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];
106 FontRecord *ftab_temp;
110 uint16_t style_entries, ftab_entries;
113 int count_s, count_f;
122 int (*decode)(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt);
125 static void mov_text_cleanup(MovTextContext *m)
128 if (m->box_flags & STYL_BOX) {
129 for(i = 0; i < m->count_s; i++) {
134 m->style_entries = 0;
138 static void mov_text_cleanup_ftab(MovTextContext *m)
142 av_freep(&m->ftab_temp->font);
143 av_freep(&m->ftab_temp);
145 for(i = 0; i < m->count_f; i++) {
146 av_freep(&m->ftab[i]->font);
147 av_freep(&m->ftab[i]);
154 static int mov_text_tx3g(AVCodecContext *avctx, MovTextContext *m)
156 uint8_t *tx3g_ptr = avctx->extradata;
157 int i, box_size, font_length;
158 int8_t v_align, h_align;
163 box_size = BOX_SIZE_INITIAL; /* Size till ftab_entries */
164 if (avctx->extradata_size < box_size)
170 h_align = *tx3g_ptr++;
171 v_align = *tx3g_ptr++;
174 m->d.alignment = TOP_LEFT;
176 m->d.alignment = MIDDLE_LEFT;
178 m->d.alignment = BOTTOM_LEFT;
182 m->d.alignment = TOP_CENTER;
184 m->d.alignment = MIDDLE_CENTER;
186 m->d.alignment = BOTTOM_CENTER;
190 m->d.alignment = TOP_RIGHT;
192 m->d.alignment = MIDDLE_RIGHT;
194 m->d.alignment = BOTTOM_RIGHT;
197 m->d.back_color = AV_RB24(tx3g_ptr);
199 m->d.back_alpha = AV_RB8(tx3g_ptr);
206 m->d.fontID = AV_RB16(tx3g_ptr);
209 s_default.style_flag = *tx3g_ptr++;
210 m->d.bold = !!(s_default.style_flag & STYLE_FLAG_BOLD);
211 m->d.italic = !!(s_default.style_flag & STYLE_FLAG_ITALIC);
212 m->d.underline = !!(s_default.style_flag & STYLE_FLAG_UNDERLINE);
214 m->d.fontsize = *tx3g_ptr++;
216 m->d.color = AV_RB24(tx3g_ptr);
218 m->d.alpha = AV_RB8(tx3g_ptr);
226 m->ftab_entries = AV_RB16(tx3g_ptr);
229 for (i = 0; i < m->ftab_entries; i++) {
232 if (avctx->extradata_size < box_size) {
233 mov_text_cleanup_ftab(m);
236 m->ftab_temp = av_mallocz(sizeof(*m->ftab_temp));
238 mov_text_cleanup_ftab(m);
239 return AVERROR(ENOMEM);
241 m->ftab_temp->fontID = AV_RB16(tx3g_ptr);
243 font_length = *tx3g_ptr++;
245 box_size = box_size + font_length;
246 if (avctx->extradata_size < box_size) {
247 mov_text_cleanup_ftab(m);
250 m->ftab_temp->font = av_malloc(font_length + 1);
251 if (!m->ftab_temp->font) {
252 mov_text_cleanup_ftab(m);
253 return AVERROR(ENOMEM);
255 memcpy(m->ftab_temp->font, tx3g_ptr, font_length);
256 m->ftab_temp->font[font_length] = '\0';
257 av_dynarray_add(&m->ftab, &m->count_f, m->ftab_temp);
259 mov_text_cleanup_ftab(m);
260 return AVERROR(ENOMEM);
263 tx3g_ptr = tx3g_ptr + font_length;
265 // In case of broken header, init default font
266 m->d.font = ASS_DEFAULT_FONT;
267 for (i = 0; i < m->ftab_entries; i++) {
268 if (m->d.fontID == m->ftab[i]->fontID)
269 m->d.font = m->ftab[i]->font;
274 static int decode_twrp(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
276 m->box_flags |= TWRP_BOX;
277 m->w.wrap_flag = *tsmb++;
281 static int decode_hlit(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
283 m->box_flags |= HLIT_BOX;
284 m->h.hlit_start = AV_RB16(tsmb);
286 m->h.hlit_end = AV_RB16(tsmb);
291 static int decode_hclr(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
293 m->box_flags |= HCLR_BOX;
294 memcpy(m->c.hlit_color, tsmb, 4);
299 static int decode_styl(const uint8_t *tsmb, MovTextContext *m, AVPacket *avpkt)
302 int style_entries = AV_RB16(tsmb);
304 // A single style record is of length 12 bytes.
