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/avstring.h"
25 #include "libavutil/common.h"
26 #include "libavutil/bprint.h"
27 #include "libavutil/intreadwrite.h"
28 #include "libavutil/mem.h"
30 #define STYLE_FLAG_BOLD (1<<0)
31 #define STYLE_FLAG_ITALIC (1<<1)
32 #define STYLE_FLAG_UNDERLINE (1<<2)
34 #define BOX_SIZE_INITIAL 40
36 #define STYL_BOX (1<<0)
37 #define HLIT_BOX (1<<1)
38 #define HCLR_BOX (1<<2)
39 #define TWRP_BOX (1<<3)
42 #define BOTTOM_CENTER 2
43 #define BOTTOM_RIGHT 3
45 #define MIDDLE_CENTER 5
46 #define MIDDLE_RIGHT 6
51 #define RGB_TO_BGR(c) (((c) & 0xff) << 16 | ((c) & 0xff00) | (((c) >> 16) & 0xff))
74 uint16_t style_fontID;
83 uint8_t hlit_color[4];
96 FontRecord *ftab_temp;
100 uint16_t style_entries, ftab_entries;
103 int count_s, count_f;
110 int (*decode)(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt);
113 static void mov_text_cleanup(MovTextContext *m)
116 if (m->box_flags & STYL_BOX) {
117 for(i = 0; i < m->count_s; i++) {
122 m->style_entries = 0;
126 static void mov_text_cleanup_ftab(MovTextContext *m)
130 av_freep(&m->ftab_temp->font);
131 av_freep(&m->ftab_temp);
133 for(i = 0; i < m->count_f; i++) {
134 av_freep(&m->ftab[i]->font);
135 av_freep(&m->ftab[i]);
141 static int mov_text_tx3g(AVCodecContext *avctx, MovTextContext *m)
143 uint8_t *tx3g_ptr = avctx->extradata;
144 int i, box_size, font_length;
145 int8_t v_align, h_align;
151 box_size = BOX_SIZE_INITIAL; /* Size till ftab_entries */
152 if (avctx->extradata_size < box_size)
158 h_align = *tx3g_ptr++;
159 v_align = *tx3g_ptr++;
162 m->d.alignment = TOP_LEFT;
164 m->d.alignment = MIDDLE_LEFT;
166 m->d.alignment = BOTTOM_LEFT;
170 m->d.alignment = TOP_CENTER;
172 m->d.alignment = MIDDLE_CENTER;
174 m->d.alignment = BOTTOM_CENTER;
178 m->d.alignment = TOP_RIGHT;
180 m->d.alignment = MIDDLE_RIGHT;
182 m->d.alignment = BOTTOM_RIGHT;
185 m->d.back_color = AV_RB24(tx3g_ptr);
192 style_fontID = AV_RB16(tx3g_ptr);
195 s_default.style_flag = *tx3g_ptr++;
196 m->d.bold = !!(s_default.style_flag & STYLE_FLAG_BOLD);
197 m->d.italic = !!(s_default.style_flag & STYLE_FLAG_ITALIC);
198 m->d.underline = !!(s_default.style_flag & STYLE_FLAG_UNDERLINE);
200 m->d.fontsize = *tx3g_ptr++;
202 m->d.color = AV_RB24(tx3g_ptr);
210 m->ftab_entries = AV_RB16(tx3g_ptr);
213 for (i = 0; i < m->ftab_entries; i++) {
216 if (avctx->extradata_size < box_size) {
217 mov_text_cleanup_ftab(m);
221 m->ftab_temp = av_mallocz(sizeof(*m->ftab_temp));
223 mov_text_cleanup_ftab(m);
224 return AVERROR(ENOMEM);
226 m->ftab_temp->fontID = AV_RB16(tx3g_ptr);
228 font_length = *tx3g_ptr++;
230 box_size = box_size + font_length;
231 if (avctx->extradata_size < box_size) {
232 mov_text_cleanup_ftab(m);
236 m->ftab_temp->font = av_malloc(font_length + 1);
237 if (!m->ftab_temp->font) {
238 mov_text_cleanup_ftab(m);
239 return AVERROR(ENOMEM);
241 memcpy(m->ftab_temp->font, tx3g_ptr, font_length);
242 m->ftab_temp->font[font_length] = '\0';
243 av_dynarray_add(&m->ftab, &m->count_f, m->ftab_temp);
245 mov_text_cleanup_ftab(m);
246 return AVERROR(ENOMEM);
249 tx3g_ptr = tx3g_ptr + font_length;
251 for (i = 0; i < m->ftab_entries; i++) {
252 if (style_fontID == m->ftab[i]->fontID)
253 m->d.font = m->ftab[i]->font;
258 static int decode_twrp(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
260 m->box_flags |= TWRP_BOX;
261 m->w.wrap_flag = *tsmb++;
265 static int decode_hlit(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
267 m->box_flags |= HLIT_BOX;
268 m->h.hlit_start = AV_RB16(tsmb);
270 m->h.hlit_end = AV_RB16(tsmb);
275 static int decode_hclr(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
277 m->box_flags |= HCLR_BOX;
278 memcpy(m->c.hlit_color, tsmb, 4);
283 static int decode_styl(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
286 int style_entries = AV_RB16(tsmb);
288 // A single style record is of length 12 bytes.
