#include <stdarg.h>
#include "avcodec.h"
+#include "libavutil/opt.h"
#include "libavutil/avassert.h"
#include "libavutil/avstring.h"
#include "libavutil/intreadwrite.h"
#define HLIT_BOX (1<<1)
#define HCLR_BOX (1<<2)
+#define DEFAULT_STYLE_FONT_ID 0x01
+#define DEFAULT_STYLE_FONTSIZE 0x12
+#define DEFAULT_STYLE_COLOR 0xffffffff
+#define DEFAULT_STYLE_FLAG 0x00
+
#define BGR_TO_RGB(c) (((c) & 0xff) << 16 | ((c) & 0xff00) | (((c) >> 16) & 0xff))
+#define FONTSIZE_SCALE(s,fs) ((fs) * (s)->font_scale_factor + 0.5)
#define av_bprint_append_any(buf, data, size) av_bprint_append_data(buf, ((const char*)data), size)
typedef struct {
uint16_t style_start;
uint16_t style_end;
uint8_t style_flag;
+ uint16_t style_fontID;
+ uint8_t style_fontsize;
+ uint32_t style_color;
} StyleBox;
typedef struct {
} HilightcolorBox;
typedef struct {
+ AVClass *class;
AVCodecContext *avctx;
ASSSplitContext *ass_ctx;
+ ASSStyle *ass_dialog_style;
AVBPrint buffer;
StyleBox **style_attributes;
StyleBox *style_attributes_temp;
HilightcolorBox hclr;
int count;
uint8_t box_flags;
- uint16_t style_fontID;
- uint8_t style_fontsize;
- uint32_t style_color;
+ StyleBox d;
uint16_t text_pos;
uint16_t byte_count;
+ char ** fonts;
+ int font_count;
+ double font_scale_factor;
+ int frame_height;
} MovTextContext;
typedef struct {
av_freep(&s->style_attributes);
}
if (s->style_attributes_temp) {
- s->style_attributes_temp->style_flag = 0;
- s->style_attributes_temp->style_start = 0;
+ *s->style_attributes_temp = s->d;
}
}
tsmb_size = s->count * STYLE_RECORD_SIZE + SIZE_ADD;
tsmb_size = AV_RB32(&tsmb_size);
style_entries = AV_RB16(&s->count);
- s->style_fontID = 0x00 | 0x01<<8;
- s->style_fontsize = 0x12;
- s->style_color = MKTAG(0xFF, 0xFF, 0xFF, 0xFF);
/*The above three attributes are hard coded for now
but will come from ASS style in the future*/
av_bprint_append_any(&s->buffer, &tsmb_size, 4);
av_bprint_append_any(&s->buffer, &tsmb_type, 4);
av_bprint_append_any(&s->buffer, &style_entries, 2);
for (j = 0; j < s->count; j++) {
- uint16_t style_start, style_end;
+ uint16_t style_start, style_end, style_fontID;
+ uint32_t style_color;
style_start = AV_RB16(&s->style_attributes[j]->style_start);
style_end = AV_RB16(&s->style_attributes[j]->style_end);
+ style_color = AV_RB32(&s->style_attributes[j]->style_color);
+ style_fontID = AV_RB16(&s->style_attributes[j]->style_fontID);
+
av_bprint_append_any(&s->buffer, &style_start, 2);
av_bprint_append_any(&s->buffer, &style_end, 2);
- av_bprint_append_any(&s->buffer, &s->style_fontID, 2);
+ av_bprint_append_any(&s->buffer, &style_fontID, 2);
av_bprint_append_any(&s->buffer, &s->style_attributes[j]->style_flag, 1);
- av_bprint_append_any(&s->buffer, &s->style_fontsize, 1);
- av_bprint_append_any(&s->buffer, &s->style_color, 4);
+ av_bprint_append_any(&s->buffer, &s->style_attributes[j]->style_fontsize, 1);
+ av_bprint_append_any(&s->buffer, &style_color, 4);
}
}
mov_text_cleanup(s);
const static size_t box_count = FF_ARRAY_ELEMS(box_types);
-static av_cold int mov_text_encode_init(AVCodecContext *avctx)
+static int mov_text_encode_close(AVCodecContext *avctx)
+{
+ MovTextContext *s = avctx->priv_data;
+ int i;
+
+ ff_ass_split_free(s->ass_ctx);
+ if (s->style_attributes) {
+ for (i = 0; i < s->count; i++) {
+ av_freep(&s->style_attributes[i]);
+ }
+ av_freep(&s->style_attributes);
+ }
+ av_freep(&s->fonts);
+ av_freep(&s->style_attributes_temp);
+ av_bprint_finalize(&s->buffer, NULL);
+ return 0;
+}
+
+static int encode_sample_description(AVCodecContext *avctx)
{
- /*
- * For now, we'll use a fixed default style. When we add styling
- * support, this will be generated from the ASS style.
