avformat/mpeg: demux ivtv captions
[ffmpeg.git] / libavcodec / movtextdec.c
blobf799252bf212938506f940a58c726dff169fa8fb
1 /*
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
22 #include "avcodec.h"
23 #include "ass.h"
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"
31 #include "codec_internal.h"
33 #define STYLE_FLAG_BOLD (1<<0)
34 #define STYLE_FLAG_ITALIC (1<<1)
35 #define STYLE_FLAG_UNDERLINE (1<<2)
37 #define BOX_SIZE_INITIAL 40
39 #define STYL_BOX (1<<0)
40 #define HLIT_BOX (1<<1)
41 #define HCLR_BOX (1<<2)
42 #define TWRP_BOX (1<<3)
44 #define BOTTOM_LEFT 1
45 #define BOTTOM_CENTER 2
46 #define BOTTOM_RIGHT 3
47 #define MIDDLE_LEFT 4
48 #define MIDDLE_CENTER 5
49 #define MIDDLE_RIGHT 6
50 #define TOP_LEFT 7
51 #define TOP_CENTER 8
52 #define TOP_RIGHT 9
54 #define RGB_TO_BGR(c) (((c) & 0xff) << 16 | ((c) & 0xff00) | (((c) >> 16) & 0xff))
56 typedef struct {
57 uint16_t font_id;
58 char *font;
59 } FontRecord;
61 typedef struct {
62 uint16_t start;
63 uint16_t end;
64 uint8_t flags;
65 uint8_t bold;
66 uint8_t italic;
67 uint8_t underline;
68 int color;
69 uint8_t alpha;
70 uint8_t fontsize;
71 uint16_t font_id;
72 } StyleBox;
74 typedef struct {
75 StyleBox style;
76 const char *font;
77 int back_color;
78 uint8_t back_alpha;
79 int alignment;
80 } MovTextDefault;
82 typedef struct {
83 uint16_t hlit_start;
84 uint16_t hlit_end;
85 } HighlightBox;
87 typedef struct {
88 uint8_t hlit_color[4];
89 } HilightcolorBox;
91 typedef struct {
92 uint8_t wrap_flag;
93 } TextWrapBox;
95 typedef struct {
96 AVClass *class;
97 StyleBox *s;
98 HighlightBox h;
99 HilightcolorBox c;
100 FontRecord *ftab;
101 TextWrapBox w;
102 MovTextDefault d;
103 uint8_t box_flags;
104 uint16_t style_entries, ftab_entries;
105 int readorder;
106 int frame_width;
107 int frame_height;
108 } MovTextContext;
110 typedef struct {
111 uint32_t type;
112 unsigned base_size;
113 int (*decode)(const uint8_t *tsmb, MovTextContext *m, uint64_t size);
114 } Box;
116 static void mov_text_cleanup(MovTextContext *m)
118 if (m->box_flags & STYL_BOX) {
119 av_freep(&m->s);
120 m->style_entries = 0;
124 static void mov_text_cleanup_ftab(MovTextContext *m)
126 for (unsigned i = 0; i < m->ftab_entries; i++)
127 av_freep(&m->ftab[i].font);
128 av_freep(&m->ftab);
129 m->ftab_entries = 0;
132 static void mov_text_parse_style_record(StyleBox *style, const uint8_t **ptr)
134 // fontID
135 style->font_id = bytestream_get_be16(ptr);
136 // face-style-flags
137 style->flags = bytestream_get_byte(ptr);
138 style->bold = !!(style->flags & STYLE_FLAG_BOLD);
139 style->italic = !!(style->flags & STYLE_FLAG_ITALIC);
140 style->underline = !!(style->flags & STYLE_FLAG_UNDERLINE);
141 // fontsize
142 style->fontsize = bytestream_get_byte(ptr);
143 // Primary color
144 style->color = bytestream_get_be24(ptr);
145 style->color = RGB_TO_BGR(style->color);
146 style->alpha = bytestream_get_byte(ptr);
149 static int mov_text_tx3g(AVCodecContext *avctx, MovTextContext *m)
151 const uint8_t *tx3g_ptr = avctx->extradata;
152 int i, j = -1, font_length, remaining = avctx->extradata_size - BOX_SIZE_INITIAL;
153 int8_t v_align, h_align;
154 unsigned ftab_entries;
156 m->ftab_entries = 0;
157 if (remaining < 0)
