203. };
  204. #undef __FTERRORS_H__
  205. #define FT_ERROR_START_LIST {
  206. #define FT_ERRORDEF(e, v, s) { (e), (s) },
  207. #define FT_ERROR_END_LIST { 0, NULL } };
  208. struct ft_error
  209. {
  210. int err;
  211. const char *err_msg;
  212. } static ft_errors[] =
  213. #include FT_ERRORS_H
  214. #define FT_ERRMSG(e) ft_errors[e].err_msg
  215. typedef struct {
  216. FT_Glyph *glyph;
  217. uint32_t code;
  218. FT_Bitmap bitmap; ///< array holding bitmaps of font
  219. FT_BBox bbox;
  220. int advance;
  221. int bitmap_left;
  222. int bitmap_top;
  223. } Glyph;
  224. static int glyph_cmp(void *key, const void *b)
  225. {
  226. const Glyph *a = key, *bb = b;
  227. int64_t diff = (int64_t)a->code - (int64_t)bb->code;
  228. return diff > 0 ? 1 : diff < 0 ? -1 : 0;
  229. }
  230. /**
  231. * Load glyphs corresponding to the UTF-32 codepoint code.
  232. */
  233. static int load_glyph(AVFilterContext *ctx, Glyph **glyph_ptr, uint32_t code)
  234. {
  235. DrawTextContext *dtext = ctx->priv;
  236. Glyph *glyph;
  237. struct AVTreeNode *node = NULL;
  238. int ret;
  239. /* load glyph into dtext->face->glyph */
  240. if (FT_Load_Char(dtext->face, code, dtext->ft_load_flags))
  241. return AVERROR(EINVAL);
  242. /* save glyph */
  243. if (!(glyph = av_mallocz(sizeof(*glyph))) ||
  244. !(glyph->glyph = av_mallocz(sizeof(*glyph->glyph)))) {
  245. ret = AVERROR(ENOMEM);
  246. goto error;
  247. }
  248. glyph->code = code;
  249. if (FT_Get_Glyph(dtext->face->glyph, glyph->glyph)) {
  250. ret = AVERROR(EINVAL);
  251. goto error;
  252. }
  253. glyph->bitmap = dtext->face->glyph->bitmap;
  254. glyph->bitmap_left = dtext->face->glyph->bitmap_left;
  255. glyph->bitmap_top = dtext->face->glyph->bitmap_top;
  256. glyph->advance = dtext->face->glyph->advance.x >> 6;
  257. /* measure text height to calculate text_height (or the maximum text height) */
  258. FT_Glyph_Get_CBox(*glyph->glyph, ft_glyph_bbox_pixels, &glyph->bbox);
  259. /* cache the newly created glyph */
  260. if (!(node = av_mallocz(av_tree_node_size))) {
  261. ret = AVERROR(ENOMEM);
  262. goto error;
  263. }
  264. av_tree_insert(&dtext->glyphs, glyph, glyph_cmp, &node);
  265. if (glyph_ptr)
  266. *glyph_ptr = glyph;
  267. return 0;
  268. error:
  269. if (glyph)
  270. av_freep(&glyph->glyph);
  271. av_freep(&glyph);
  272. av_freep(&node);
  273. return ret;
  274. }
  275. static av_cold int init(AVFilterContext *ctx, const char *args, void *opaque)
  276. {
  277. int err;
  278. DrawTextContext *dtext = ctx->priv;
  279. Glyph *glyph;
  280. dtext->class = &drawtext_class;
  281. av_opt_set_defaults(dtext);
  282. if ((err = (av_set_options_string(dtext, args, "=", ":"))) < 0) {
  283. av_log(ctx, AV_LOG_ERROR, "Error parsing options string: '%s'\n", args);
  284. return err;
  285. }
  286. if (!dtext->fontfile) {