- */
- static const uint8_t text_sample_entry[] = {
+ ASS * ass;
+ ASSStyle * style;
+ int i, j;
+ uint32_t tsmb_size, tsmb_type, back_color, style_color;
+ uint16_t style_start, style_end, fontID, count;
+ int font_names_total_len = 0;
+ MovTextContext *s = avctx->priv_data;
+
+ static const uint8_t display_and_justification[] = {
0x00, 0x00, 0x00, 0x00, // uint32_t displayFlags
0x01, // int8_t horizontal-justification
0xFF, // int8_t vertical-justification
- 0x00, 0x00, 0x00, 0x00, // uint8_t background-color-rgba[4]
- // BoxRecord {
+ };
+ // 0x00, 0x00, 0x00, 0x00, // uint8_t background-color-rgba[4]
+ static const uint8_t box_record[] = {
+ // BoxRecord {
0x00, 0x00, // int16_t top
0x00, 0x00, // int16_t left
0x00, 0x00, // int16_t bottom
0x00, 0x00, // int16_t right
- // };
- // StyleRecord {
- 0x00, 0x00, // uint16_t startChar
- 0x00, 0x00, // uint16_t endChar
- 0x00, 0x01, // uint16_t font-ID
- 0x00, // uint8_t face-style-flags
- 0x12, // uint8_t font-size
- 0xFF, 0xFF, 0xFF, 0xFF, // uint8_t text-color-rgba[4]
- // };
- // FontTableBox {
- 0x00, 0x00, 0x00, 0x12, // uint32_t size
- 'f', 't', 'a', 'b', // uint8_t name[4]
- 0x00, 0x01, // uint16_t entry-count
- // FontRecord {
- 0x00, 0x01, // uint16_t font-ID
- 0x05, // uint8_t font-name-length
- 'S', 'e', 'r', 'i', 'f',// uint8_t font[font-name-length]
- // };
- // };
+ // };
};
+ // StyleRecord {
+ // 0x00, 0x00, // uint16_t startChar
+ // 0x00, 0x00, // uint16_t endChar
+ // 0x00, 0x01, // uint16_t font-ID
+ // 0x00, // uint8_t face-style-flags
+ // 0x12, // uint8_t font-size
+ // 0xFF, 0xFF, 0xFF, 0xFF, // uint8_t text-color-rgba[4]
+ // };
+ // FontTableBox {
+ // 0x00, 0x00, 0x00, 0x12, // uint32_t size
+ // 'f', 't', 'a', 'b', // uint8_t name[4]
+ // 0x00, 0x01, // uint16_t entry-count
+ // FontRecord {
+ // 0x00, 0x01, // uint16_t font-ID
+ // 0x05, // uint8_t font-name-length
+ // 'S', 'e', 'r', 'i', 'f',// uint8_t font[font-name-length]
+ // };
+ // };
+
+ // Populate sample description from ASS header
+ ass = (ASS*)s->ass_ctx;
+ // Compute font scaling factor based on (optionally) provided
+ // output video height and ASS script play_res_y
+ if (s->frame_height && ass->script_info.play_res_y)
+ s->font_scale_factor = (double)s->frame_height / ass->script_info.play_res_y;
+ else
+ s->font_scale_factor = 1;
+
+ style = ff_ass_style_get(s->ass_ctx, "Default");
+ if (!style && ass->styles_count) {
+ style = &ass->styles[0];
+ }
+ s->d.style_fontID = DEFAULT_STYLE_FONT_ID;
+ s->d.style_fontsize = DEFAULT_STYLE_FONTSIZE;
+ s->d.style_color = DEFAULT_STYLE_COLOR;
+ s->d.style_flag = DEFAULT_STYLE_FLAG;
+ if (style) {
+ s->d.style_fontsize = FONTSIZE_SCALE(s, style->font_size);
+ s->d.style_color = BGR_TO_RGB(style->primary_color & 0xffffff) << 8 |
+ 255 - ((uint32_t)style->primary_color >> 24);
+ s->d.style_flag = (!!style->bold * STYLE_FLAG_BOLD) |
+ (!!style->italic * STYLE_FLAG_ITALIC) |
+ (!!style->underline * STYLE_FLAG_UNDERLINE);
+ back_color = (BGR_TO_RGB(style->back_color & 0xffffff) << 8) |
+ (255 - ((uint32_t)style->back_color >> 24));
+ }
- MovTextContext *s = avctx->priv_data;
- s->avctx = avctx;
