Cygames’ CEDEC presentation on the English localization of Umamusume: Pretty Derby is unusually useful because it says the quiet part out loud. The deck does not merely show a few lines that were loosened for natural English. It lays out a method: separate “translation” from “localization,” treat the former as the transfer of information, then use the latter to recreate a character’s “essence” through target-language dialects, internet communities, period slang, memes, and newly invented verbal quirks. Near the end, the team summarizes the result by saying it “created unique speech styles and quirks for each character.”

That choice of verb matters. A translator is supposed to discover the character in the Japanese and render that character in English. Once the process becomes the creation of a new English persona based on what a localization team thinks will feel recognizable or entertaining to an Anglophone audience, the work has crossed a line. It may still be clever copywriting. It may still make people laugh. It may even produce individual lines that some players prefer. None of that turns it back into a faithful translation. The American Translators Association explicitly treats additions, omissions, changes of tone, and “creative” renderings that alter meaning as meaning-transfer or faithfulness errors, while separately requiring idiomatic target-language writing.

The distinction is not between stiff literalism and lively localization. That is the false choice on which the presentation rests. Natural English can preserve meaning, register, hesitation, roughness, politeness, humor, dialectal flavor, and characterization without replacing Japanese social and cultural signals with an American costume.

The examples in this deck repeatedly demonstrate the opposite approach. Where the Japanese is difficult, the English is not asked to carry the difficulty. The Japanese feature is replaced with something more immediately legible to the localization team: Kansai speech becomes a fabricated written accent, rural Iwate becomes mid-century American idiom, otaku enthusiasm becomes English internet fandom language, gyaru slang becomes Gen-Z meme speech, and a Japanese nickname becomes whatever “appropriately zoomer-y” alias gets the biggest laugh in Slack.

The presentation calls this recognizability and authenticity. The more accurate description is cultural substitution.

The distinction matters because the presentation itself sets a much stronger standard than simple entertainment. The CEDEC session frames its method around “Recognizability” and “Authenticity,” while Allison Cottrell’s official speaker statement says that, when localizing Japanese material, she aims for natural, consistent translation while faithfully preserving the original work’s appeal and cultural background. Those goals are compatible only if the English voice remains answerable to the Japanese one. If “authenticity” instead means authenticity to a target-culture persona invented by the localization team, the word has changed meaning halfway through the argument.

The Deck Begins by Defining Translation Out of Existence

The conceptual problem appears before Umamusume itself enters the discussion. Slides 4 through 6 present a miniature lesson titled “What Is Localization?” The first Japanese sample reads:

貴方は優秀です。
これは贈り物です。

The slide renders it as:

You were good.
Here’s a gift.

Then the next slide reveals a Santa-like figure and changes the Japanese to:

ほっほっほ、
良い子にしておったな。
わしからのプレゼントじゃ！

The English becomes:

Hohoho!
Someone’s been good this year!
Here’s a present for you!

The deck then concludes that translation conveys information while localization reproduces emotion, dialect, and speech quirks, thereby preserving the “essence” of the character.

The demonstration does not establish that. It quietly swaps the source sentence.

貴方は優秀です and 良い子にしておったな do not say the same thing in two different voices. 優秀 describes excellence or outstanding ability; 良い子 is the language of a “good child,” exactly the moral framing associated with Santa rewarding good behavior. The first uses neutral です; the second gives the speaker an old-man/Santa voice through わし, じゃ, the laughter, and the phrasing. The first says これは贈り物です, “this is a gift.” The second says わしからのプレゼントじゃ, “a present from me,” which adds a relation between giver and recipient. The English also adds “this year,” which is not in the Japanese shown on the slide.

So there are only two possible readings of the example, and neither supports the presentation’s thesis. If slide 5 is supposed to be a localization of slide 4, then the localization has rewritten the Japanese source before congratulating itself for preserving characterization. If slide 5 is a different Japanese source, then its stronger characterization comes from the Japanese writer. A competent translation naturally carries that extra characterization over. The comparison proves that characterful source text produces characterful target text. It does not prove that “localization” must invent character on top of translation.

Even the supposedly plain line on slide 4 is already less precise than it could be. “You were good” shifts both the tense and the semantic field of 優秀, pulling the sentence toward Santa’s moral “goodness” before Santa has even appeared.

Faithful natural English: “You did very well. This is a gift for you.”

Or, if the context genuinely calls for a statement of ability rather than performance:

Faithful natural English: “You’re excellent. This is a gift.”