158 return -1;
160 // Display Flags
161 tx3g_ptr += 4;
162 // Alignment
163 h_align = bytestream_get_byte(&tx3g_ptr);
164 v_align = bytestream_get_byte(&tx3g_ptr);
165 if (h_align == 0) {
166 if (v_align == 0)
167 m->d.alignment = TOP_LEFT;
168 if (v_align == 1)
169 m->d.alignment = MIDDLE_LEFT;
170 if (v_align == -1)
171 m->d.alignment = BOTTOM_LEFT;
173 if (h_align == 1) {
174 if (v_align == 0)
175 m->d.alignment = TOP_CENTER;
176 if (v_align == 1)
177 m->d.alignment = MIDDLE_CENTER;
178 if (v_align == -1)
179 m->d.alignment = BOTTOM_CENTER;
181 if (h_align == -1) {
182 if (v_align == 0)
183 m->d.alignment = TOP_RIGHT;
184 if (v_align == 1)
185 m->d.alignment = MIDDLE_RIGHT;
186 if (v_align == -1)
187 m->d.alignment = BOTTOM_RIGHT;
189 // Background Color
190 m->d.back_color = bytestream_get_be24(&tx3g_ptr);
191 m->d.back_color = RGB_TO_BGR(m->d.back_color);
192 m->d.back_alpha = bytestream_get_byte(&tx3g_ptr);
193 // BoxRecord
194 tx3g_ptr += 8;
195 // StyleRecord
196 tx3g_ptr += 4;
197 mov_text_parse_style_record(&m->d.style, &tx3g_ptr);
198 // FontRecord
199 // FontRecord Size
200 tx3g_ptr += 4;
201 // ftab
202 tx3g_ptr += 4;
204 // In case of broken header, init default font
205 m->d.font = ASS_DEFAULT_FONT;
207 ftab_entries = bytestream_get_be16(&tx3g_ptr);
208 if (!ftab_entries)
209 return 0;
210 remaining -= 3 * ftab_entries;
211 if (remaining < 0)
212 return AVERROR_INVALIDDATA;
213 m->ftab = av_calloc(ftab_entries, sizeof(*m->ftab));
214 if (!m->ftab)
215 return AVERROR(ENOMEM);
216 m->ftab_entries = ftab_entries;
218 for (i = 0; i < m->ftab_entries; i++) {
219 m->ftab[i].font_id = bytestream_get_be16(&tx3g_ptr);
220 if (m->ftab[i].font_id == m->d.style.font_id)
221 j = i;
222 font_length = bytestream_get_byte(&tx3g_ptr);
224 remaining -= font_length;
225 if (remaining < 0) {
226 mov_text_cleanup_ftab(m);
227 return -1;
229 m->ftab[i].font = av_malloc(font_length + 1);
230 if (!m->ftab[i].font) {
231 mov_text_cleanup_ftab(m);
232 return AVERROR(ENOMEM);
234 bytestream_get_buffer(&tx3g_ptr, m->ftab[i].font, font_length);
235 m->ftab[i].font[font_length] = '\0';
237 if (j >= 0)
238 m->d.font = m->ftab[j].font;
239 return 0;
242 static int decode_twrp(const uint8_t *tsmb, MovTextContext *m, uint64_t size)
244 m->box_flags |= TWRP_BOX;
245 m->w.wrap_flag = bytestream_get_byte(&tsmb);
246 return 0;
249 static int decode_hlit(const uint8_t *tsmb, MovTextContext *m, uint64_t size)
251 m->box_flags |= HLIT_BOX;
252 m->h.hlit_start = bytestream_get_be16(&tsmb);
253 m->h.hlit_end = bytestream_get_be16(&tsmb);
254 return 0;
257 static int decode_hclr(const uint8_t *tsmb, MovTextContext *m, uint64_t size)
259 m->box_flags |= HCLR_BOX;
260 bytestream_get_buffer(&tsmb, m->c.hlit_color, 4);
261 return 0;
264 static int styles_equivalent(const StyleBox *a, const StyleBox *b)
266 #define CMP(field) ((a)->field == (b)->field)
267 return CMP(bold) && CMP(italic) && CMP(underline) && CMP(color) &&
268 CMP(alpha) && CMP(fontsize) && CMP(font_id);
269 #undef CMP
272 static int decode_styl(const uint8_t *tsmb, MovTextContext *m, uint64_t size)
274 int i;
275 int style_entries = bytestream_get_be16(&tsmb);
276 StyleBox *tmp;
278 // A single style record is of length 12 bytes.