+ av_bprint_append_any(&s->buffer, display_and_justification,
+ sizeof(display_and_justification));
+ back_color = AV_RB32(&back_color);
+ av_bprint_append_any(&s->buffer, &back_color, 4);
+ // BoxRecord {
+ av_bprint_append_any(&s->buffer, box_record, sizeof(box_record));
+ // };
+ // StyleRecord {
+ style_start = AV_RB16(&s->d.style_start);
+ style_end = AV_RB16(&s->d.style_end);
+ fontID = AV_RB16(&s->d.style_fontID);
+ style_color = AV_RB32(&s->d.style_color);
+ av_bprint_append_any(&s->buffer, &style_start, 2);
+ av_bprint_append_any(&s->buffer, &style_end, 2);
+ av_bprint_append_any(&s->buffer, &fontID, 2);
+ av_bprint_append_any(&s->buffer, &s->d.style_flag, 1);
+ av_bprint_append_any(&s->buffer, &s->d.style_fontsize, 1);
+ av_bprint_append_any(&s->buffer, &style_color, 4);
+ // };
+
+ // Build font table
+ // We can't build a complete font table since that would require
+ // scanning all dialogs first. But we can at least fill in what
+ // is avaiable in the ASS header
+ if (style && ass->styles_count) {
+ // Find unique font names
+ av_dynarray_add(&s->fonts, &s->font_count, style->font_name);
+ font_names_total_len += strlen(style->font_name);
+ for (i = 0; i < ass->styles_count; i++) {
+ int found = 0;
+ for (j = 0; j < s->font_count; j++) {
+ if (!strcmp(s->fonts[j], ass->styles[i].font_name)) {
+ found = 1;
+ break;
+ }
+ }
+ if (!found) {
+ av_dynarray_add(&s->fonts, &s->font_count,
+ ass->styles[i].font_name);
+ font_names_total_len += strlen(ass->styles[i].font_name);
+ }
+ }
+ } else
+ av_dynarray_add(&s->fonts, &s->font_count, (char*)"Serif");
+
+ // FontTableBox {
+ tsmb_size = SIZE_ADD + 3 * s->font_count + font_names_total_len;
+ tsmb_size = AV_RB32(&tsmb_size);
+ tsmb_type = MKTAG('f','t','a','b');
+ count = AV_RB16(&s->font_count);
+ av_bprint_append_any(&s->buffer, &tsmb_size, 4);
+ av_bprint_append_any(&s->buffer, &tsmb_type, 4);
+ av_bprint_append_any(&s->buffer, &count, 2);
+ // FontRecord {
+ for (i = 0; i < s->font_count; i++) {
+ int len;
+ fontID = i + 1;
+ fontID = AV_RB16(&fontID);
+ av_bprint_append_any(&s->buffer, &fontID, 2);
+ len = strlen(s->fonts[i]);
+ av_bprint_append_any(&s->buffer, &len, 1);
+ av_bprint_append_any(&s->buffer, s->fonts[i], len);
+ }
+ // };
+ // };
- s->style_attributes_temp = av_mallocz(sizeof(*s->style_attributes_temp));
- if (!s->style_attributes_temp) {
+ if (!av_bprint_is_complete(&s->buffer)) {
return AVERROR(ENOMEM);
}
- avctx->extradata_size = sizeof text_sample_entry;
+ avctx->extradata_size = s->buffer.len;
avctx->extradata = av_mallocz(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
- if (!avctx->extradata)
+ if (!avctx->extradata) {
return AVERROR(ENOMEM);
+ }
+
+ memcpy(avctx->extradata, s->buffer.str, avctx->extradata_size);
+ av_bprint_clear(&s->buffer);
+
+ return 0;
+}
+
+static av_cold int mov_text_encode_init(AVCodecContext *avctx)
+{
+ int ret;
+ MovTextContext *s = avctx->priv_data;
+ s->avctx = avctx;
av_bprint_init(&s->buffer, 0, AV_BPRINT_SIZE_UNLIMITED);
- memcpy(avctx->extradata, text_sample_entry, avctx->extradata_size);
+ s->style_attributes_temp = av_mallocz(sizeof(*s->style_attributes_temp));