The characterful Santa line can likewise remain lively without additions:

Faithful natural English: “Ho ho ho! You’ve been a good child, haven’t you? Here’s a present from me!”

Nothing about that requires a theory in which translation handles facts and localization handles personality. Personality is part of what is being translated. Professional translation standards make the same distinction in less theatrical terms: preserving meaning and intent is a source-target requirement, while idiomatic wording, register, and style are also evaluated. A natural target sentence is not exempt from fidelity because it sounds good.

That false split poisons the rest of the deck. Emotion, register, dialect, verbal habits, implication, and social position are not decorative material sitting outside “information.” They are part of meaning. A character who says the equivalent of “What are you doing?” in clipped, deferential speech is not saying the same thing as a character who spits out a rough dialectal challenge. A translation that preserves the dictionary proposition while replacing the social voice has not preserved the source in any useful artistic sense. Translation guidance from the Chartered Institute of Linguists likewise treats register, vocabulary, idiom, spirit, and intention as part of a successful rendering rather than extras to be invented after “information” has been transferred.

Slide 9 makes the contradiction harder to miss. One of the fan comments selected by Cygames says that the poster hopes the long-awaited global release will not have a translation so bad that it breaks immersion in the story. That anxiety is presented as evidence of high expectations. Fair enough. But a fan asking not to be pulled out of a Japanese story by bad translation is not asking a localization team to replace Japanese characterization with whatever English-language subculture seems analogous. The presentation takes a demand for a trustworthy bridge to the source and treats it as a mandate for more intervention.

The industry’s own quality frameworks do not require that leap. MQM places mistranslation under accuracy and separately recognizes omissions and additions; ATA likewise distinguishes source-meaning transfer from target-language writing quality. In other words, fluent prose and faithfulness are simultaneous obligations, not competing philosophies from which a localization team must choose one.

That distinction matters because the deck repeatedly turns the character dossier from a constraint into a content generator.

A useful character dossier tells a translator how to resolve ambiguity. If the same character repeatedly uses a certain sentence ending, degree of politeness, joke pattern or verbal tic, English should reflect that pattern where possible. Background can guide lexical choices when the source gives the translator genuine latitude.

A character dossier should not become permission to generate new content whenever the Japanese line looks too ordinary.

A Fake Accent Is Not Kansai Dialect

Slides 11 through 13 identify a real problem. Kansai dialect is often represented in English localization through some variety of American Southern speech. Umamusume makes that shortcut especially awkward because its cast also includes characters whose imagery is explicitly connected with the United States and, in Taiki Shuttle’s case, a cowgirl persona. Giving Tamamo Cross and an American-coded character the same “Howdy, y’all!” voice would collapse distinctions that exist in the setting.

So far, so good.

Dialects do not have clean one-to-one equivalents across languages. A Japanese regional variety carries geographical, social, historical, comic, and interpersonal associations that an English regional variety cannot reproduce without bringing its own baggage. Kansai is not “the Japanese South.” Osaka is not Texas. Research specifically examining English translations of Kansai speech describes the geographical, cultural, and social specificity of the dialect as a persistent translation problem and documents strategies other than simple target-dialect substitution.

A target dialect can sometimes serve a particular dramatic function, but it always imports associations that were never present in the source. Sometimes standard English with carefully chosen compensatory markers is less distorting than a fake geographical transplant. Cygames recognizes the problem and then solves it by inventing a dialect that does not exist.

Slide 16 explains the method. The team says it needed to convey differences in speech “through text alone,” while avoiding established English dialects. Its solution was to reproduce the “rhythm, speed, and phonetics” of the Japanese through non-standard English spelling. Slide 19 describes the procedure more concretely: read the English line aloud at the rhythm and speed of the original, observe how the English changes, and write those changes into the spelling.

That sounds scientific until the actual example appears.

Tamamo Cross says:

なっ……なに抜かしとんねん！　真剣勝負やぞ！
楽しんで走る余裕なんて、あるわけないやろ！！

The presentation gives this “plain translation”:

Wh-what are you talking about?!
This isn’t the time to be having fun!

Cygames’ version is:

Wh-what’re ya talking about?!
Dis isn’t the time to be havin’ fun!

The localized line is worse, but the “plain” line is not good enough either.

Both versions omit 真剣勝負やぞ！, a complete emphatic clause: this is a serious contest, a serious race, the real thing. Both also flatten the final sentence. Tamamo is not simply announcing that now is an inappropriate moment for fun. She says there is no way she has the 余裕, the leeway or composure, to run while enjoying herself. That difference matters. The line reveals how seriously she is taking the race and how little mental space she has left for carefree enjoyment.

