Doom9's Forum - View Single Post

ai4spam · 21st September 2005, 04:58

@TiaoMacaleh: If you mean that you often get the prompt to type the entire dialogue, it's probably due to the fact that the lines don't have enough distance between them (usually that happens if you have accentuated characters on the second line). Try lowering the "min interline height" value in the OCR options. Sometimes it's impossible to draw a straight line between the two lines of text no matter what you do, so you'll need to type in those dialogues manually. It is indeed possible to separate text lines with a line that "goes around" accents and the like, but these cases are too rare to justify the programming effort at the moment. And since now you also have the special characters table in that window, it's a lot easier to type the entire line.

@LeMoi:
1) You probably mean the last CodePage, not CharSet. The French language usually triggers the UniCode flag due to the "oe" ligature character. Answer "No" when asked to save as UniCode, then just leave "DEFAULT_CHARSET" and "1252 - ANSI Latin I". You should have no problems with the resulting subtitle. You should never need to choose CodePage 65000 or 65001 (which are UTF-7 and UTF-8, by the way). When you change the CharSet, SubRip automatically sets the corresponding CodePage for you, and normally you should never have to change the CodePage yourself (it's there only for more flexibility). Alternatively, you can save the text as UniCode, then load it in Word and save it as text with the encoding (CodePage) of your choice. Word highlights the characters that cannot be converted for you (it shows them as red question marks).

2) The translator didn't respect the requirement to keep the number of characters when translating, so the best guess text is overlapping the text that says "Meilleure reponse". It should be a bit better in the last beta, because now the best guess has spaces before and after it, like " l ". The font and color are different also, so you should easily see the best guess. Or... edit "Lang\Francais.lng" with NotePad and put "Meil. repo." instead of "Meilleure reponse".

21st September 2005, 04:58	#221 \| Link
ai4spam Programmer Join Date: Sep 2003 Posts: 382	@TiaoMacaleh: If you mean that you often get the prompt to type the entire dialogue, it's probably due to the fact that the lines don't have enough distance between them (usually that happens if you have accentuated characters on the second line). Try lowering the "min interline height" value in the OCR options. Sometimes it's impossible to draw a straight line between the two lines of text no matter what you do, so you'll need to type in those dialogues manually. It is indeed possible to separate text lines with a line that "goes around" accents and the like, but these cases are too rare to justify the programming effort at the moment. And since now you also have the special characters table in that window, it's a lot easier to type the entire line. @LeMoi: 1) You probably mean the last CodePage, not CharSet. The French language usually triggers the UniCode flag due to the "oe" ligature character. Answer "No" when asked to save as UniCode, then just leave "DEFAULT_CHARSET" and "1252 - ANSI Latin I". You should have no problems with the resulting subtitle. You should never need to choose CodePage 65000 or 65001 (which are UTF-7 and UTF-8, by the way). When you change the CharSet, SubRip automatically sets the corresponding CodePage for you, and normally you should never have to change the CodePage yourself (it's there only for more flexibility). Alternatively, you can save the text as UniCode, then load it in Word and save it as text with the encoding (CodePage) of your choice. Word highlights the characters that cannot be converted for you (it shows them as red question marks). 2) The translator didn't respect the requirement to keep the number of characters when translating, so the best guess text is overlapping the text that says "Meilleure reponse". It should be a bit better in the last beta, because now the best guess has spaces before and after it, like " l ". The font and color are different also, so you should easily see the best guess. Or... edit "Lang\Francais.lng" with NotePad and put "Meil. repo." instead of "Meilleure reponse".