+ if (!s->style_attributes_temp) {
+ ret = AVERROR(ENOMEM);
+ goto fail;
+ }
s->ass_ctx = ff_ass_split(avctx->subtitle_header);
- return s->ass_ctx ? 0 : AVERROR_INVALIDDATA;
+ if (!s->ass_ctx) {
+ ret = AVERROR_INVALIDDATA;
+ goto fail;
+ }
+ ret = encode_sample_description(avctx);
+ if (ret < 0)
+ goto fail;
+
+ return 0;
+
+fail:
+ mov_text_encode_close(avctx);
+ return ret;
}
// Start a new style box if needed
if (s->style_attributes_temp->style_start == s->text_pos)
// Still at same text pos, use same entry
return 1;
- if (s->style_attributes_temp->style_flag) {
+ if (s->style_attributes_temp->style_flag != s->d.style_flag ||
+ s->style_attributes_temp->style_color != s->d.style_color ||
+ s->style_attributes_temp->style_fontID != s->d.style_fontID ||
+ s->style_attributes_temp->style_fontsize != s->d.style_fontsize) {
// last style != defaults, end the style entry and start a new one
s->box_flags |= STYL_BOX;
s->style_attributes_temp->style_end = s->text_pos;
return 0;
}
- s->style_attributes_temp->style_flag = s->style_attributes[s->count - 1]->style_flag;
+ *s->style_attributes_temp = s->d;
s->style_attributes_temp->style_start = s->text_pos;
} else { // style entry matches defaults, drop entry
- s->style_attributes_temp->style_flag = 0;
+ *s->style_attributes_temp = s->d;
s->style_attributes_temp->style_start = s->text_pos;
}
return 1;
return style_flag;
}
+static void mov_text_style_set(MovTextContext *s, uint8_t style_flags)
+{
+ if (!s->style_attributes_temp ||
+ !((s->style_attributes_temp->style_flag & style_flags) ^ style_flags)) {
+ // setting flags that that are already set
+ return;
+ }
+ if (mov_text_style_start(s))
+ s->style_attributes_temp->style_flag |= style_flags;
+}
+
static void mov_text_style_cb(void *priv, const char style, int close)
{
MovTextContext *s = priv;
}
}
+static void mov_text_color_set(MovTextContext *s, uint32_t color)
+{
+ if (!s->style_attributes_temp ||
+ (s->style_attributes_temp->style_color & 0xffffff00) == color) {
+ // color hasn't changed
+ return;
+ }
+ if (mov_text_style_start(s))
+ s->style_attributes_temp->style_color = (color & 0xffffff00) |
+ (s->style_attributes_temp->style_color & 0xff);
+}
+
static void mov_text_color_cb(void *priv, unsigned int color, unsigned int color_id)
{
MovTextContext *s = priv;
color = BGR_TO_RGB(color) << 8;
- if (color_id == 2) { //secondary color changes
- if (s->box_flags & HLIT_BOX) { //close tag
+ if (color_id == 1) { //primary color changes
+ mov_text_color_set(s, color);
+ } else if (color_id == 2) { //secondary color changes
+ if (!(s->box_flags & HCLR_BOX))
+ // Highlight alpha not set yet, use current primary alpha
+ s->hclr.color = s->style_attributes_temp->style_color;
+ if (!(s->box_flags & HLIT_BOX) || s->hlit.start == s->text_pos) {
+ s->box_flags |= HCLR_BOX;
+ s->box_flags |= HLIT_BOX;
+ s->hlit.start = s->text_pos;
+ s->hclr.color = color | (s->hclr.color & 0xFF);
+ }
+ else //close tag
s->hlit.end = s->text_pos;
- } else {
+ /* If there are more than one secondary color changes in ASS,
+ take start of first section and end of last section. Movtext
+ allows only one highlight box per sample.
