2 * 3GPP TS 26.245 Timed Text decoder
3 * Copyright (c) 2012 Philip Langdale <philipl@overt.org>
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 #include "libavutil/opt.h"
25 #include "libavutil/avstring.h"
26 #include "libavutil/common.h"
27 #include "libavutil/bprint.h"
28 #include "libavutil/intreadwrite.h"
29 #include "libavutil/mem.h"
31 #define STYLE_FLAG_BOLD (1<<0)
32 #define STYLE_FLAG_ITALIC (1<<1)
33 #define STYLE_FLAG_UNDERLINE (1<<2)
35 #define BOX_SIZE_INITIAL 40
37 #define STYL_BOX (1<<0)
38 #define HLIT_BOX (1<<1)
39 #define HCLR_BOX (1<<2)
40 #define TWRP_BOX (1<<3)
43 #define BOTTOM_CENTER 2
44 #define BOTTOM_RIGHT 3
46 #define MIDDLE_CENTER 5
47 #define MIDDLE_RIGHT 6
52 #define RGB_TO_BGR(c) (((c) & 0xff) << 16 | ((c) & 0xff00) | (((c) >> 16) & 0xff))
83 uint16_t style_fontID;
92 uint8_t hlit_color[4];
108 uint16_t style_entries, ftab_entries;
119 int (*decode)(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt);
122 static void mov_text_cleanup(MovTextContext *m)
124 if (m->box_flags & STYL_BOX) {
126 m->style_entries = 0;
130 static void mov_text_cleanup_ftab(MovTextContext *m)
132 for (unsigned i = 0; i < m->ftab_entries; i++)
133 av_freep(&m->ftab[i].font);
138 static int mov_text_tx3g(AVCodecContext *avctx, MovTextContext *m)
140 uint8_t *tx3g_ptr = avctx->extradata;
141 int i, j = -1, font_length, remaining = avctx->extradata_size - BOX_SIZE_INITIAL;
142 int8_t v_align, h_align;
143 unsigned ftab_entries;
153 h_align = *tx3g_ptr++;
154 v_align = *tx3g_ptr++;
157 m->d.alignment = TOP_LEFT;
159 m->d.alignment = MIDDLE_LEFT;
161 m->d.alignment = BOTTOM_LEFT;
165 m->d.alignment = TOP_CENTER;
167 m->d.alignment = MIDDLE_CENTER;
169 m->d.alignment = BOTTOM_CENTER;
173 m->d.alignment = TOP_RIGHT;
175 m->d.alignment = MIDDLE_RIGHT;
177 m->d.alignment = BOTTOM_RIGHT;
180 m->d.back_color = AV_RB24(tx3g_ptr);
182 m->d.back_alpha = AV_RB8(tx3g_ptr);
189 m->d.fontID = AV_RB16(tx3g_ptr);
192 s_default.style_flag = *tx3g_ptr++;
193 m->d.bold = !!(s_default.style_flag & STYLE_FLAG_BOLD);
194 m->d.italic = !!(s_default.style_flag & STYLE_FLAG_ITALIC);
195 m->d.underline = !!(s_default.style_flag & STYLE_FLAG_UNDERLINE);
197 m->d.fontsize = *tx3g_ptr++;
199 m->d.color = AV_RB24(tx3g_ptr);
201 m->d.alpha = AV_RB8(tx3g_ptr);
209 // In case of broken header, init default font
210 m->d.font = ASS_DEFAULT_FONT;
212 ftab_entries = AV_RB16(tx3g_ptr);
215 remaining -= 3 * ftab_entries;
217 return AVERROR_INVALIDDATA;
218 m->ftab = av_calloc(ftab_entries, sizeof(*m->ftab));