  287. av_log(ctx, AV_LOG_ERROR, "No font filename provided\n");
  288. return AVERROR(EINVAL);
  289. }
  290. if (dtext->textfile) {
  291. uint8_t *textbuf;
  292. size_t textbuf_size;
  293. if (dtext->text) {
  294. av_log(ctx, AV_LOG_ERROR,
  295. "Both text and text file provided. Please provide only one\n");
  296. return AVERROR(EINVAL);
  297. }
  298. if ((err = av_file_map(dtext->textfile, &textbuf, &textbuf_size, 0, ctx)) < 0) {
  299. av_log(ctx, AV_LOG_ERROR,
  300. "The text file '%s' could not be read or is empty\n",
  301. dtext->textfile);
  302. return err;
  303. }
  304. if (!(dtext->text = av_malloc(textbuf_size+1)))
  305. return AVERROR(ENOMEM);
  306. memcpy(dtext->text, textbuf, textbuf_size);
  307. dtext->text[textbuf_size] = 0;
  308. av_file_unmap(textbuf, textbuf_size);
  309. }
  310. if (dtext->tc_opt_string) {
  311. int ret = av_timecode_init_from_string(&dtext->tc, dtext->tc_rate,
  312. dtext->tc_opt_string, ctx);
  313. if (ret < 0)
  314. return ret;
  315. if (dtext->tc24hmax)
  316. dtext->tc.flags |= AV_TIMECODE_FLAG_24HOURSMAX;
  317. if (!dtext->text)
  318. dtext->text = av_strdup("");
  319. }
  320. if (!dtext->text) {
  321. av_log(ctx, AV_LOG_ERROR,
  322. "Either text, a valid file or a timecode must be provided\n");
  323. return AVERROR(EINVAL);
  324. }
  325. if ((err = av_parse_color(dtext->fontcolor_rgba, dtext->fontcolor_string, -1, ctx))) {
  326. av_log(ctx, AV_LOG_ERROR,
  327. "Invalid font color '%s'\n", dtext->fontcolor_string);
  328. return err;
  329. }
  330. if ((err = av_parse_color(dtext->boxcolor_rgba, dtext->boxcolor_string, -1, ctx))) {
  331. av_log(ctx, AV_LOG_ERROR,
  332. "Invalid box color '%s'\n", dtext->boxcolor_string);
  333. return err;
  334. }
  335. if ((err = av_parse_color(dtext->shadowcolor_rgba, dtext->shadowcolor_string, -1, ctx))) {
  336. av_log(ctx, AV_LOG_ERROR,
  337. "Invalid shadow color '%s'\n", dtext->shadowcolor_string);
  338. return err;
  339. }
  340. if ((err = FT_Init_FreeType(&(dtext->library)))) {
  341. av_log(ctx, AV_LOG_ERROR,
  342. "Could not load FreeType: %s\n", FT_ERRMSG(err));
  343. return AVERROR(EINVAL);
  344. }
  345. /* load the face, and set up the encoding, which is by default UTF-8 */
  346. if ((err = FT_New_Face(dtext->library, dtext->fontfile, 0, &dtext->face))) {
  347. av_log(ctx, AV_LOG_ERROR, "Could not load fontface from file '%s': %s\n",
  348. dtext->fontfile, FT_ERRMSG(err));
  349. return AVERROR(EINVAL);
  350. }
  351. if ((err = FT_Set_Pixel_Sizes(dtext->face, 0, dtext->fontsize))) {
  352. av_log(ctx, AV_LOG_ERROR, "Could not set font size to %d pixels: %s\n",
  353. dtext->fontsize, FT_ERRMSG(err));
  354. return AVERROR(EINVAL);
  355. }
  356. dtext->use_kerning = FT_HAS_KERNING(dtext->face);
  357. /* load the fallback glyph with code 0 */
  358. load_glyph(ctx, NULL, 0);