305 if (m->tracksize + m->size_var + 2 + style_entries * 12 > avpkt->size)
308 m->style_entries = style_entries;
310 m->box_flags |= STYL_BOX;
311 for(i = 0; i < m->style_entries; i++) {
312 m->s_temp = av_malloc(sizeof(*m->s_temp));
315 return AVERROR(ENOMEM);
317 m->s_temp->style_start = AV_RB16(tsmb);
319 m->s_temp->style_end = AV_RB16(tsmb);
321 if ( m->s_temp->style_end < m->s_temp->style_start
322 || (m->count_s && m->s_temp->style_start < m->s[m->count_s - 1]->style_end)) {
323 av_freep(&m->s_temp);
325 return AVERROR(ENOMEM);
329 m->s_temp->style_fontID = AV_RB16(tsmb);
331 m->s_temp->style_flag = AV_RB8(tsmb);
332 m->s_temp->bold = !!(m->s_temp->style_flag & STYLE_FLAG_BOLD);
333 m->s_temp->italic = !!(m->s_temp->style_flag & STYLE_FLAG_ITALIC);
334 m->s_temp->underline = !!(m->s_temp->style_flag & STYLE_FLAG_UNDERLINE);
336 m->s_temp->fontsize = AV_RB8(tsmb);
338 m->s_temp->color = AV_RB24(tsmb);
340 m->s_temp->alpha = AV_RB8(tsmb);
342 av_dynarray_add(&m->s, &m->count_s, m->s_temp);
345 return AVERROR(ENOMEM);
351 static const Box box_types[] = {
352 { MKBETAG('s','t','y','l'), 2, decode_styl },
353 { MKBETAG('h','l','i','t'), 4, decode_hlit },
354 { MKBETAG('h','c','l','r'), 4, decode_hclr },
355 { MKBETAG('t','w','r','p'), 1, decode_twrp }
358 const static size_t box_count = FF_ARRAY_ELEMS(box_types);
360 // Return byte length of the UTF-8 sequence starting at text[0]. 0 on error.
361 static int get_utf8_length_at(const char *text, const char *text_end)
363 const char *start = text;
366 GET_UTF8(c, text < text_end ? (uint8_t)*text++ : (err = 1, 0), goto error;);
374 static int text_to_ass(AVBPrint *buf, const char *text, const char *text_end,
375 AVCodecContext *avctx)
377 MovTextContext *m = avctx->priv_data;
380 int style_active = 0;
382 int color = m->d.color;
384 if (text < text_end && m->box_flags & TWRP_BOX) {
385 if (m->w.wrap_flag == 1) {
386 av_bprintf(buf, "{\\q1}"); /* End of line wrap */
388 av_bprintf(buf, "{\\q2}"); /* No wrap */
392 while (text < text_end) {
395 if ((m->box_flags & STYL_BOX) && entry < m->style_entries) {
396 if (text_pos == m->s[entry]->style_start) {
398 if (m->s[entry]->bold ^ m->d.bold)
399 av_bprintf(buf, "{\\b%d}", m->s[entry]->bold);
400 if (m->s[entry]->italic ^ m->d.italic)
401 av_bprintf(buf, "{\\i%d}", m->s[entry]->italic);
402 if (m->s[entry]->underline ^ m->d.underline)
403 av_bprintf(buf, "{\\u%d}", m->s[entry]->underline);
404 if (m->s[entry]->fontsize != m->d.fontsize)
405 av_bprintf(buf, "{\\fs%d}", m->s[entry]->fontsize);
406 if (m->s[entry]->style_fontID != m->d.fontID)
407 for (i = 0; i < m->ftab_entries; i++) {
408 if (m->s[entry]->style_fontID == m->ftab[i]->fontID)
409 av_bprintf(buf, "{\\fn%s}", m->ftab[i]->font);
