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"
30 #include "bytestream.h"
32 #define STYLE_FLAG_BOLD (1<<0)
33 #define STYLE_FLAG_ITALIC (1<<1)
34 #define STYLE_FLAG_UNDERLINE (1<<2)
36 #define BOX_SIZE_INITIAL 40
38 #define STYL_BOX (1<<0)
39 #define HLIT_BOX (1<<1)
40 #define HCLR_BOX (1<<2)
41 #define TWRP_BOX (1<<3)
44 #define BOTTOM_CENTER 2
45 #define BOTTOM_RIGHT 3
47 #define MIDDLE_CENTER 5
48 #define MIDDLE_RIGHT 6
53 #define RGB_TO_BGR(c) (((c) & 0xff) << 16 | ((c) & 0xff00) | (((c) >> 16) & 0xff))
84 uint16_t style_fontID;
93 uint8_t hlit_color[4];
109 uint16_t style_entries, ftab_entries;
120 int (*decode)(const uint8_t *tsmb, MovTextContext *m, const AVPacket *avpkt);
123 static void mov_text_cleanup(MovTextContext *m)
125 if (m->box_flags & STYL_BOX) {
127 m->style_entries = 0;
131 static void mov_text_cleanup_ftab(MovTextContext *m)
133 for (unsigned i = 0; i < m->ftab_entries; i++)
134 av_freep(&m->ftab[i].font);
139 static int mov_text_tx3g(AVCodecContext *avctx, MovTextContext *m)
141 const uint8_t *tx3g_ptr = avctx->extradata;
142 int i, j = -1, font_length, remaining = avctx->extradata_size - BOX_SIZE_INITIAL;
143 int8_t v_align, h_align;
144 unsigned ftab_entries;
154 h_align = bytestream_get_byte(&tx3g_ptr);
155 v_align = bytestream_get_byte(&tx3g_ptr);
158 m->d.alignment = TOP_LEFT;
160 m->d.alignment = MIDDLE_LEFT;
162 m->d.alignment = BOTTOM_LEFT;
166 m->d.alignment = TOP_CENTER;
168 m->d.alignment = MIDDLE_CENTER;
170 m->d.alignment = BOTTOM_CENTER;
174 m->d.alignment = TOP_RIGHT;
176 m->d.alignment = MIDDLE_RIGHT;
178 m->d.alignment = BOTTOM_RIGHT;
181 m->d.back_color = bytestream_get_be24(&tx3g_ptr);
182 m->d.back_alpha = bytestream_get_byte(&tx3g_ptr);
188 m->d.fontID = bytestream_get_be16(&tx3g_ptr);
190 s_default.style_flag = bytestream_get_byte(&tx3g_ptr);
191 m->d.bold = !!(s_default.style_flag & STYLE_FLAG_BOLD);
192 m->d.italic = !!(s_default.style_flag & STYLE_FLAG_ITALIC);
193 m->d.underline = !!(s_default.style_flag & STYLE_FLAG_UNDERLINE);
195 m->d.fontsize = bytestream_get_byte(&tx3g_ptr);
197 m->d.color = bytestream_get_be24(&tx3g_ptr);
198 m->d.alpha = bytestream_get_byte(&tx3g_ptr);
205 // In case of broken header, init default font
206 m->d.font = ASS_DEFAULT_FONT;
208 ftab_entries = bytestream_get_be16(&tx3g_ptr);
211 remaining -= 3 * ftab_entries;
213 return AVERROR_INVALIDDATA;
214 m->ftab = av_calloc(ftab_entries, sizeof(*m->ftab));
216 return AVERROR(ENOMEM);
217 m->ftab_entries = ftab_entries;
219 for (i = 0; i < m->ftab_entries; i++) {
220 m->ftab[i].fontID = bytestream_get_be16(&tx3g_ptr);
221 if (m->ftab[i].fontID == m->d.fontID)