  359. /* set the tabsize in pixels */
  360. if ((err = load_glyph(ctx, &glyph, ' ')) < 0) {
  361. av_log(ctx, AV_LOG_ERROR, "Could not set tabsize.\n");
  362. return err;
  363. }
  364. dtext->tabsize *= glyph->advance;
  365. return 0;
  366. }
  367. static int query_formats(AVFilterContext *ctx)
  368. {
  369. static const enum PixelFormat pix_fmts[] = {
  370. PIX_FMT_ARGB, PIX_FMT_RGBA,
  371. PIX_FMT_ABGR, PIX_FMT_BGRA,
  372. PIX_FMT_RGB24, PIX_FMT_BGR24,
  373. PIX_FMT_YUV420P, PIX_FMT_YUV444P,
  374. PIX_FMT_YUV422P, PIX_FMT_YUV411P,
  375. PIX_FMT_YUV410P, PIX_FMT_YUV440P,
  376. PIX_FMT_NONE
  377. };
  378. avfilter_set_common_pixel_formats(ctx, avfilter_make_format_list(pix_fmts));
  379. return 0;
  380. }
  381. static int glyph_enu_free(void *opaque, void *elem)
  382. {
  383. av_free(elem);
  384. return 0;
  385. }
  386. static av_cold void uninit(AVFilterContext *ctx)
  387. {
  388. DrawTextContext *dtext = ctx->priv;
  389. int i;
  390. av_expr_free(dtext->x_pexpr); dtext->x_pexpr = NULL;
  391. av_expr_free(dtext->y_pexpr); dtext->y_pexpr = NULL;
  392. av_freep(&dtext->boxcolor_string);
  393. av_freep(&dtext->expanded_text);
  394. av_freep(&dtext->fontcolor_string);
  395. av_freep(&dtext->fontfile);
  396. av_freep(&dtext->shadowcolor_string);
  397. av_freep(&dtext->text);
  398. av_freep(&dtext->x_expr);
  399. av_freep(&dtext->y_expr);
  400. av_freep(&dtext->positions);
  401. dtext->nb_positions = 0;
  402. av_tree_enumerate(dtext->glyphs, NULL, NULL, glyph_enu_free);
  403. av_tree_destroy(dtext->glyphs);
  404. dtext->glyphs = NULL;
  405. FT_Done_Face(dtext->face);
  406. FT_Done_FreeType(dtext->library);
  407. for (i = 0; i < 4; i++) {
  408. av_freep(&dtext->box_line[i]);
  409. dtext->pixel_step[i] = 0;
  410. }
  411. }
  412. static inline int is_newline(uint32_t c)
  413. {
  414. return c == '\n' || c == '\r' || c == '\f' || c == '\v';
  415. }
  416. static int config_input(AVFilterLink *inlink)
  417. {
  418. AVFilterContext *ctx = inlink->dst;
  419. DrawTextContext *dtext = ctx->priv;
  420. const AVPixFmtDescriptor *pix_desc = &av_pix_fmt_descriptors[inlink->format];
  421. int ret;
  422. dtext->hsub = pix_desc->log2_chroma_w;
  423. dtext->vsub = pix_desc->log2_chroma_h;
  424. if ((ret =
  425. ff_fill_line_with_color(dtext->box_line, dtext->pixel_step,
  426. inlink->w, dtext->boxcolor,
  427. inlink->format, dtext->boxcolor_rgba,
  428. &dtext->is_packed_rgb, dtext->rgba_map)) < 0)
  429. return ret;
  430. if (!dtext->is_packed_rgb) {
  431. uint8_t *rgba = dtext->fontcolor_rgba;
  432. dtext->fontcolor[0] = RGB_TO_Y_CCIR(rgba[0], rgba[1], rgba[2]);
  433. dtext->fontcolor[1] = RGB_TO_U_CCIR(rgba[0], rgba[1], rgba[2], 0);
  434. dtext->fontcolor[2] = RGB_TO_V_CCIR(rgba[0], rgba[1], rgba[2], 0);
  435. dtext->fontcolor[3] = rgba[3];
  436. rgba = dtext->shadowcolor_rgba;