何抜かしとんねん is rougher than a neutral “what are you talking about,” too. 抜かす can function as a coarse or disparaging verb of saying, alongside expressions such as ほざく; the source therefore supplies abrasive force without requiring the translator to manufacture an unrelated English regional identity.

A source-first rendering can preserve all of that without turning her into a collection of pseudo-phonetic spellings:

Faithful natural English: “W-what are you talking about?! This is a serious race! Like I’ve got any time to enjoy myself out there!!”

A slightly freer version can sharpen the roughness of なに抜かしとんねん while staying inside the source:

Natural localized variant: “W-what are you on about?! This is a serious race! Like I’ve got time to just enjoy the run!!”

That line already has voice. The stammer is there. The aggression is there. The clipped rhetorical structure is there. The emphatic repetition is there. “What are you on about?” gives the opening a rougher colloquial edge without assigning Tamamo to a specific English-speaking region. “Like I’ve got time…” reproduces the incredulous structure of the final clause.

No one needs to spell “this” as dis to understand that this is not a calm standard-register speaker.

The dis is where the presentation’s claim about “sonic similarity” becomes especially shaky. English TH-stopping, where voiced th can be realized closer to d, is not some culturally empty by-product of talking fast. Linguistic research documents the feature in multiple real English varieties and examines its regional, ethnic, and social indexing. Research on eye dialect likewise warns that nonstandard spellings are not socially neutral visual representations of sound; they can index class, region, race, education, or caricature even when the underlying pronunciation is not uniquely “incorrect.”

That does not mean Cygames has simply “given Tamamo AAVE,” or any other single real dialect. The problem is subtler. The method claims to escape target-culture dialect baggage by dismantling accents into isolated spellings, but those spellings already belong to the history of written English. Dis, ya, and systematic -in' arrive with associations of their own. The cultural baggage does not disappear because the localization team assembled it à la carte.

This is also a readability problem disguised as characterization. In Japanese, Tamamo’s Kansai speech is part of an established written convention for representing regional speech. The English player is instead asked to decode a deliberately misspelled pseudo-accent. The visual deviation becomes more conspicuous than the underlying sentence.

There are better tools available. Dialectal characterization can be carried through contractions, rhythm, sentence structure, characteristic interjections, lexical preferences, degrees of directness, recurring turns of phrase, and selective compensation elsewhere. None of those techniques requires pretending that a Kansai speaker would transform English th into d.

The point is not to erase Tamamo’s dialect. The point is to stop treating dialect as a bag of phonetic particles that can be poured into another language.

The presentation is right that “Kansai = Southern American” is crude. Its own method merely replaces one crude mapping with a bespoke one.

Yukino Bijin and the Invention of an American Past

The Yukino Bijin example is where the presentation stops merely over-stylizing a line and begins assigning a Japanese character a target-culture history she never had.

Slide 21 says that sound alone is not enough to communicate a character’s “depth or charm.” The solution is to research the character’s history, personality, and cultural background, then let those things influence the English lexicon. That principle sounds reasonable. The problem is visible on the very next example: the team researches Yukino’s Japanese background and concludes that she should speak in dated American expressions from the 1930s through the 1960s.

Yukino is explicitly characterized in Cygames’ Japanese material as a sincere girl from rural Iwate who has come to the city. Her provincial background belongs to a specific Japanese place and social context. None of that makes her an American from the first half of the twentieth century.

The Japanese line on slide 23 is:

は……はひぃぃ～……！！
（どうすりゃいいンだぁ～～～！？）

The presentation’s “plain translation” is:

O-okay!
(Ohhh, I’m in trouble now!)

Cygames changes that to:

O-okey dokeyyyy!
(Ohhh, I’m in the soup!)

Here the supposedly plain translation is already semantically off.

はひぃぃ is a panicked squeal or flustered vocalization, not a clear statement of assent. どうすりゃいいんだ is a question: “What should I do?”, “What am I supposed to do?”, “What do I do now?” Turning it into “I’m in trouble now!” changes a desperate question into a declarative summary of her situation.

The localization then takes that weakened paraphrase and coats it in two pieces of imported American period flavor.

“Okey-dokey” is primarily an expression of assent. Merriam-Webster dates its first known use to 1931. “In the soup” means being in trouble, and Dictionary.com dates the slang expression to the late nineteenth century.