223 font_length = bytestream_get_byte(&tx3g_ptr);
225 remaining -= font_length;
227 mov_text_cleanup_ftab(m);
230 m->ftab[i].font = av_malloc(font_length + 1);
231 if (!m->ftab[i].font) {
232 mov_text_cleanup_ftab(m);
233 return AVERROR(ENOMEM);
235 bytestream_get_buffer(&tx3g_ptr, m->ftab[i].font, font_length);
236 m->ftab[i].font[font_length] = '\0';
239 m->d.font = m->ftab[j].font;
243 static int decode_twrp(const uint8_t *tsmb, MovTextContext *m, const AVPacket *avpkt)
245 m->box_flags |= TWRP_BOX;
246 m->w.wrap_flag = bytestream_get_byte(&tsmb);
250 static int decode_hlit(const uint8_t *tsmb, MovTextContext *m, const AVPacket *avpkt)
252 m->box_flags |= HLIT_BOX;
253 m->h.hlit_start = bytestream_get_be16(&tsmb);
254 m->h.hlit_end = bytestream_get_be16(&tsmb);
258 static int decode_hclr(const uint8_t *tsmb, MovTextContext *m, const AVPacket *avpkt)
260 m->box_flags |= HCLR_BOX;
261 bytestream_get_buffer(&tsmb, m->c.hlit_color, 4);
265 static int decode_styl(const uint8_t *tsmb, MovTextContext *m, const AVPacket *avpkt)
268 int style_entries = bytestream_get_be16(&tsmb);
271 // A single style record is of length 12 bytes.
272 if (m->tracksize + m->size_var + 2 + style_entries * 12 > avpkt->size)
275 tmp = av_realloc_array(m->s, style_entries, sizeof(*m->s));
277 return AVERROR(ENOMEM);
279 m->style_entries = style_entries;
281 m->box_flags |= STYL_BOX;
282 for(i = 0; i < m->style_entries; i++) {
283 StyleBox *style = &m->s[i];
285 style->style_start = bytestream_get_be16(&tsmb);
286 style->style_end = bytestream_get_be16(&tsmb);
287 if ( style->style_end < style->style_start
288 || (i && style->style_start < m->s[i - 1].style_end)) {
290 return AVERROR(ENOMEM);
292 if (style->style_start == style->style_end) {
293 /* Skip this style as it applies to no character */
300 style->style_fontID = bytestream_get_be16(&tsmb);
301 style->style_flag = bytestream_get_byte(&tsmb);
302 style->bold = !!(style->style_flag & STYLE_FLAG_BOLD);
303 style->italic = !!(style->style_flag & STYLE_FLAG_ITALIC);
304 style->underline = !!(style->style_flag & STYLE_FLAG_UNDERLINE);
305 style->fontsize = bytestream_get_byte(&tsmb);
306 style->color = bytestream_get_be24(&tsmb);
307 style->alpha = bytestream_get_byte(&tsmb);
312 static const Box box_types[] = {
313 { MKBETAG('s','t','y','l'), 2, decode_styl },
314 { MKBETAG('h','l','i','t'), 4, decode_hlit },
315 { MKBETAG('h','c','l','r'), 4, decode_hclr },
316 { MKBETAG('t','w','r','p'), 1, decode_twrp }
319 const static size_t box_count = FF_ARRAY_ELEMS(box_types);
321 // Return byte length of the UTF-8 sequence starting at text[0]. 0 on error.