+ */
+ }
+ // Movtext does not support changes to other color_id (outline, background)
+}
+
+static void mov_text_alpha_set(MovTextContext *s, uint8_t alpha)
+{
+ if (!s->style_attributes_temp ||
+ (s->style_attributes_temp->style_color & 0xff) == alpha) {
+ // color hasn't changed
+ return;
+ }
+ if (mov_text_style_start(s))
+ s->style_attributes_temp->style_color =
+ (s->style_attributes_temp->style_color & 0xffffff00) | alpha;
+}
+
+static void mov_text_alpha_cb(void *priv, int alpha, int alpha_id)
+{
+ MovTextContext *s = priv;
+
+ alpha = 255 - alpha;
+ if (alpha_id == 1) // primary alpha changes
+ mov_text_alpha_set(s, alpha);
+ else if (alpha_id == 2) { //secondary alpha changes
+ if (!(s->box_flags & HCLR_BOX))
+ // Highlight color not set yet, use current primary color
+ s->hclr.color = s->style_attributes_temp->style_color;
+ if (!(s->box_flags & HLIT_BOX) || s->hlit.start == s->text_pos) {
s->box_flags |= HCLR_BOX;
s->box_flags |= HLIT_BOX;
s->hlit.start = s->text_pos;
- s->hclr.color = color | 0xFF; //set alpha value to FF
+ s->hclr.color = (s->hclr.color & 0xffffff00) | alpha;
}
+ else //close tag
+ s->hlit.end = s->text_pos;
}
- /* If there are more than one secondary color changes in ASS, take start of
- first section and end of last section. Movtext allows only one
- highlight box per sample.
- */
+ // Movtext does not support changes to other alpha_id (outline, background)
+}
+
+static uint16_t find_font_id(MovTextContext * s, const char * name)
+{
+ int i;
+ for (i = 0; i < s->font_count; i++) {
+ if (!strcmp(name, s->fonts[i]))
+ return i + 1;
+ }
+ return 1;
+}
+
+static void mov_text_font_name_set(MovTextContext *s, const char *name)
+{
+ int fontID = find_font_id(s, name);
+ if (!s->style_attributes_temp ||
+ s->style_attributes_temp->style_fontID == fontID) {
+ // color hasn't changed
+ return;
+ }
+ if (mov_text_style_start(s))
+ s->style_attributes_temp->style_fontID = fontID;
+}
+
+static void mov_text_font_name_cb(void *priv, const char *name)
+{
+ mov_text_font_name_set((MovTextContext*)priv, name);
+}
+
+static void mov_text_font_size_set(MovTextContext *s, int size)
+{
+ size = FONTSIZE_SCALE(s, size);
+ if (!s->style_attributes_temp ||
+ s->style_attributes_temp->style_fontsize == size) {
+ // color hasn't changed
+ return;
+ }
+ if (mov_text_style_start(s))
+ s->style_attributes_temp->style_fontsize = size;
+}
+
+static void mov_text_font_size_cb(void *priv, int size)
+{
+ mov_text_font_size_set((MovTextContext*)priv, size);
+}
+
+static void mov_text_end_cb(void *priv)
+{
+ // End of text, close any open style record
+ mov_text_style_start((MovTextContext*)priv);
+}
+
+static void mov_text_ass_style_set(MovTextContext *s, ASSStyle *style)
+{
+ uint8_t style_flags, alpha;
+ uint32_t color;
+
+ if (style) {
+ style_flags = (!!style->bold * STYLE_FLAG_BOLD) |
+ (!!style->italic * STYLE_FLAG_ITALIC) |
+ (!!style->underline * STYLE_FLAG_UNDERLINE);
+ mov_text_style_set(s, style_flags);
+ color = BGR_TO_RGB(style->primary_color & 0xffffff) << 8;
+ mov_text_color_set(s, color);
+ alpha = 255 - ((uint32_t)style->primary_color >> 24);