  437. dtext->shadowcolor[0] = RGB_TO_Y_CCIR(rgba[0], rgba[1], rgba[2]);
  438. dtext->shadowcolor[1] = RGB_TO_U_CCIR(rgba[0], rgba[1], rgba[2], 0);
  439. dtext->shadowcolor[2] = RGB_TO_V_CCIR(rgba[0], rgba[1], rgba[2], 0);
  440. dtext->shadowcolor[3] = rgba[3];
  441. }
  442. dtext->var_values[VAR_w] = dtext->var_values[VAR_W] = dtext->var_values[VAR_MAIN_W] = inlink->w;
  443. dtext->var_values[VAR_h] = dtext->var_values[VAR_H] = dtext->var_values[VAR_MAIN_H] = inlink->h;
  444. dtext->var_values[VAR_SAR] = inlink->sample_aspect_ratio.num ? av_q2d(inlink->sample_aspect_ratio) : 1;
  445. dtext->var_values[VAR_DAR] = (double)inlink->w / inlink->h * dtext->var_values[VAR_SAR];
  446. dtext->var_values[VAR_HSUB] = 1<<pix_desc->log2_chroma_w;
  447. dtext->var_values[VAR_VSUB] = 1<<pix_desc->log2_chroma_h;
  448. dtext->var_values[VAR_X] = NAN;
  449. dtext->var_values[VAR_Y] = NAN;
  450. if (!dtext->reinit)
  451. dtext->var_values[VAR_N] = 0;
  452. dtext->var_values[VAR_T] = NAN;
  453. av_lfg_init(&dtext->prng, av_get_random_seed());
  454. if ((ret = av_expr_parse(&dtext->x_pexpr, dtext->x_expr, var_names,
  455. NULL, NULL, fun2_names, fun2, 0, ctx)) < 0 ||
  456. (ret = av_expr_parse(&dtext->y_pexpr, dtext->y_expr, var_names,
  457. NULL, NULL, fun2_names, fun2, 0, ctx)) < 0 ||
  458. (ret = av_expr_parse(&dtext->d_pexpr, dtext->d_expr, var_names,
  459. NULL, NULL, fun2_names, fun2, 0, ctx)) < 0)
  460. return AVERROR(EINVAL);
  461. return 0;
  462. }
  463. static int command(AVFilterContext *ctx, const char *cmd, const char *arg, char *res, int res_len, int flags)
  464. {
  465. DrawTextContext *dtext = ctx->priv;
  466. if (!strcmp(cmd, "reinit")) {
  467. int ret;
  468. uninit(ctx);
  469. dtext->reinit = 1;
  470. if ((ret = init(ctx, arg, NULL)) < 0)
  471. return ret;
  472. return config_input(ctx->inputs[0]);
  473. }
  474. return AVERROR(ENOSYS);
  475. }
  476. #define GET_BITMAP_VAL(r, c) \
  477. bitmap->pixel_mode == FT_PIXEL_MODE_MONO ? \
  478. (bitmap->buffer[(r) * bitmap->pitch + ((c)>>3)] & (0x80 >> ((c)&7))) * 255 : \
  479. bitmap->buffer[(r) * bitmap->pitch + (c)]
  480. #define SET_PIXEL_YUV(picref, yuva_color, val, x, y, hsub, vsub) { \
  481. luma_pos = ((x) ) + ((y) ) * picref->linesize[0]; \
  482. alpha = yuva_color[3] * (val) * 129; \
  483. picref->data[0][luma_pos] = (alpha * yuva_color[0] + (255*255*129 - alpha) * picref->data[0][luma_pos] ) >> 23; \
  484. if (((x) & ((1<<(hsub)) - 1)) == 0 && ((y) & ((1<<(vsub)) - 1)) == 0) {\
  485. chroma_pos1 = ((x) >> (hsub)) + ((y) >> (vsub)) * picref->linesize[1]; \
  486. chroma_pos2 = ((x) >> (hsub)) + ((y) >> (vsub)) * picref->linesize[2]; \
  487. picref->data[1][chroma_pos1] = (alpha * yuva_color[1] + (255*255*129 - alpha) * picref->data[1][chroma_pos1]) >> 23; \
  488. picref->data[2][chroma_pos2] = (alpha * yuva_color[2] + (255*255*129 - alpha) * picref->data[2][chroma_pos2]) >> 23; \