220 return AVERROR(ENOMEM);
221 m->ftab_entries = ftab_entries;
224 for (i = 0; i < m->ftab_entries; i++) {
225 m->ftab[i].fontID = AV_RB16(tx3g_ptr);
226 if (m->ftab[i].fontID == m->d.fontID)
229 font_length = *tx3g_ptr++;
231 remaining -= font_length;
233 mov_text_cleanup_ftab(m);
236 m->ftab[i].font = av_malloc(font_length + 1);
237 if (!m->ftab[i].font) {
238 mov_text_cleanup_ftab(m);
239 return AVERROR(ENOMEM);
241 memcpy(m->ftab[i].font, tx3g_ptr, font_length);
242 m->ftab[i].font[font_length] = '\0';
243 tx3g_ptr = tx3g_ptr + font_length;
246 m->d.font = m->ftab[j].font;
250 static int decode_twrp(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
252 m->box_flags |= TWRP_BOX;
253 m->w.wrap_flag = *tsmb++;
257 static int decode_hlit(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
259 m->box_flags |= HLIT_BOX;
260 m->h.hlit_start = AV_RB16(tsmb);
262 m->h.hlit_end = AV_RB16(tsmb);
267 static int decode_hclr(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
269 m->box_flags |= HCLR_BOX;
270 memcpy(m->c.hlit_color, tsmb, 4);
275 static int decode_styl(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
278 int style_entries = AV_RB16(tsmb);
281 // A single style record is of length 12 bytes.
282 if (m->tracksize + m->size_var + 2 + style_entries * 12 > avpkt->size)
285 tmp = av_realloc_array(m->s, style_entries, sizeof(*m->s));
287 return AVERROR(ENOMEM);
289 m->style_entries = style_entries;
291 m->box_flags |= STYL_BOX;
292 for(i = 0; i < m->style_entries; i++) {
293 StyleBox *style = &m->s[i];
294 style->style_start = AV_RB16(tsmb);
296 style->style_end = AV_RB16(tsmb);
298 if ( style->style_end < style->style_start
299 || (i && style->style_start < m->s[i - 1].style_end)) {
301 return AVERROR(ENOMEM);
305 style->style_fontID = AV_RB16(tsmb);
307 style->style_flag = AV_RB8(tsmb);
308 style->bold = !!(style->style_flag & STYLE_FLAG_BOLD);
309 style->italic = !!(style->style_flag & STYLE_FLAG_ITALIC);
310 style->underline = !!(style->style_flag & STYLE_FLAG_UNDERLINE);
312 style->fontsize = AV_RB8(tsmb);
314 style->color = AV_RB24(tsmb);
316 style->alpha = AV_RB8(tsmb);
322 static const Box box_types[] = {
323 { MKBETAG('s','t','y','l'), 2, decode_styl },
324 { MKBETAG('h','l','i','t'), 4, decode_hlit },
325 { MKBETAG('h','c','l','r'), 4, decode_hclr },
326 { MKBETAG('t','w','r','p'), 1, decode_twrp }
329 const static size_t box_count = FF_ARRAY_ELEMS(box_types);
331 // Return byte length of the UTF-8 sequence starting at text[0]. 0 on error.
