Grading Scales for Fansub Reviews
Sample perceived difficulty and scores for releases I’ve watched multiple times or recently:
[Nozomi Entertainment] Shoujo Kakumei Utena Ep1 = Difficulty 7 = A+, LS-8
[Various] Studio Ghibli theatrical / DVD subs = Difficulty 5~6 = A+ LS-5~7
[Eclipse] Claymore = Difficulty 5 = A, LS-5
[gg] Mawaru Penguindrum = Difficulty 6 (w/Puns and Quirks 8 ) = A- (B+), LS-7
[Koharubi] Bakemonogatari = Difficulty 8 (w/Wordplay 10) = A- (B-), LS-5
[Crunchyroll] Steins;Gate = Difficulty 7 = B+, LS-4
[UTW] Kamisama no Memochou = Difficulty 8 = B+, LS-5 (Note: I try harder and spend more time on KamiMemo than Penguindrum, but it has lower grade due to difficulty and sometimes fatigue from blogging lol)
[AniYoshi] Ga-rei: Zero = Difficulty 6 = B, LS-4
[SS-Eclipse] Shakugan no Shana III = Difficulty 8 = B-, LS-4
[Fansub Average] Of all I’ve seen so far = Difficulty 6 = C, LS-4 (Note: Today’s average for popular releases are closer to B- since most of the shitty groups died)
[gg] Sacred Seven = Difficulty 5 (w/Quirkiness 8 ) = C- (B+), LS-8 (Note: Higher grade for higher difficulty is possible, as quirkiness is allowed to have higher LS tolerance and neutralize errors)
[Doki] AnoHana Ep1, 7 = Difficulty 5 = D+, LS-3
[SHiN-gx] Hanasaku Iroha Ep1 = Difficulty 5 = D-, LS-8
[gg] Hidan no Aria = Difficulty 4 = F+, LS-10
[Nyoro~n] This shit = Difficulty 6 = F, LS-4
[GoogleTranslate] Everything = Difficulty any = F-, LS-1
*(brackets): Quirk / Wordplay included
*Bold: Simulcast / Speed subs (Release within 24 hours of first broadcast)
A+: Professional level translations that are nearly flawless. Over-localized without losing details or nuance. No fansub release has ever achieved this standard as far as I have seen. LS>=6
A and A-: Reserved only for the best of the best. Releases that need very few corrections and has 99%+ accuracy (No more than two major mistake per episode, 5 major/minor combined). LS >= 5.
B- to B+: “Good”. 97.5%+ accuracy (about less than 10 critical and minor errors combined per episode) with decent nuance and flow.
C- to C: “Watchable”. 95%+ accuracy (about less than major 20 critical errors per episode) with passable nuance and flow.
D- to D+: “Poor”. 90%+ accuracy, poor phrasing. Avoid these if possible.
F- to F+: “Unusable”. Trollsubs or sheer incompetence.
Note: In reality, “97.5%” accuracy should be “F+” grade, but there is no way to fix this without making most group’s releases “F”-ranged.
Accuracy rate and Liberal Scale (LS) listed are simply the minimum requirement. This ranking system is heavily dependent on accuracy (or lack of translation mistakes, lines that doesn’t represent the meaning of what’s actually being said) because I firmly believe that’s the most important thing in translation.
Flow (how smoothly it reads, how well lines are shortened without loss of meanings) and nuance (the amount of detail, tones, and quirks retained) come after.
The translations must meet certain standards in all four fields be eligible for respective grades.
9 to 10: Extremely hard, if not impossible to translate. Heavy wordplay and jargon usage, or archaic discourse. These will always show up in brackets and the translator may choose to ignore them.
7 to 8: Very difficult series to translate. Complex Japanese dialogue. Quirky speeches. Wordplays etc.
5 to 6: Average late-night series. Requires fluency in both source and target to properly translate.
3 to 4: Easy to translate series, often daytime series or slice of life.
1 to 2: Children’s shows.
Liberal Scale (LS)
The perceived amount of localization present. Liberal translation is NOT “shit”, they are better since the whole point of translation is localizing a foreign language into the native language, but it’s harder (and sometimes impossible) to execute, often resulting in shit translations when done by inexperienced translators.
LS-1 is as literal as it gets on the scale, or “dictionary/Google translation.” LS-10 is as liberal as it gets on the scale without being completely wrong, or “almost made up.” The higher up in the scale, the more difficult to translate accurately.
LS1~2 and LS9~10 are far enough to the original meanings that they are usually considered to be a “minor mistake” even if when they’re arguably accurate.
>>> LS-2 is LS-1 arranged properly in English grammar. I skipped LS-4, 6, and 8 because it’s difficult to determine the exact scale and it wouldn’t make a very good example.
MPD Ep2 「手段を選んでいる場合か」
1: “Measures of choosing time be?”
2: “Is this the time to be choosing measures?”
3: “This isn’t the time to be picky about our methods.”
5: “We can’t be picky about our methods.”
7: “The end justifies the means.”
9: “Himari is dead if we played it safe.”
10: “Fuck Ringo, we only care about Himari.”
>>> Japanese is a low-context language. The subject is not needed because the language assumes that the listener/reader would understand the context already, but they must be added in English to make sense.
This line, in context, means “We don’t have the choice of doing things morally right now, and we have to do what it takes to achieve our goal.”
“The end justifies the means.” is the ideal liberal translation here because it’s a common English phrase that’s used exactly in this situation. However, if “Whatever it takes to get the job done” was chosen, there would be less emphasis on the moral issue they’re facing right now, not to mention other things like nuance and text length. The difficulty of selecting the right idiom for each situation is why viewers often perceive liberal translations to be “shit”, because the translator must have chosen the wrong one (or an appropriate one didn’t exist to begin with). It’s impossible for non-fluent source language speakers to notice liberal translations when they’re properly done, as the top-tier translators who release “A+” quality are consistently capable of.
1: “Who is yours prey something.”
2: “Who will become a prey of you?”
3: “Like I’ll become your prey?”
5: “I’m not your prey.”
7: “How dare you call me prey?”
9: “Who you calling prey, ho?”
10: ” Prey my ass, bitch.”
>>> The line literally translates to “Who will become a prey of you?”, but it must be translated to make sense in English. What the character meant was “You just called me your prey, but that’s not what I am” to someone he despises. The challenge is fitting this into appropriate length for ~2 seconds (ideally under 12 characters per second, absolutely no more than 18 for professional standards).
As mentioned, the extremes of both ends are minor errors as you can see. LS-1 to 3 are awkard or don’t make sense. LS-5 and LS-7 are clearly the best choices, however some meanings are lost. LS-9 and 10 convey better nuance of the line, but interrupts the flow of conversation and character personality.
You may have noticed that I have chosen LS-9 for the gg release knowing it was a mistake because well, I have that freedom to add lulz. It also hints his disdain while only barely missing the ideal text length. Adding “to someone like you” would’ve instantly been too long for any of those translations. LS-7 would’ve been chosen if I was translating professionally.
LS-6~7 are the ideal levels in this scale for professional level translations. Somewhat localized, but not completely ruin the nuance or lose details. Godly translators can achieve good accuracy with higher level of localization.
The LS ratings I will give to individual releases will be perceived average of lines that can be localized or need to be localized. Higher LS score is better only when accuracy/overall grades are nearly the same.