  489. }\
  490. }
  491. static inline int draw_glyph_yuv(AVFilterBufferRef *picref, FT_Bitmap *bitmap,
  492. int x, int y, int width, int height,
  493. const uint8_t yuva_color[4], int hsub, int vsub)
  494. {
  495. int r, c, alpha;
  496. unsigned int luma_pos, chroma_pos1, chroma_pos2;
  497. uint8_t src_val;
  498. for (r = 0; r < bitmap->rows && r+y < height; r++) {
  499. for (c = 0; c < bitmap->width && c+x < width; c++) {
  500. if (c+x < 0 || r+y < 0)
  501. continue;
  502. /* get intensity value in the glyph bitmap (source) */
  503. src_val = GET_BITMAP_VAL(r, c);
  504. if (!src_val)
  505. continue;
  506. SET_PIXEL_YUV(picref, yuva_color, src_val, c+x, y+r, hsub, vsub);
  507. }
  508. }
  509. return 0;
  510. }
  511. #define SET_PIXEL_RGB(picref, rgba_color, val, x, y, pixel_step, r_off, g_off, b_off, a_off) { \
  512. p = picref->data[0] + (x) * pixel_step + ((y) * picref->linesize[0]); \
  513. alpha = rgba_color[3] * (val) * 129; \
  514. *(p+r_off) = (alpha * rgba_color[0] + (255*255*129 - alpha) * *(p+r_off)) >> 23; \
  515. *(p+g_off) = (alpha * rgba_color[1] + (255*255*129 - alpha) * *(p+g_off)) >> 23; \
  516. *(p+b_off) = (alpha * rgba_color[2] + (255*255*129 - alpha) * *(p+b_off)) >> 23; \
  517. }
  518. static inline int draw_glyph_rgb(AVFilterBufferRef *picref, FT_Bitmap *bitmap,
  519. int x, int y, int width, int height, int pixel_step,
  520. const uint8_t rgba_color[4], const uint8_t rgba_map[4])
  521. {
  522. int r, c, alpha;
  523. uint8_t *p;
  524. uint8_t src_val;
  525. for (r = 0; r < bitmap->rows && r+y < height; r++) {
  526. for (c = 0; c < bitmap->width && c+x < width; c++) {
  527. if (c+x < 0 || r+y < 0)
  528. continue;
  529. /* get intensity value in the glyph bitmap (source) */
  530. src_val = GET_BITMAP_VAL(r, c);
  531. if (!src_val)
  532. continue;
  533. SET_PIXEL_RGB(picref, rgba_color, src_val, c+x, y+r, pixel_step,
  534. rgba_map[0], rgba_map[1], rgba_map[2], rgba_map[3]);
  535. }
  536. }
  537. return 0;
  538. }
  539. static inline void drawbox(AVFilterBufferRef *picref, int x, int y,
  540. int width, int height,
  541. uint8_t *line[4], int pixel_step[4], uint8_t color[4],
  542. int hsub, int vsub, int is_rgba_packed, uint8_t rgba_map[4])
  543. {
  544. int i, j, alpha;
  545. if (color[3] != 0xFF) {
  546. if (is_rgba_packed) {
  547. uint8_t *p;
  548. for (j = 0; j < height; j++)
  549. for (i = 0; i < width; i++)
  550. SET_PIXEL_RGB(picref, color, 255, i+x, y+j, pixel_step[0],
  551. rgba_map[0], rgba_map[1], rgba_map[2], rgba_map[3]);
  552. } else {
  553. unsigned int luma_pos, chroma_pos1, chroma_pos2;
  554. for (j = 0; j < height; j++)
  555. for (i = 0; i < width; i++)
  556. SET_PIXEL_YUV(picref, color, 255, i+x, y+j, hsub, vsub);
  557. }
  558. } else {
  559. ff_draw_rectangle(picref->data, picref->linesize,