279 if (2 + style_entries * 12 > size)
280 return -1;
282 tmp = av_realloc_array(m->s, style_entries, sizeof(*m->s));
283 if (!tmp)
284 return AVERROR(ENOMEM);
285 m->s = tmp;
286 m->style_entries = style_entries;
288 m->box_flags |= STYL_BOX;
289 for(i = 0; i < m->style_entries; i++) {
290 StyleBox *style = &m->s[i];
292 style->start = bytestream_get_be16(&tsmb);
293 style->end = bytestream_get_be16(&tsmb);
294 if (style->end < style->start ||
295 (i && style->start < m->s[i - 1].end)) {
296 mov_text_cleanup(m);
297 return AVERROR_INVALIDDATA;
299 if (style->start == style->end) {
300 /* Skip this style as it applies to no character */
301 tsmb += 8;
302 m->style_entries--;
303 i--;
304 continue;
307 mov_text_parse_style_record(style, &tsmb);
308 if (styles_equivalent(style, &m->d.style)) {
309 /* Skip this style as it is equivalent to the default style */
310 m->style_entries--;
311 i--;
312 continue;
313 } else if (i && style->start == style[-1].end &&
314 styles_equivalent(style, &style[-1])) {
315 /* Merge the two adjacent styles */
316 style[-1].end = style->end;
317 m->style_entries--;
318 i--;
319 continue;
322 return 0;
325 static const Box box_types[] = {
326 { MKBETAG('s','t','y','l'), 2, decode_styl },
327 { MKBETAG('h','l','i','t'), 4, decode_hlit },
328 { MKBETAG('h','c','l','r'), 4, decode_hclr },
329 { MKBETAG('t','w','r','p'), 1, decode_twrp }
332 const static size_t box_count = FF_ARRAY_ELEMS(box_types);
334 // Return byte length of the UTF-8 sequence starting at text[0]. 0 on error.
335 static int get_utf8_length_at(const char *text, const char *text_end)
337 const char *start = text;
338 int err = 0;
339 uint32_t c;
340 GET_UTF8(c, text < text_end ? (uint8_t)*text++ : (err = 1, 0), goto error;);
341 if (err)
342 goto error;
343 return text - start;
344 error:
345 return 0;
348 static int text_to_ass(AVBPrint *buf, const char *text, const char *text_end,
349 AVCodecContext *avctx)
351 MovTextContext *m = avctx->priv_data;
352 const StyleBox *const default_style = &m->d.style;
353 int i = 0;
354 int text_pos = 0;
355 int entry = 0;
356 int color = default_style->color;
358 if (text < text_end && m->box_flags & TWRP_BOX) {
359 if (m->w.wrap_flag == 1) {
360 av_bprintf(buf, "{\\q1}"); /* End of line wrap */
361 } else {
362 av_bprintf(buf, "{\\q2}"); /* No wrap */
366 while (text < text_end) {
367 int len;
369 if ((m->box_flags & STYL_BOX) && entry < m->style_entries) {
370 const StyleBox *style = &m->s[entry];
371 if (text_pos == style->end) {
372 av_bprintf(buf, "{\\r}");
373 color = default_style->color;
374 entry++;
375 style++;
377 if (entry < m->style_entries && text_pos == style->start) {
378 if (style->bold ^ default_style->bold)
379 av_bprintf(buf, "{\\b%d}", style->bold);
380 if (style->italic ^ default_style->italic)
381 av_bprintf(buf, "{\\i%d}", style->italic);
382 if (style->underline ^ default_style->underline)
383 av_bprintf(buf, "{\\u%d}", style->underline);
384 if (style->fontsize != default_style->fontsize)
385 av_bprintf(buf, "{\\fs%d}", style->fontsize);
386 if (style->font_id != default_style->font_id)
387 for (i = 0; i < m->ftab_entries; i++) {
388 if (style->font_id == m->ftab[i].font_id)
389 av_bprintf(buf, "{\\fn%s}", m->ftab[i].font);
391 if (default_style->color != style->color) {
392 color = style->color;
393 av_bprintf(buf, "{\\1c&H%X&}", color);
395 if (default_style->alpha != style->alpha)
396 av_bprintf(buf, "{\\1a&H%02X&}", 255 - style->alpha);
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
405 * be implemented.