332 static int get_utf8_length_at(const char *text, const char *text_end)
334 const char *start = text;
337 GET_UTF8(c, text < text_end ? (uint8_t)*text++ : (err = 1, 0), goto error;);
345 static int text_to_ass(AVBPrint *buf, const char *text, const char *text_end,
346 AVCodecContext *avctx)
348 MovTextContext *m = avctx->priv_data;
351 int style_active = 0;
353 int color = m->d.color;
355 if (text < text_end && m->box_flags & TWRP_BOX) {
356 if (m->w.wrap_flag == 1) {
357 av_bprintf(buf, "{\\q1}"); /* End of line wrap */
359 av_bprintf(buf, "{\\q2}"); /* No wrap */
363 while (text < text_end) {
366 if ((m->box_flags & STYL_BOX) && entry < m->style_entries) {
367 const StyleBox *style = &m->s[entry];
368 if (text_pos == style->style_start) {
370 if (style->bold ^ m->d.bold)
371 av_bprintf(buf, "{\\b%d}", style->bold);
372 if (style->italic ^ m->d.italic)
373 av_bprintf(buf, "{\\i%d}", style->italic);
374 if (style->underline ^ m->d.underline)
375 av_bprintf(buf, "{\\u%d}", style->underline);
376 if (style->fontsize != m->d.fontsize)
377 av_bprintf(buf, "{\\fs%d}", style->fontsize);
378 if (style->style_fontID != m->d.fontID)
379 for (i = 0; i < m->ftab_entries; i++) {
380 if (style->style_fontID == m->ftab[i].fontID)
381 av_bprintf(buf, "{\\fn%s}", m->ftab[i].font);
383 if (m->d.color != style->color) {
384 color = style->color;
385 av_bprintf(buf, "{\\1c&H%X&}", RGB_TO_BGR(color));
387 if (m->d.alpha != style->alpha)
388 av_bprintf(buf, "{\\1a&H%02X&}", 255 - style->alpha);
390 if (text_pos == style->style_end) {
392 av_bprintf(buf, "{\\r}");
399 if (m->box_flags & HLIT_BOX) {
400 if (text_pos == m->h.hlit_start) {
401 /* If hclr box is present, set the secondary color to the color
402 * specified. Otherwise, set primary color to white and secondary
403 * color to black. These colors will come from TextSampleModifier
404 * boxes in future and inverse video technique for highlight will
407 if (m->box_flags & HCLR_BOX) {
408 av_bprintf(buf, "{\\2c&H%02x%02x%02x&}", m->c.hlit_color[2],
409 m->c.hlit_color[1], m->c.hlit_color[0]);
411 av_bprintf(buf, "{\\1c&H000000&}{\\2c&HFFFFFF&}");
414 if (text_pos == m->h.hlit_end) {
415 if (m->box_flags & HCLR_BOX) {
416 av_bprintf(buf, "{\\2c&H%X&}", RGB_TO_BGR(m->d.color));
418 av_bprintf(buf, "{\\1c&H%X&}{\\2c&H%X&}",
419 RGB_TO_BGR(color), RGB_TO_BGR(m->d.color));
424 len = get_utf8_length_at(text, text_end);
426 av_log(avctx, AV_LOG_ERROR, "invalid UTF-8 byte in subtitle\n");
429 for (i = 0; i < len; i++) {
434 av_bprintf(buf, "\\N");
437 av_bprint_chars(buf, *text, 1);
448 static int mov_text_init(AVCodecContext *avctx) {
450 * TODO: Handle the default text style.
451 * NB: Most players ignore styles completely, with the result that
452 * it's very common to find files where the default style is broken
453 * and respecting it results in a worse experience than ignoring it.
456 MovTextContext *m = avctx->priv_data;
457 ret = mov_text_tx3g(avctx, m);
459 if (!m->frame_width || !m->frame_height) {
460 m->frame_width = ASS_DEFAULT_PLAYRESX;
461 m->frame_height = ASS_DEFAULT_PLAYRESY;
463 return ff_ass_subtitle_header_full(avctx,
464 m->frame_width, m->frame_height,
465 m->d.font, m->d.fontsize,
466 (255U - m->d.alpha) << 24 | RGB_TO_BGR(m->d.color),
467 (255U - m->d.alpha) << 24 | RGB_TO_BGR(m->d.color),
468 (255U - m->d.back_alpha) << 24 | RGB_TO_BGR(m->d.back_color),
469 (255U - m->d.back_alpha) << 24 | RGB_TO_BGR(m->d.back_color),
470 m->d.bold, m->d.italic, m->d.underline,
471 ASS_DEFAULT_BORDERSTYLE, m->d.alignment);
473 return ff_ass_subtitle_header_default(avctx);
476 static int mov_text_decode_frame(AVCodecContext *avctx,
477 void *data, int *got_sub_ptr, AVPacket *avpkt)
479 AVSubtitle *sub = data;
480 MovTextContext *m = avctx->priv_data;
483 char *ptr = avpkt->data;
485 int text_length, tsmb_type, ret_tsmb;
490 if (!ptr || avpkt->size < 2)
491 return AVERROR_INVALIDDATA;
494 * A packet of size two with value zero is an empty subtitle
495 * used to mark the end of the previous non-empty subtitle.