  560. line, pixel_step, hsub, vsub,
  561. x, y, width, height);
  562. }
  563. }
  564. static int draw_glyphs(DrawTextContext *dtext, AVFilterBufferRef *picref,
  565. int width, int height, const uint8_t rgbcolor[4], const uint8_t yuvcolor[4], int x, int y)
  566. {
  567. char *text = dtext->expanded_text;
  568. uint32_t code = 0;
  569. int i, x1, y1;
  570. uint8_t *p;
  571. Glyph *glyph = NULL;
  572. for (i = 0, p = text; *p; i++) {
  573. Glyph dummy = { 0 };
  574. GET_UTF8(code, *p++, continue;);
  575. /* skip new line chars, just go to new line */
  576. if (code == '\n' || code == '\r' || code == '\t')
  577. continue;
  578. dummy.code = code;
  579. glyph = av_tree_find(dtext->glyphs, &dummy, (void *)glyph_cmp, NULL);
  580. if (glyph->bitmap.pixel_mode != FT_PIXEL_MODE_MONO &&
  581. glyph->bitmap.pixel_mode != FT_PIXEL_MODE_GRAY)
  582. return AVERROR(EINVAL);
  583. x1 = dtext->positions[i].x+dtext->x+x;
  584. y1 = dtext->positions[i].y+dtext->y+y;
  585. if (dtext->is_packed_rgb) {
  586. draw_glyph_rgb(picref, &glyph->bitmap,
  587. x1, y1, width, height,
  588. dtext->pixel_step[0], rgbcolor, dtext->rgba_map);
  589. } else {
  590. draw_glyph_yuv(picref, &glyph->bitmap,
  591. x1, y1, width, height,
  592. yuvcolor, dtext->hsub, dtext->vsub);
  593. }
  594. }
  595. return 0;
  596. }
  597. static int draw_text(AVFilterContext *ctx, AVFilterBufferRef *picref,
  598. int width, int height)
  599. {
  600. DrawTextContext *dtext = ctx->priv;
  601. uint32_t code = 0, prev_code = 0;
  602. int x = 0, y = 0, i = 0, ret;
  603. int max_text_line_w = 0, len;
  604. int box_w, box_h;
  605. char *text = dtext->text;
  606. uint8_t *p;
  607. int y_min = 32000, y_max = -32000;
  608. int x_min = 32000, x_max = -32000;
  609. FT_Vector delta;
  610. Glyph *glyph = NULL, *prev_glyph = NULL;
  611. Glyph dummy = { 0 };
  612. time_t now = time(0);
  613. struct tm ltime;
  614. uint8_t *buf = dtext->expanded_text;
  615. int buf_size = dtext->expanded_text_size;
  616. if(dtext->basetime != AV_NOPTS_VALUE)
  617. now= picref->pts*av_q2d(ctx->inputs[0]->time_base) + dtext->basetime/1000000;
  618. if (!buf) {
  619. buf_size = 2*strlen(dtext->text)+1;
  620. buf = av_malloc(buf_size);
  621. }
  622. #if HAVE_LOCALTIME_R
  623. localtime_r(&now, &ltime);
  624. #else
  625. if(strchr(dtext->text, '%'))
  626. ltime= *localtime(&now);
  627. #endif
  628. do {
  629. *buf = 1;
  630. if (strftime(buf, buf_size, dtext->text, &ltime) != 0 || *buf == 0)
  631. break;
  632. buf_size *= 2;
  633. } while ((buf = av_realloc(buf, buf_size)));
  634. if (dtext->tc_opt_string) {
  635. char tcbuf[AV_TIMECODE_STR_SIZE];
  636. av_timecode_make_string(&dtext->tc, tcbuf, dtext->frame_id++);
  637. buf = av_asprintf("%s%s", dtext->text, tcbuf);
  638. }
  639. if (!buf)
  640. return AVERROR(ENOMEM);