322 static int get_utf8_length_at(const char *text, const char *text_end)
324 const char *start = text;
327 GET_UTF8(c, text < text_end ? (uint8_t)*text++ : (err = 1, 0), goto error;);
335 static int text_to_ass(AVBPrint *buf, const char *text, const char *text_end,
336 AVCodecContext *avctx)
338 MovTextContext *m = avctx->priv_data;
342 int color = m->d.color;
344 if (text < text_end && m->box_flags & TWRP_BOX) {
345 if (m->w.wrap_flag == 1) {
346 av_bprintf(buf, "{\\q1}"); /* End of line wrap */
348 av_bprintf(buf, "{\\q2}"); /* No wrap */
352 while (text < text_end) {
355 if ((m->box_flags & STYL_BOX) && entry < m->style_entries) {
356 const StyleBox *style = &m->s[entry];
357 if (text_pos == style->style_end) {
358 av_bprintf(buf, "{\\r}");
363 if (entry < m->style_entries && text_pos == style->style_start) {
364 if (style->bold ^ m->d.bold)
365 av_bprintf(buf, "{\\b%d}", style->bold);
366 if (style->italic ^ m->d.italic)
367 av_bprintf(buf, "{\\i%d}", style->italic);
368 if (style->underline ^ m->d.underline)
369 av_bprintf(buf, "{\\u%d}", style->underline);
370 if (style->fontsize != m->d.fontsize)
371 av_bprintf(buf, "{\\fs%d}", style->fontsize);
372 if (style->style_fontID != m->d.fontID)
373 for (i = 0; i < m->ftab_entries; i++) {
374 if (style->style_fontID == m->ftab[i].fontID)
375 av_bprintf(buf, "{\\fn%s}", m->ftab[i].font);
377 if (m->d.color != style->color) {
378 color = style->color;
379 av_bprintf(buf, "{\\1c&H%X&}", RGB_TO_BGR(color));
381 if (m->d.alpha != style->alpha)
382 av_bprintf(buf, "{\\1a&H%02X&}", 255 - style->alpha);
385 if (m->box_flags & HLIT_BOX) {
386 if (text_pos == m->h.hlit_start) {
387 /* If hclr box is present, set the secondary color to the color
388 * specified. Otherwise, set primary color to white and secondary
389 * color to black. These colors will come from TextSampleModifier
390 * boxes in future and inverse video technique for highlight will
393 if (m->box_flags & HCLR_BOX) {
394 av_bprintf(buf, "{\\2c&H%02x%02x%02x&}", m->c.hlit_color[2],
395 m->c.hlit_color[1], m->c.hlit_color[0]);
397 av_bprintf(buf, "{\\1c&H000000&}{\\2c&HFFFFFF&}");
400 if (text_pos == m->h.hlit_end) {
401 if (m->box_flags & HCLR_BOX) {
402 av_bprintf(buf, "{\\2c&H%X&}", RGB_TO_BGR(m->d.color));
404 av_bprintf(buf, "{\\1c&H%X&}{\\2c&H%X&}",
405 RGB_TO_BGR(color), RGB_TO_BGR(m->d.color));
410 len = get_utf8_length_at(text, text_end);
412 av_log(avctx, AV_LOG_ERROR, "invalid UTF-8 byte in subtitle\n");
419 av_bprintf(buf, "\\N");
422 av_bprint_append_data(buf, text, len);
432 static int mov_text_init(AVCodecContext *avctx) {
434 * TODO: Handle the default text style.
435 * NB: Most players ignore styles completely, with the result that
436 * it's very common to find files where the default style is broken
437 * and respecting it results in a worse experience than ignoring it.
440 MovTextContext *m = avctx->priv_data;
441 ret = mov_text_tx3g(avctx, m);
443 if (!m->frame_width || !m->frame_height) {
444 m->frame_width = ASS_DEFAULT_PLAYRESX;
445 m->frame_height = ASS_DEFAULT_PLAYRESY;
447 return ff_ass_subtitle_header_full(avctx,
448 m->frame_width, m->frame_height,
449 m->d.font, m->d.fontsize,
450 (255U - m->d.alpha) << 24 | RGB_TO_BGR(m->d.color),
451 (255U - m->d.alpha) << 24 | RGB_TO_BGR(m->d.color),
452 (255U - m->d.back_alpha) << 24 | RGB_TO_BGR(m->d.back_color),
453 (255U - m->d.back_alpha) << 24 | RGB_TO_BGR(m->d.back_color),
454 m->d.bold, m->d.italic, m->d.underline,
455 ASS_DEFAULT_BORDERSTYLE, m->d.alignment);
457 return ff_ass_subtitle_header_default(avctx);
460 static int mov_text_decode_frame(AVCodecContext *avctx,
461 void *data, int *got_sub_ptr, AVPacket *avpkt)
463 AVSubtitle *sub = data;
464 MovTextContext *m = avctx->priv_data;
467 char *ptr = avpkt->data;
469 int text_length, tsmb_type, ret_tsmb;
474 if (!ptr || avpkt->size < 2)
475 return AVERROR_INVALIDDATA;
478 * A packet of size two with value zero is an empty subtitle
479 * used to mark the end of the previous non-empty subtitle.