That produces two separate failures.

First, neither phrase recovers what the Japanese actually says. はひぃぃ becomes an affirmative expression, and an explicit “What am I supposed to do?” question disappears altogether.

Second, the supposedly research-driven historical palette is incoherent on its own stated terms. Slide 24 says the team uses expressions from the 1930s through the 1960s to represent Yukino’s “small-town upbringing and innocence,” yet one of its two showcase idioms is documented decades earlier.

More fundamentally, why should a teenager from rural Iwate need a fabricated American time period in the first place?

A direct translation is both more accurate and more characterful because the character’s panic is already written into the Japanese:

Faithful natural English: “H-hyiiii…! (Wh-what am I supposed to dooo?!)”

Or, with slightly less phonetic imitation:

Natural localized variant: “A-ahhh…! (What am I gonna dooo?!)”

The stretched vowels, stammer, and frantic internal monologue carry the scene. Nothing is gained by making a girl from Iwate sound as though she wandered in from a mid-century American comedy.

The deck’s justification is more damaging than the individual phrase. Slide 24 states outright that dated expressions are being used “to show Yukino Bijin’s small-town upbringing and innocence.” That equation is culturally arbitrary. Rural Japanese does not map to old-fashioned American. “Small town” is not a universal dialect code.

The method takes a real Japanese social identity, decides that English readers cannot be trusted to perceive it through translation and context, then substitutes a familiar Anglophone stereotype.

It also damages distinctions within Umamusume itself. Cygames has characters whose Japanese characterization genuinely depends on deliberately dated language. Maruzensky’s official Japanese material, for example, explicitly makes old-fashioned wording part of her comic identity. If dated English becomes a generic localization shorthand for provincial innocence, genuinely dated source speech loses some of its distinctive function.

That is exactly the category of problem the deck identified when it rejected “Kansai = American South”: one English shorthand begins swallowing several different Japanese identities.

The better answer is not to make everyone speak colorless textbook English. Yukino can have a recognizable voice. Her syntax can be hesitant. Her vowels can stretch when she panics. Her vocabulary can remain modest and earnest. Where the Japanese text uses a regional form that has a useful functional equivalent in English, restrained compensation may be appropriate.

What should not happen is the invention of an American temporal identity that the source never gave her.

There is no requirement that every isolated sentence advertise “country girl” through a linguistic costume.

That requirement is one of the deck’s hidden assumptions: a good localized voice must be identifiable immediately from a single line.

But a character voice does not work like a logo.

Characters have signature phrases and neutral phrases. A believable idiolect has range. When every ordinary utterance is pressed into service as a character-identification device, translators start over-marking speech. The character no longer sounds like a person who happens to have particular traits. She sounds like a bundle of traits performing themselves every time she opens her mouth.

The plain version is therefore much better than “okey-dokey / in the soup” in one limited sense: it introduces less foreign characterization. But it still needs correction. “Less rewritten” is not the same thing as “accurate.”

The right comparison is not between boring fidelity and colorful localization. It is between a flawed paraphrase, an even more invasive rewrite, and a faithful line that is already expressive because the Japanese was expressive.

Otaku, Gyaru, and the Replacement of Japanese Culture with Internet English

The “Creating Authenticity” section makes the deck’s target-culture logic explicit. Slide 26 says an English character must possess as much “charm and flavor” as the Japanese character, so the team researches the character’s culture and surroundings and makes the voice “authentic to real life.”

That sounds harmless until the question becomes: authentic to whose real life?

For Agnes Digital, the answer is contemporary English-speaking online fandom.

The Japanese example on slide 28 is:

（ハァ～～～ッ——た ま ら ん ッッッ！！！！）

The plain translation reads:

(Ahhh! This is amazing!)

Cygames gives her:

(Aaaaahasfdhh… The fanservice is unreal!)

Slide 29 explains the process. Find the platforms an otaku character would probably use, then shape her speech with expressions people on those platforms are actually using. The result is labeled “convincingly authentic speech.”

Agnes Digital is indeed explicitly described as an ウマ娘オタク in Cygames’ own Japanese character profile. Her Japanese characterization is already packed with otaku-coded enthusiasm and verbal quirks.

That makes the localization’s reasoning less defensible, not more.

The source does not need the translator to decide what an otaku would say in this situation. The Japanese writer has already decided what this otaku says.

たまらん is a colloquial form of たまらない, carrying the sense of something being unbearable or irresistible and, depending on context, expressing overwhelming positive or negative intensity. Here the positive reading is obvious from the scene: “I can’t take it,” “this is too good,” “I can’t get enough.”

