Skip to content

Releases

DazedTL 2.75.0

Choose a tag to compare

@dazedanon dazedanon released this 11 Nov 16:04

Added

  • Release: an Assistant-led patch that ships a BepInEx plugin waits until the game's own check shows every line in English.

DazedTL 2.74.0

Choose a tag to compare

@dazedanon dazedanon released this 11 Nov 06:54

Added

  • Settings > Publishing: a Translator name titles each translated game's window with its English title, version and your name.

DazedTL 2.73.1

Choose a tag to compare

@dazedanon dazedanon released this 10 Nov 19:27

Fixed

  • Text QA: one unusable correction no longer stops the whole apply; the others apply, and Text QA lists the one left out.
  • Text QA: a line break written in a correction as ⏎ no longer goes into the game as that symbol.

DazedTL 2.73.0

Choose a tag to compare

@dazedanon dazedanon released this 10 Nov 09:46

Added

  • Text QA: Assistant-led's Options can narrow its review or skip it, and both methods say how much of your assistant's plan a review takes.
  • Text QA: each main character's lines are now also read together across scenes against their glossary voice note, so drifting voices get corrected.
  • Assistant-led: a game translated before requests carried the lines before them can translate again the lines sent without them, keeping your corrections.
  • Tool feedback: your assistant ends each translation with a short note on what DazedTL could do better, to copy or save and post on Discord.
  • Text QA: Assistant-led games on engines other than RPG Maker now get the same review, its corrections saved for your assistant to write in.

Changed

  • Assistant-led: each request now carries the lines just before it in its map or area, so a conversation split across requests reads as one.
  • Set up: voice notes now fix one English choice for each archaic pronoun, ending or tic, so a character sounds the same in every request.

Fixed

  • Assistant-led: Progress no longer lists speaker names as unextracted when their glossary rows carry a note.

DazedTL 2.72.0

Choose a tag to compare

@dazedanon dazedanon released this 09 Nov 20:04

Added

  • Check: Text QA's new Start over sets a QA task aside, even mid-run, so your assistant reviews every line again from scratch.

DazedTL 2.71.1

Choose a tag to compare

@dazedanon dazedanon released this 09 Nov 16:21

Fixed

  • Translation: messages that pause between words or dots no longer show stray spaces inside the pause, even when translated again.

DazedTL 2.71.0

Choose a tag to compare

@dazedanon dazedanon released this 09 Nov 15:07

Changed

  • Compare: while names are hidden, each line shows its translations in a new order, so you can't tell a model by its side.

Fixed

  • Glossary: a glossary saved with Windows line ends or as UTF-16 no longer shows the base glossary among your own entries or stops a run.
  • Glossary: a glossary saved before this fix may list a copy of the base glossary among your own entries, which you can delete.
  • Context: game context, style and quirks and custom skills saved as UTF-16 no longer stop a run, Plugin files or Text QA reading them.
  • Glossary: saving the glossary of an English game no longer adds the Japanese base glossary, and translations no longer send its terms.

DazedTL 2.70.0

Choose a tag to compare

@dazedanon dazedanon released this 09 Nov 08:24

Added

  • Settings: an LM Studio connection translates for free with no key, and PLaMo 3 Translate models, such as plamo-3-translate-8b, work in Guided and Assistant-led.

Changed

  • Project page: Export and Import have their own Export/Import tab, Change method follows the method's name, and clicking the game's path opens its folder.
  • Compare: one workspace replaces the three tabs: each column is a model with its own settings, and unchanged columns are reused, not paid again.

DazedTL 2.69.0

Choose a tag to compare

@dazedanon dazedanon released this 09 Nov 04:42

Added

  • Compare: a new page in the sidebar translates a few lines of your game with up to four models side by side, from any of your connections, or one model with a changed thinking effort, glossary or instruction, sent as a real run sends them, so you can pick a model in equal conditions; Blind hides the names until you've picked the best line of each.

Fixed

  • Translate: translating a message again with Skip already translated text off no longer sends a quest name such as 【柚月の心残り】 on the message's second line as a speaker and leaves it out of the text, or sends the English of a row after a blank first row as that row's text, where Assistant-led, a translation file or an earlier run wrote it, and with First-line names on a translated name before speech goes to the speaker step whatever its speech's English opens with: each text is sent with its own Japanese and speaker, as the translation file lists it; runs estimated before this version send what they were approved with.
  • Text QA: a correction to a name on a row of its own and the speech after it, where Assistant-led or an earlier translation file import wrote them under one Japanese source, no longer wraps the name into the speech: QA reviews and fits each as a line of its own, and Assistant-led now keeps each one's own source when it writes the game. A QA task still in review is prepared again after this update; a finished one applies as before. A Guided translation file of a game Assistant-led wrote lists such a name and its speech as two lines even with First-line names off, so export the file again: an older file's line holding both no longer matches, and its import skips it.

DazedTL 2.68.4

Choose a tag to compare

@dazedanon dazedanon released this 09 Nov 01:48

Fixed

  • Translate: translating a message again with Skip already translated text off no longer sends its speech's English as its text, or the speech as a name, where Assistant-led or a translation file wrote it over its rows, such as a name on a row of its own and its speech, and collecting speaker names no longer takes that speech for a name either: each is sent with its own Japanese, as the translation file lists it; runs estimated before this version send what they were approved with.