+ mov_text_alpha_set(s, alpha);
+ mov_text_font_size_set(s, style->font_size);
+ mov_text_font_name_set(s, style->font_name);
+ } else {
+ // End current style record, go back to defaults
+ mov_text_style_start(s);
+ }
+}
+
+static void mov_text_dialog(MovTextContext *s, ASSDialog *dialog)
+{
+ ASSStyle * style = ff_ass_style_get(s->ass_ctx, dialog->style);
+
+ s->ass_dialog_style = style;
+ mov_text_ass_style_set(s, style);
+}
+
+static void mov_text_cancel_overrides_cb(void *priv, const char * style_name)
+{
+ MovTextContext *s = priv;
+ ASSStyle * style;
+
+ if (!style_name || !*style_name)
+ style = s->ass_dialog_style;
+ else
+ style= ff_ass_style_get(s->ass_ctx, style_name);
+
+ mov_text_ass_style_set(s, style);
}
static uint16_t utf8_strlen(const char *text, int len)
}
static const ASSCodesCallbacks mov_text_callbacks = {
- .text = mov_text_text_cb,
- .new_line = mov_text_new_line_cb,
- .style = mov_text_style_cb,
- .color = mov_text_color_cb,
+ .text = mov_text_text_cb,
+ .new_line = mov_text_new_line_cb,
+ .style = mov_text_style_cb,
+ .color = mov_text_color_cb,
+ .alpha = mov_text_alpha_cb,
+ .font_name = mov_text_font_name_cb,
+ .font_size = mov_text_font_size_cb,
+ .cancel_overrides = mov_text_cancel_overrides_cb,
+ .end = mov_text_end_cb,
};
static int mov_text_encode_frame(AVCodecContext *avctx, unsigned char *buf,
if (sub->rects[i]->type != SUBTITLE_ASS) {
av_log(avctx, AV_LOG_ERROR, "Only SUBTITLE_ASS type supported.\n");
- return AVERROR(ENOSYS);
+ return AVERROR(EINVAL);
}
#if FF_API_ASS_TIMING
int num;
dialog = ff_ass_split_dialog(s->ass_ctx, ass, 0, &num);
for (; dialog && num--; dialog++) {
+ mov_text_dialog(s, dialog);
ff_ass_split_override_codes(&mov_text_callbacks, s, dialog->text);
}
} else {
dialog = ff_ass_split_dialog2(s->ass_ctx, ass);
if (!dialog)
return AVERROR(ENOMEM);
+ mov_text_dialog(s, dialog);
ff_ass_split_override_codes(&mov_text_callbacks, s, dialog->text);
ff_ass_free_dialog(&dialog);
#if FF_API_ASS_TIMING
if (s->buffer.len > bufsize - 3) {
av_log(avctx, AV_LOG_ERROR, "Buffer too small for ASS event.\n");
- length = AVERROR(EINVAL);
+ length = AVERROR_BUFFER_TOO_SMALL;
goto exit;
}
return length;
}
-static int mov_text_encode_close(AVCodecContext *avctx)
-{
- MovTextContext *s = avctx->priv_data;
- ff_ass_split_free(s->ass_ctx);
- av_bprint_finalize(&s->buffer, NULL);
- return 0;
-}
+#define OFFSET(x) offsetof(MovTextContext, x)
+#define FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_SUBTITLE_PARAM
+static const AVOption options[] = {
+ { "height", "Frame height, usually video height", OFFSET(frame_height), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
+ { NULL },
+};
+
+static const AVClass mov_text_encoder_class = {
+ .class_name = "MOV text enoder",
+ .item_name = av_default_item_name,
+ .option = options,
+ .version = LIBAVUTIL_VERSION_INT,
+};
AVCodec ff_movtext_encoder = {
.name = "mov_text",
.type = AVMEDIA_TYPE_SUBTITLE,
.id = AV_CODEC_ID_MOV_TEXT,
.priv_data_size = sizeof(MovTextContext),
+ .priv_class = &mov_text_encoder_class,
.init = mov_text_encode_init,
.encode_sub = mov_text_encode_frame,
.close = mov_text_encode_close,