  641. text = dtext->expanded_text = buf;
  642. dtext->expanded_text_size = buf_size;
  643. if ((len = strlen(text)) > dtext->nb_positions) {
  644. if (!(dtext->positions =
  645. av_realloc(dtext->positions, len*sizeof(*dtext->positions))))
  646. return AVERROR(ENOMEM);
  647. dtext->nb_positions = len;
  648. }
  649. x = 0;
  650. y = 0;
  651. /* load and cache glyphs */
  652. for (i = 0, p = text; *p; i++) {
  653. GET_UTF8(code, *p++, continue;);
  654. /* get glyph */
  655. dummy.code = code;
  656. glyph = av_tree_find(dtext->glyphs, &dummy, glyph_cmp, NULL);
  657. if (!glyph) {
  658. load_glyph(ctx, &glyph, code);
  659. }
  660. y_min = FFMIN(glyph->bbox.yMin, y_min);
  661. y_max = FFMAX(glyph->bbox.yMax, y_max);
  662. x_min = FFMIN(glyph->bbox.xMin, x_min);
  663. x_max = FFMAX(glyph->bbox.xMax, x_max);
  664. }
  665. dtext->max_glyph_h = y_max - y_min;
  666. dtext->max_glyph_w = x_max - x_min;
  667. /* compute and save position for each glyph */
  668. glyph = NULL;
  669. for (i = 0, p = text; *p; i++) {
  670. GET_UTF8(code, *p++, continue;);
  671. /* skip the \n in the sequence \r\n */
  672. if (prev_code == '\r' && code == '\n')
  673. continue;
  674. prev_code = code;
  675. if (is_newline(code)) {
  676. max_text_line_w = FFMAX(max_text_line_w, x);
  677. y += dtext->max_glyph_h;
  678. x = 0;
  679. continue;
  680. }
  681. /* get glyph */
  682. prev_glyph = glyph;
  683. dummy.code = code;
  684. glyph = av_tree_find(dtext->glyphs, &dummy, glyph_cmp, NULL);
  685. /* kerning */
  686. if (dtext->use_kerning && prev_glyph && glyph->code) {
  687. FT_Get_Kerning(dtext->face, prev_glyph->code, glyph->code,
  688. ft_kerning_default, &delta);
  689. x += delta.x >> 6;
  690. }
  691. /* save position */
  692. dtext->positions[i].x = x + glyph->bitmap_left;
  693. dtext->positions[i].y = y - glyph->bitmap_top + y_max;
  694. if (code == '\t') x = (x / dtext->tabsize + 1)*dtext->tabsize;
  695. else x += glyph->advance;
  696. }
  697. max_text_line_w = FFMAX(x, max_text_line_w);
  698. dtext->var_values[VAR_TW] = dtext->var_values[VAR_TEXT_W] = max_text_line_w;
  699. dtext->var_values[VAR_TH] = dtext->var_values[VAR_TEXT_H] = y + dtext->max_glyph_h;
  700. dtext->var_values[VAR_MAX_GLYPH_W] = dtext->max_glyph_w;
  701. dtext->var_values[VAR_MAX_GLYPH_H] = dtext->max_glyph_h;
  702. dtext->var_values[VAR_MAX_GLYPH_A] = dtext->var_values[VAR_ASCENT ] = y_max;
  703. dtext->var_values[VAR_MAX_GLYPH_D] = dtext->var_values[VAR_DESCENT] = y_min;
  704. dtext->var_values[VAR_LINE_H] = dtext->var_values[VAR_LH] = dtext->max_glyph_h;
  705. dtext->x = dtext->var_values[VAR_X] = av_expr_eval(dtext->x_pexpr, dtext->var_values, &dtext->prng);
  706. dtext->y = dtext->var_values[VAR_Y] = av_expr_eval(dtext->y_pexpr, dtext->var_values, &dtext->prng);
  707. dtext->x = dtext->var_values[VAR_X] = av_expr_eval(dtext->x_pexpr, dtext->var_values, &dtext->prng);