289 if (m->tracksize + m->size_var + 2 + style_entries * 12 > avpkt->size)
292 m->style_entries = style_entries;
294 m->box_flags |= STYL_BOX;
295 for(i = 0; i < m->style_entries; i++) {
296 m->s_temp = av_malloc(sizeof(*m->s_temp));
299 return AVERROR(ENOMEM);
301 m->s_temp->style_start = AV_RB16(tsmb);
303 m->s_temp->style_end = AV_RB16(tsmb);
305 if ( m->s_temp->style_end < m->s_temp->style_start
306 || (m->count_s && m->s_temp->style_start < m->s[m->count_s - 1]->style_end)) {
307 av_freep(&m->s_temp);
309 return AVERROR(ENOMEM);
313 m->s_temp->style_fontID = AV_RB16(tsmb);
315 m->s_temp->style_flag = AV_RB8(tsmb);
317 m->s_temp->fontsize = AV_RB8(tsmb);
318 av_dynarray_add(&m->s, &m->count_s, m->s_temp);
321 return AVERROR(ENOMEM);
330 static const Box box_types[] = {
331 { MKBETAG('s','t','y','l'), 2, decode_styl },
332 { MKBETAG('h','l','i','t'), 4, decode_hlit },
333 { MKBETAG('h','c','l','r'), 4, decode_hclr },
334 { MKBETAG('t','w','r','p'), 1, decode_twrp }
337 const static size_t box_count = FF_ARRAY_ELEMS(box_types);
339 // Return byte length of the UTF-8 sequence starting at text[0]. 0 on error.
340 static int get_utf8_length_at(const char *text, const char *text_end)
342 const char *start = text;
345 GET_UTF8(c, text < text_end ? (uint8_t)*text++ : (err = 1, 0), goto error;);
353 static int text_to_ass(AVBPrint *buf, const char *text, const char *text_end,
354 AVCodecContext *avctx)
356 MovTextContext *m = avctx->priv_data;
359 int style_active = 0;
362 if (text < text_end && m->box_flags & TWRP_BOX) {
363 if (m->w.wrap_flag == 1) {
364 av_bprintf(buf, "{\\q1}"); /* End of line wrap */
366 av_bprintf(buf, "{\\q2}"); /* No wrap */
370 while (text < text_end) {
373 if ((m->box_flags & STYL_BOX) && entry < m->style_entries) {
374 if (text_pos == m->s[entry]->style_start) {
376 if (m->s[entry]->style_flag & STYLE_FLAG_BOLD)
377 av_bprintf(buf, "{\\b1}");
378 if (m->s[entry]->style_flag & STYLE_FLAG_ITALIC)
379 av_bprintf(buf, "{\\i1}");
380 if (m->s[entry]->style_flag & STYLE_FLAG_UNDERLINE)
381 av_bprintf(buf, "{\\u1}");
382 av_bprintf(buf, "{\\fs%d}", m->s[entry]->fontsize);
383 for (i = 0; i < m->ftab_entries; i++) {
384 if (m->s[entry]->style_fontID == m->ftab[i]->fontID)
385 av_bprintf(buf, "{\\fn%s}", m->ftab[i]->font);
388 if (text_pos == m->s[entry]->style_end) {
390 av_bprintf(buf, "{\\r}");
396 if (m->box_flags & HLIT_BOX) {
397 if (text_pos == m->h.hlit_start) {
398 /* If hclr box is present, set the secondary color to the color
399 * specified. Otherwise, set primary color to white and secondary
400 * color to black. These colors will come from TextSampleModifier
401 * boxes in future and inverse video technique for highlight will
404 if (m->box_flags & HCLR_BOX) {
405 av_bprintf(buf, "{\\2c&H%02x%02x%02x&}", m->c.hlit_color[2],
406 m->c.hlit_color[1], m->c.hlit_color[0]);
408 av_bprintf(buf, "{\\1c&H000000&}{\\2c&HFFFFFF&}");
411 if (text_pos == m->h.hlit_end) {
412 if (m->box_flags & HCLR_BOX) {
413 av_bprintf(buf, "{\\2c&H000000&}");
415 av_bprintf(buf, "{\\1c&HFFFFFF&}{\\2c&H000000&}");
420 len = get_utf8_length_at(text, text_end);
422 av_log(avctx, AV_LOG_ERROR, "invalid UTF-8 byte in subtitle\n");
425 for (i = 0; i < len; i++) {
430 av_bprintf(buf, "\\N");
433 av_bprint_chars(buf, *text, 1);
444 static int mov_text_init(AVCodecContext *avctx) {
446 * TODO: Handle the default text style.
