When choosing a Mac text editor for Japanese input, do not promote the candidate that wins one short speed test. Keep the editor that remains stable through composition, proper-name conversion, Japanese dialogue punctuation, reconversion, saving, and reopening. Use the same IME and dictionary, then reverse candidate order on day two so learning and fatigue are less likely to decide the result.
Fix the IME before comparing editors
Japanese conversion accuracy is not controlled by the editor alone. Record the macOS input source, IME, user-dictionary state, live-conversion setting, and test time before opening candidate A or B.
Do not teach ten test names to one candidate but not the other. Either preserve the same dictionary state or choose unlearned names. This article does not rank IMEs. It compares how two editors behave when they receive the same Japanese input conditions.
Build an 800-character task with visible answers
Prepare an 800-character passage containing dialogue, a long conversion phrase, ten proper names, Japanese brackets, and ellipses. Keep a separate answer key for the names and punctuation.
Ordinary prose can reduce the result to a vague feeling. The controlled task exposes specific failures: losing the active composition, confirming a wrong name, failing to reconvert it, or moving the insertion point after repeated punctuation. Use copies, not the authoritative manuscript, and record the baseline character count and line breaks.
Separate speed from correction cost
Type for 15 minutes in each candidate. Record completed characters, wrong conversions, reconversions, punctuation failures, and lost insertion positions in separate fields.
An illustrative row might show A with 800 characters, three wrong conversions, two reconversions, and one punctuation failure; B with 760 characters, one wrong conversion, no reconversions, and no punctuation failures. These are examples of the sheet, not benchmark results. A is not automatically better because its first number is larger.
Do not change conditions when a failure appears. Name the failure and continue the same task so the two candidates remain comparable.
Saving and reopening are part of input stability
After the timed run, save and quit. On the next day, reopen the file and compare its final text, insertion point, line breaks, proper names, and punctuation with the answer key.
A conversion mistake recorded at confirmation is different from a saved-text mismatch found after reopening. Keep those failures separate. The first belongs to the input session; the second is a persistence problem.
Tests across UTF-8, BOM, and Shift_JIS belong to the encoding articles. This comparison keeps one UTF-8 baseline and follows Japanese input through the same-format restart.
Reverse the order on day two
If day one runs A then B, run B then A on day two. Use a second passage with the same difficulty and repeat the timed input, name checks, punctuation checks, save, and reopen.
Read results by day as well as in total. A candidate may be fast only when tested first, or show mistakes only after fatigue. Exclude a candidate that repeatedly loses the active composition or fails the reopened-content check even if its total character count is higher.
Keep the Rune Studio claim narrow
Rune Studio has documented support for Japanese long-form editing, font-size and line-spacing adjustment, search, and saving in one editing environment. That makes it eligible for this controlled task.
The evidence does not establish that Rune Studio is fastest with a particular IME, has the fewest conversion errors, or wins the two-day test. IME accuracy is not a Rune Studio guarantee. The recommendation must come from the completed sheet under the reader’s actual IME conditions.
Read blocking failures before speed
Check reopened mismatches, lost composition position, and missed proper-name errors first. Then compare wrong conversions and reconversions. Read raw typing volume last.
This ordering prevents a slightly faster but correction-heavy candidate from being called recommended. STUDIO-333 chooses stage-specific display values. This article chooses a complete Japanese-input route. Promote only an editor that survives the same IME, equivalent 800-character tasks, reversed test order, and next-day reopening.