411 if (m->d.color != m->s[entry]->color) {
412 color = m->s[entry]->color;
413 av_bprintf(buf, "{\\1c&H%X&}", RGB_TO_BGR(color));
415 if (m->d.alpha != m->s[entry]->alpha)
416 av_bprintf(buf, "{\\1a&H%02X&}", 255 - m->s[entry]->alpha);
418 if (text_pos == m->s[entry]->style_end) {
420 av_bprintf(buf, "{\\r}");
427 if (m->box_flags & HLIT_BOX) {
428 if (text_pos == m->h.hlit_start) {
429 /* If hclr box is present, set the secondary color to the color
430 * specified. Otherwise, set primary color to white and secondary
431 * color to black. These colors will come from TextSampleModifier
432 * boxes in future and inverse video technique for highlight will
435 if (m->box_flags & HCLR_BOX) {
436 av_bprintf(buf, "{\\2c&H%02x%02x%02x&}", m->c.hlit_color[2],
437 m->c.hlit_color[1], m->c.hlit_color[0]);
439 av_bprintf(buf, "{\\1c&H000000&}{\\2c&HFFFFFF&}");
442 if (text_pos == m->h.hlit_end) {
443 if (m->box_flags & HCLR_BOX) {
444 av_bprintf(buf, "{\\2c&H%X&}", RGB_TO_BGR(m->d.color));
446 av_bprintf(buf, "{\\1c&H%X&}{\\2c&H%X&}",
447 RGB_TO_BGR(color), RGB_TO_BGR(m->d.color));
452 len = get_utf8_length_at(text, text_end);
454 av_log(avctx, AV_LOG_ERROR, "invalid UTF-8 byte in subtitle\n");
457 for (i = 0; i < len; i++) {
462 av_bprintf(buf, "\\N");
465 av_bprint_chars(buf, *text, 1);
476 static int mov_text_init(AVCodecContext *avctx) {
478 * TODO: Handle the default text style.
479 * NB: Most players ignore styles completely, with the result that
480 * it's very common to find files where the default style is broken
481 * and respecting it results in a worse experience than ignoring it.
484 MovTextContext *m = avctx->priv_data;
485 ret = mov_text_tx3g(avctx, m);
487 if (!m->frame_width || !m->frame_height) {
488 m->frame_width = ASS_DEFAULT_PLAYRESX;
489 m->frame_height = ASS_DEFAULT_PLAYRESY;
491 return ff_ass_subtitle_header_full(avctx,
492 m->frame_width, m->frame_height,
493 m->d.font, m->d.fontsize,
494 (255U - m->d.alpha) << 24 | RGB_TO_BGR(m->d.color),
495 (255U - m->d.alpha) << 24 | RGB_TO_BGR(m->d.color),
496 (255U - m->d.back_alpha) << 24 | RGB_TO_BGR(m->d.back_color),
497 (255U - m->d.back_alpha) << 24 | RGB_TO_BGR(m->d.back_color),
498 m->d.bold, m->d.italic, m->d.underline,
499 ASS_DEFAULT_BORDERSTYLE, m->d.alignment);
501 return ff_ass_subtitle_header_default(avctx);
504 static int mov_text_decode_frame(AVCodecContext *avctx,
505 void *data, int *got_sub_ptr, AVPacket *avpkt)
507 AVSubtitle *sub = data;
508 MovTextContext *m = avctx->priv_data;
511 char *ptr = avpkt->data;
513 int text_length, tsmb_type, ret_tsmb;
518 if (!ptr || avpkt->size < 2)
519 return AVERROR_INVALIDDATA;
522 * A packet of size two with value zero is an empty subtitle
523 * used to mark the end of the previous non-empty subtitle.