447 * NB: Most players ignore styles completely, with the result that
448 * it's very common to find files where the default style is broken
449 * and respecting it results in a worse experience than ignoring it.
452 MovTextContext *m = avctx->priv_data;
453 ret = mov_text_tx3g(avctx, m);
455 return ff_ass_subtitle_header(avctx, m->d.font, m->d.fontsize,
456 RGB_TO_BGR(m->d.color),
457 RGB_TO_BGR(m->d.back_color),
458 m->d.bold, m->d.italic, m->d.underline,
459 ASS_DEFAULT_BORDERSTYLE, m->d.alignment);
461 return ff_ass_subtitle_header_default(avctx);
464 static int mov_text_decode_frame(AVCodecContext *avctx,
465 void *data, int *got_sub_ptr, AVPacket *avpkt)
467 AVSubtitle *sub = data;
468 MovTextContext *m = avctx->priv_data;
471 char *ptr = avpkt->data;
473 int text_length, tsmb_type, ret_tsmb;
478 if (!ptr || avpkt->size < 2)
479 return AVERROR_INVALIDDATA;
482 * A packet of size two with value zero is an empty subtitle
483 * used to mark the end of the previous non-empty subtitle.
484 * We can just drop them here as we have duration information
485 * already. If the value is non-zero, then it's technically a
488 if (avpkt->size == 2)
489 return AV_RB16(ptr) == 0 ? 0 : AVERROR_INVALIDDATA;
492 * The first two bytes of the packet are the length of the text string
493 * In complex cases, there are style descriptors appended to the string
494 * so we can't just assume the packet size is the string size.
496 text_length = AV_RB16(ptr);
497 end = ptr + FFMIN(2 + text_length, avpkt->size);
503 m->tracksize = 2 + text_length;
504 m->style_entries = 0;
507 // Note that the spec recommends lines be no longer than 2048 characters.
508 av_bprint_init(&buf, 0, AV_BPRINT_SIZE_UNLIMITED);
509 if (text_length + 2 != avpkt->size) {
510 while (m->tracksize + 8 <= avpkt->size) {
511 // A box is a minimum of 8 bytes.
512 tsmb = ptr + m->tracksize - 2;
513 tsmb_size = AV_RB32(tsmb);
515 tsmb_type = AV_RB32(tsmb);
518 if (tsmb_size == 1) {
519 if (m->tracksize + 16 > avpkt->size)
521 tsmb_size = AV_RB64(tsmb);
526 //size_var is equal to 8 or 16 depending on the size of box
528 if (tsmb_size == 0) {
529 av_log(avctx, AV_LOG_ERROR, "tsmb_size is 0\n");
530 return AVERROR_INVALIDDATA;
533 if (tsmb_size > avpkt->size - m->tracksize)
536 for (i = 0; i < box_count; i++) {
537 if (tsmb_type == box_types[i].type) {
538 if (m->tracksize + m->size_var + box_types[i].base_size > avpkt->size)
540 ret_tsmb = box_types[i].decode(tsmb, m, avpkt);
545 m->tracksize = m->tracksize + tsmb_size;
547 text_to_ass(&buf, ptr, end, avctx);
550 text_to_ass(&buf, ptr, end, avctx);
552 ret = ff_ass_add_rect(sub, buf.str, m->readorder++, 0, NULL, NULL);
553 av_bprint_finalize(&buf, NULL);
556 *got_sub_ptr = sub->num_rects > 0;
560 static int mov_text_decode_close(AVCodecContext *avctx)
562 MovTextContext *m = avctx->priv_data;
563 mov_text_cleanup_ftab(m);
568 static void mov_text_flush(AVCodecContext *avctx)
570 MovTextContext *m = avctx->priv_data;
571 if (!(avctx->flags2 & AV_CODEC_FLAG2_RO_FLUSH_NOOP))
575 AVCodec ff_movtext_decoder = {
577 .long_name = NULL_IF_CONFIG_SMALL("3GPP Timed Text subtitle"),
578 .type = AVMEDIA_TYPE_SUBTITLE,
579 .id = AV_CODEC_ID_MOV_TEXT,
580 .priv_data_size = sizeof(MovTextContext),
581 .init = mov_text_init,
582 .decode = mov_text_decode_frame,
583 .close = mov_text_decode_close,
584 .flush = mov_text_flush,