496 * We can just drop them here as we have duration information
497 * already. If the value is non-zero, then it's technically a
500 if (avpkt->size == 2)
501 return AV_RB16(ptr) == 0 ? 0 : AVERROR_INVALIDDATA;
504 * The first two bytes of the packet are the length of the text string
505 * In complex cases, there are style descriptors appended to the string
506 * so we can't just assume the packet size is the string size.
508 text_length = AV_RB16(ptr);
509 end = ptr + FFMIN(2 + text_length, avpkt->size);
515 m->tracksize = 2 + text_length;
516 m->style_entries = 0;
518 // Note that the spec recommends lines be no longer than 2048 characters.
519 av_bprint_init(&buf, 0, AV_BPRINT_SIZE_UNLIMITED);
520 if (text_length + 2 != avpkt->size) {
521 while (m->tracksize + 8 <= avpkt->size) {
522 // A box is a minimum of 8 bytes.
523 tsmb = ptr + m->tracksize - 2;
524 tsmb_size = AV_RB32(tsmb);
526 tsmb_type = AV_RB32(tsmb);
529 if (tsmb_size == 1) {
530 if (m->tracksize + 16 > avpkt->size)
532 tsmb_size = AV_RB64(tsmb);
537 //size_var is equal to 8 or 16 depending on the size of box
539 if (tsmb_size == 0) {
540 av_log(avctx, AV_LOG_ERROR, "tsmb_size is 0\n");
541 return AVERROR_INVALIDDATA;
544 if (tsmb_size > avpkt->size - m->tracksize)
547 for (i = 0; i < box_count; i++) {
548 if (tsmb_type == box_types[i].type) {
549 if (m->tracksize + m->size_var + box_types[i].base_size > avpkt->size)
551 ret_tsmb = box_types[i].decode(tsmb, m, avpkt);
556 m->tracksize = m->tracksize + tsmb_size;
558 text_to_ass(&buf, ptr, end, avctx);
561 text_to_ass(&buf, ptr, end, avctx);
563 ret = ff_ass_add_rect(sub, buf.str, m->readorder++, 0, NULL, NULL);
564 av_bprint_finalize(&buf, NULL);
567 *got_sub_ptr = sub->num_rects > 0;
571 static int mov_text_decode_close(AVCodecContext *avctx)
573 MovTextContext *m = avctx->priv_data;
574 mov_text_cleanup_ftab(m);
579 static void mov_text_flush(AVCodecContext *avctx)
581 MovTextContext *m = avctx->priv_data;
582 if (!(avctx->flags2 & AV_CODEC_FLAG2_RO_FLUSH_NOOP))
586 #define OFFSET(x) offsetof(MovTextContext, x)
587 #define FLAGS AV_OPT_FLAG_DECODING_PARAM | AV_OPT_FLAG_SUBTITLE_PARAM
588 static const AVOption options[] = {
589 { "width", "Frame width, usually video width", OFFSET(frame_width), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
590 { "height", "Frame height, usually video height", OFFSET(frame_height), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
594 static const AVClass mov_text_decoder_class = {
595 .class_name = "MOV text decoder",
596 .item_name = av_default_item_name,
598 .version = LIBAVUTIL_VERSION_INT,
601 AVCodec ff_movtext_decoder = {
603 .long_name = NULL_IF_CONFIG_SMALL("3GPP Timed Text subtitle"),
604 .type = AVMEDIA_TYPE_SUBTITLE,
605 .id = AV_CODEC_ID_MOV_TEXT,
606 .priv_data_size = sizeof(MovTextContext),
607 .priv_class = &mov_text_decoder_class,
608 .init = mov_text_init,
609 .decode = mov_text_decode_frame,
610 .close = mov_text_decode_close,
611 .flush = mov_text_flush,