480 * We can just drop them here as we have duration information
481 * already. If the value is non-zero, then it's technically a
484 if (avpkt->size == 2)
485 return AV_RB16(ptr) == 0 ? 0 : AVERROR_INVALIDDATA;
488 * The first two bytes of the packet are the length of the text string
489 * In complex cases, there are style descriptors appended to the string
490 * so we can't just assume the packet size is the string size.
492 text_length = AV_RB16(ptr);
493 end = ptr + FFMIN(2 + text_length, avpkt->size);
499 m->tracksize = 2 + text_length;
500 m->style_entries = 0;
502 // Note that the spec recommends lines be no longer than 2048 characters.
503 av_bprint_init(&buf, 0, AV_BPRINT_SIZE_UNLIMITED);
504 if (text_length + 2 != avpkt->size) {
505 while (m->tracksize + 8 <= avpkt->size) {
506 // A box is a minimum of 8 bytes.
507 tsmb = ptr + m->tracksize - 2;
508 tsmb_size = AV_RB32(tsmb);
510 tsmb_type = AV_RB32(tsmb);
513 if (tsmb_size == 1) {
514 if (m->tracksize + 16 > avpkt->size)
516 tsmb_size = AV_RB64(tsmb);
521 //size_var is equal to 8 or 16 depending on the size of box
523 if (tsmb_size == 0) {
524 av_log(avctx, AV_LOG_ERROR, "tsmb_size is 0\n");
525 return AVERROR_INVALIDDATA;
528 if (tsmb_size > avpkt->size - m->tracksize)
531 for (i = 0; i < box_count; i++) {
532 if (tsmb_type == box_types[i].type) {
533 if (m->tracksize + m->size_var + box_types[i].base_size > avpkt->size)
535 ret_tsmb = box_types[i].decode(tsmb, m, avpkt);
540 m->tracksize = m->tracksize + tsmb_size;
542 text_to_ass(&buf, ptr, end, avctx);
545 text_to_ass(&buf, ptr, end, avctx);
547 ret = ff_ass_add_rect(sub, buf.str, m->readorder++, 0, NULL, NULL);
548 av_bprint_finalize(&buf, NULL);
551 *got_sub_ptr = sub->num_rects > 0;
555 static int mov_text_decode_close(AVCodecContext *avctx)
557 MovTextContext *m = avctx->priv_data;
558 mov_text_cleanup_ftab(m);
563 static void mov_text_flush(AVCodecContext *avctx)
565 MovTextContext *m = avctx->priv_data;
566 if (!(avctx->flags2 & AV_CODEC_FLAG2_RO_FLUSH_NOOP))
570 #define OFFSET(x) offsetof(MovTextContext, x)
571 #define FLAGS AV_OPT_FLAG_DECODING_PARAM | AV_OPT_FLAG_SUBTITLE_PARAM
572 static const AVOption options[] = {
573 { "width", "Frame width, usually video width", OFFSET(frame_width), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
574 { "height", "Frame height, usually video height", OFFSET(frame_height), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
578 static const AVClass mov_text_decoder_class = {
579 .class_name = "MOV text decoder",
580 .item_name = av_default_item_name,
582 .version = LIBAVUTIL_VERSION_INT,
585 const AVCodec ff_movtext_decoder = {
587 .long_name = NULL_IF_CONFIG_SMALL("3GPP Timed Text subtitle"),
588 .type = AVMEDIA_TYPE_SUBTITLE,
589 .id = AV_CODEC_ID_MOV_TEXT,
590 .priv_data_size = sizeof(MovTextContext),
591 .priv_class = &mov_text_decoder_class,
592 .init = mov_text_init,
593 .decode = mov_text_decode_frame,
594 .close = mov_text_decode_close,
595 .flush = mov_text_flush,