524 * We can just drop them here as we have duration information
525 * already. If the value is non-zero, then it's technically a
528 if (avpkt->size == 2)
529 return AV_RB16(ptr) == 0 ? 0 : AVERROR_INVALIDDATA;
532 * The first two bytes of the packet are the length of the text string
533 * In complex cases, there are style descriptors appended to the string
534 * so we can't just assume the packet size is the string size.
536 text_length = AV_RB16(ptr);
537 end = ptr + FFMIN(2 + text_length, avpkt->size);
543 m->tracksize = 2 + text_length;
544 m->style_entries = 0;
547 // Note that the spec recommends lines be no longer than 2048 characters.
548 av_bprint_init(&buf, 0, AV_BPRINT_SIZE_UNLIMITED);
549 if (text_length + 2 != avpkt->size) {
550 while (m->tracksize + 8 <= avpkt->size) {
551 // A box is a minimum of 8 bytes.
552 tsmb = ptr + m->tracksize - 2;
553 tsmb_size = AV_RB32(tsmb);
555 tsmb_type = AV_RB32(tsmb);
558 if (tsmb_size == 1) {
559 if (m->tracksize + 16 > avpkt->size)
561 tsmb_size = AV_RB64(tsmb);
566 //size_var is equal to 8 or 16 depending on the size of box
568 if (tsmb_size == 0) {
569 av_log(avctx, AV_LOG_ERROR, "tsmb_size is 0\n");
570 return AVERROR_INVALIDDATA;
573 if (tsmb_size > avpkt->size - m->tracksize)
576 for (i = 0; i < box_count; i++) {
577 if (tsmb_type == box_types[i].type) {
578 if (m->tracksize + m->size_var + box_types[i].base_size > avpkt->size)
580 ret_tsmb = box_types[i].decode(tsmb, m, avpkt);
585 m->tracksize = m->tracksize + tsmb_size;
587 text_to_ass(&buf, ptr, end, avctx);
590 text_to_ass(&buf, ptr, end, avctx);
592 ret = ff_ass_add_rect(sub, buf.str, m->readorder++, 0, NULL, NULL);
593 av_bprint_finalize(&buf, NULL);
596 *got_sub_ptr = sub->num_rects > 0;
600 static int mov_text_decode_close(AVCodecContext *avctx)
602 MovTextContext *m = avctx->priv_data;
603 mov_text_cleanup_ftab(m);
608 static void mov_text_flush(AVCodecContext *avctx)
610 MovTextContext *m = avctx->priv_data;
611 if (!(avctx->flags2 & AV_CODEC_FLAG2_RO_FLUSH_NOOP))
615 #define OFFSET(x) offsetof(MovTextContext, x)
616 #define FLAGS AV_OPT_FLAG_DECODING_PARAM | AV_OPT_FLAG_SUBTITLE_PARAM
617 static const AVOption options[] = {
618 { "width", "Frame width, usually video width", OFFSET(frame_width), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
619 { "height", "Frame height, usually video height", OFFSET(frame_height), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
623 static const AVClass mov_text_decoder_class = {
624 .class_name = "MOV text decoder",
625 .item_name = av_default_item_name,
627 .version = LIBAVUTIL_VERSION_INT,
630 AVCodec ff_movtext_decoder = {
632 .long_name = NULL_IF_CONFIG_SMALL("3GPP Timed Text subtitle"),
633 .type = AVMEDIA_TYPE_SUBTITLE,
634 .id = AV_CODEC_ID_MOV_TEXT,
635 .priv_data_size = sizeof(MovTextContext),
636 .priv_class = &mov_text_decoder_class,
637 .init = mov_text_init,
638 .decode = mov_text_decode_frame,
639 .close = mov_text_decode_close,
640 .flush = mov_text_flush,