  708. dtext->draw = av_expr_eval(dtext->d_pexpr, dtext->var_values, &dtext->prng);
  709. if(!dtext->draw)
  710. return 0;
  711. dtext->x &= ~((1 << dtext->hsub) - 1);
  712. dtext->y &= ~((1 << dtext->vsub) - 1);
  713. box_w = FFMIN(width - 1 , max_text_line_w);
  714. box_h = FFMIN(height - 1, y + dtext->max_glyph_h);
  715. /* draw box */
  716. if (dtext->draw_box)
  717. drawbox(picref, dtext->x, dtext->y, box_w, box_h,
  718. dtext->box_line, dtext->pixel_step, dtext->is_packed_rgb ? dtext->boxcolor_rgba : dtext->boxcolor,
  719. dtext->hsub, dtext->vsub, dtext->is_packed_rgb, dtext->rgba_map);
  720. if (dtext->shadowx || dtext->shadowy) {
  721. if ((ret = draw_glyphs(dtext, picref, width, height, dtext->shadowcolor_rgba,
  722. dtext->shadowcolor, dtext->shadowx, dtext->shadowy)) < 0)
  723. return ret;
  724. }
  725. if ((ret = draw_glyphs(dtext, picref, width, height, dtext->fontcolor_rgba,
  726. dtext->fontcolor, 0, 0)) < 0)
  727. return ret;
  728. return 0;
  729. }
  730. static void null_draw_slice(AVFilterLink *link, int y, int h, int slice_dir) { }
  731. static void end_frame(AVFilterLink *inlink)
  732. {
  733. AVFilterContext *ctx = inlink->dst;
  734. AVFilterLink *outlink = ctx->outputs[0];
  735. DrawTextContext *dtext = ctx->priv;
  736. AVFilterBufferRef *picref = inlink->cur_buf;
  737. dtext->var_values[VAR_T] = picref->pts == AV_NOPTS_VALUE ?
  738. NAN : picref->pts * av_q2d(inlink->time_base);
  739. draw_text(ctx, picref, picref->video->w, picref->video->h);
  740. av_log(ctx, AV_LOG_DEBUG, "n:%d t:%f text_w:%d text_h:%d x:%d y:%d\n",
  741. (int)dtext->var_values[VAR_N], dtext->var_values[VAR_T],
  742. (int)dtext->var_values[VAR_TEXT_W], (int)dtext->var_values[VAR_TEXT_H],
  743. dtext->x, dtext->y);
  744. dtext->var_values[VAR_N] += 1.0;
  745. avfilter_draw_slice(outlink, 0, picref->video->h, 1);
  746. avfilter_end_frame(outlink);
  747. }
  748. AVFilter avfilter_vf_drawtext = {
  749. .name = "drawtext",
  750. .description = NULL_IF_CONFIG_SMALL("Draw text on top of video frames using libfreetype library."),
  751. .priv_size = sizeof(DrawTextContext),
  752. .init = init,
  753. .uninit = uninit,
  754. .query_formats = query_formats,
  755. .inputs = (const AVFilterPad[]) {{ .name = "default",
  756. .type = AVMEDIA_TYPE_VIDEO,
  757. .get_video_buffer = avfilter_null_get_video_buffer,
  758. .start_frame = avfilter_null_start_frame,
  759. .draw_slice = null_draw_slice,
  760. .end_frame = end_frame,
  761. .config_props = config_input,
  762. .min_perms = AV_PERM_WRITE |
  763. AV_PERM_READ,
  764. .rej_perms = AV_PERM_PRESERVE },
  765. { .name = NULL}},
  766. .outputs = (const AVFilterPad[]) {{ .name = "default",
  767. .type = AVMEDIA_TYPE_VIDEO, },
  768. { .name = NULL}},
  769. .process_command = command,
  770. };