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]);
410 } else {
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&}", default_style->color);
417 } else {
418 av_bprintf(buf, "{\\1c&H%X&}{\\2c&H%X&}",
419 color, default_style->color);
424 len = get_utf8_length_at(text, text_end);
425 if (len < 1) {
426 av_log(avctx, AV_LOG_ERROR, "invalid UTF-8 byte in subtitle\n");
427 len = 1;
429 switch (*text) {
430 case '\r':
431 break;
432 case '\n':
433 av_bprintf(buf, "\\N");
434 break;
435 default:
436 av_bprint_append_data(buf, text, len);
437 break;
439 text += len;
440 text_pos++;
443 return 0;
446 static int mov_text_init(AVCodecContext *avctx) {
448 * TODO: Handle the default text style.
449 * NB: Most players ignore styles completely, with the result that
450 * it's very common to find files where the default style is broken
451 * and respecting it results in a worse experience than ignoring it.
453 int ret;
454 MovTextContext *m = avctx->priv_data;
455 ret = mov_text_tx3g(avctx, m);
456 if (ret == 0) {
457 const StyleBox *const default_style = &m->d.style;
458 if (!m->frame_width || !m->frame_height) {
459 m->frame_width = ASS_DEFAULT_PLAYRESX;
460 m->frame_height = ASS_DEFAULT_PLAYRESY;
462 return ff_ass_subtitle_header_full(avctx,
463 m->frame_width, m->frame_height,
464 m->d.font, default_style->fontsize,
465 (255U - default_style->alpha) << 24 | default_style->color,
466 (255U - default_style->alpha) << 24 | default_style->color,
467 (255U - m->d.back_alpha) << 24 | m->d.back_color,
468 (255U - m->d.back_alpha) << 24 | m->d.back_color,
469 default_style->bold, default_style->italic, default_style->underline,
470 ASS_DEFAULT_BORDERSTYLE, m->d.alignment);
471 } else
472 return ff_ass_subtitle_header_default(avctx);
475 static int mov_text_decode_frame(AVCodecContext *avctx, AVSubtitle *sub,
476 int *got_sub_ptr, const AVPacket *avpkt)
478 MovTextContext *m = avctx->priv_data;
479 int ret;
480 AVBPrint buf;
481 const char *ptr = avpkt->data, *end;
482 int text_length;
483 size_t i;
485 if (!ptr || avpkt->size < 2)
486 return AVERROR_INVALIDDATA;
489 * A packet of size two with value zero is an empty subtitle
490 * used to mark the end of the previous non-empty subtitle.
491 * We can just drop them here as we have duration information
492 * already. If the value is non-zero, then it's technically a
493 * bad packet.
495 if (avpkt->size == 2)
496 return AV_RB16(ptr) == 0 ? 0 : AVERROR_INVALIDDATA;
499 * The first two bytes of the packet are the length of the text string
500 * In complex cases, there are style descriptors appended to the string
501 * so we can't just assume the packet size is the string size.
503 text_length = AV_RB16(ptr);
504 end = ptr + FFMIN(2 + text_length, avpkt->size);
505 ptr += 2;
507 mov_text_cleanup(m);
509 m->style_entries = 0;
510 m->box_flags = 0;
511 // Note that the spec recommends lines be no longer than 2048 characters.
512 av_bprint_init(&buf, 0, AV_BPRINT_SIZE_UNLIMITED);
513 if (text_length + 2 < avpkt->size) {
514 const uint8_t *tsmb = end;
515 const uint8_t *const tsmb_end = avpkt->data + avpkt->size;
516 // A box is a minimum of 8 bytes.
517 while (tsmb_end - tsmb >= 8) {
518 uint64_t tsmb_size = bytestream_get_be32(&tsmb);
519 uint32_t tsmb_type = bytestream_get_be32(&tsmb);
520 int size_var, ret_tsmb;
522 if (tsmb_size == 1) {
523 if (tsmb_end - tsmb < 8)
524 break;
525 tsmb_size = bytestream_get_be64(&tsmb);
526 size_var = 16;
527 } else
528 size_var = 8;
529 //size_var is equal to 8 or 16 depending on the size of box
531 if (tsmb_size < size_var) {
532 av_log(avctx, AV_LOG_ERROR, "tsmb_size invalid\n");
533 return AVERROR_INVALIDDATA;
535 tsmb_size -= size_var;
537 if (tsmb_end - tsmb < tsmb_size)
538 break;
540 for (i = 0; i < box_count; i++) {
541 if (tsmb_type == box_types[i].type) {
542 if (tsmb_size < box_types[i].base_size)
543 break;
544 ret_tsmb = box_types[i].decode(tsmb, m, tsmb_size);
545 if (ret_tsmb == -1)
546 break;
549 tsmb += tsmb_size;
551 text_to_ass(&buf, ptr, end, avctx);
552 mov_text_cleanup(m);
553 } else
554 text_to_ass(&buf, ptr, end, avctx);
556 ret = ff_ass_add_rect(sub, buf.str, m->readorder++, 0, NULL, NULL);
557 av_bprint_finalize(&buf, NULL);
558 if (ret < 0)
559 return ret;
560 *got_sub_ptr = sub->num_rects > 0;
561 return avpkt->size;
564 static int mov_text_decode_close(AVCodecContext *avctx)
566 MovTextContext *m = avctx->priv_data;
567 mov_text_cleanup_ftab(m);
568 mov_text_cleanup(m);
569 return 0;
572 static void mov_text_flush(AVCodecContext *avctx)
574 MovTextContext *m = avctx->priv_data;
575 if (!(avctx->flags2 & AV_CODEC_FLAG2_RO_FLUSH_NOOP))
576 m->readorder = 0;
579 #define OFFSET(x) offsetof(MovTextContext, x)
580 #define FLAGS AV_OPT_FLAG_DECODING_PARAM | AV_OPT_FLAG_SUBTITLE_PARAM
581 static const AVOption options[] = {
582 { "width", "Frame width, usually video width", OFFSET(frame_width), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
583 { "height", "Frame height, usually video height", OFFSET(frame_height), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
584 { NULL },
587 static const AVClass mov_text_decoder_class = {
588 .class_name = "MOV text decoder",
589 .item_name = av_default_item_name,
590 .option = options,
591 .version = LIBAVUTIL_VERSION_INT,
594 const FFCodec ff_movtext_decoder = {
595 .p.name = "mov_text",
596 CODEC_LONG_NAME("3GPP Timed Text subtitle"),
597 .p.type = AVMEDIA_TYPE_SUBTITLE,
598 .p.id = AV_CODEC_ID_MOV_TEXT,
599 .priv_data_size = sizeof(MovTextContext),
600 .p.priv_class = &mov_text_decoder_class,
601 .init = mov_text_init,
602 FF_CODEC_DECODE_SUB_CB(mov_text_decode_frame),
603 .close = mov_text_decode_close,
604 .flush = mov_text_flush,