ファンサービス, however, is not present in the Japanese displayed on the slide.

Neither is a keyboard smash.

Those are new pieces of characterization supplied from an English-language mental model of online fandom.

The plain translation is closer because “This is amazing!” at least stays within the emotional content of the line. It is also too bland. There is no need to choose between blandness and invention:

Faithful natural English: “Haaahhh… I can’t take it! This is too good!!”

Or:

Natural localized variant: “Aaaahhh… I can’t get enough of this!!”

Both preserve the overwhelming delight of たまらん. Neither needs to name “fanservice” unless the Japanese names fanservice.

The key rule is simple: a character’s established subculture can guide how genuinely subcultural source language is translated. It does not give the translator a standing license to insert subcultural vocabulary into lines where the writer did not use it.

That distinction disappears almost entirely in the Daitaku Helios example.

Slide 31 says gyaru culture is not very well known outside Japan. The localization team’s answer is to find a “parallel culture” capable of expressing Helios’s chaotic, fun personality. The chosen parallel is “Gen Z.”

A subculture and a generation are not parallel units.

Some gyaru are Gen Z.

Most Gen Z people are not gyaru.

The substitution discards the specific feature and preserves an abstraction:

young + energetic + online + slangy.

This is cultural replacement stated as a production method.

A Japanese gyaru does not become more authentic when rewritten as an American or global-English Zoomer. She becomes more familiar to one target audience.

Worse, the specific Japanese line they choose already contains documented gyaru/teen slang.

Helios says:

ターボいい波ノッてんね～！
こーなったら、『ペガ盛り☆レインボーカツカレー』作るしか！

The deck’s plain translation is:

That’s a great idea! Why not go even further
and make it a pretty rainbow curry?!

The localized version is:

Don’t hate—let her cook! Why not go all in and
make it an Umasta-worthy rainbow curry?!

The Japanese いい波乗ってんね～ is not merely a generic statement that something is a good idea. A 2018 egg survey of 100 models and reader-models ranked いい波乗ってんね～ among that year’s notable gyaru expressions and glossed it as 調子いいね～、イケイケじゃん: roughly “you’re doing great,” “you’re on a roll,” “you’re killing it.” The same report notes that the expression had taken first place in another gyaru buzzword ranking.

That matters enormously.

The cultural specificity that slide 31 treats as a localization problem is sitting right there in the source, doing exactly the characterization work the presentation says it wants to preserve.

A translator does not preserve that by deleting it and importing a different meme.

The Cygames line changes several things at once.

The Japanese directly addresses Turbo by name. The localization drops her name.

The Japanese praises Turbo’s momentum. “Don’t hate” invents critics or opposition who are not present.

“Let her cook” changes the speech act into a third-person instruction to bystanders. It is not culturally neutral contemporary English, either. The expression has a documented history in Anglophone internet slang associated with rapper Lil B from at least 2010 before spreading through music, sports discourse, social media, and memes.

The original proposes making レインボーカツカレー, rainbow katsu curry. Both English versions reduce that to “rainbow curry,” deleting katsu.

The localization then adds “Umasta-worthy,” introducing an imagined social-media criterion that is absent from the displayed source.

The “plain” version is therefore much better only because it does less damage. It still strips out Turbo, the slangy “riding a good wave” idea, katsu, and much of the source’s deliberately flashy menu-name flavor.

A better translation can be lively without pretending Helios grew up on English meme feeds:

Faithful natural English: “Turbo, you’re on a roll! At this point, we’ve gotta make a Pega-size☆Rainbow Katsu Curry!”

A slightly punchier option:

Natural localized variant: “Turbo, you’re killing it! Now we’ve gotta make a Pega-size☆rainbow katsu curry!”

The exact handling of the coined ペガ盛り should be decided from wider in-game context and terminology, not guessed from this isolated slide. That uncertainty is precisely why replacing it with “Umasta-worthy” is so hard to defend. Preserving the coinage, or rendering its size/intensifier function consistently once established, is safer than silently changing it into a statement about social-media appeal.

What can be established from the displayed Japanese is straightforward: Turbo is there, direct praise is there, katsu curry is there, and “let her cook” is not.

This is the deck’s central cultural error in miniature.

Japanese works do not become accessible by pretending that Japanese subcultures are failed drafts of American ones.

Gyaru culture does not need to be replaced because some players are unfamiliar with it. The fact that it is unfamiliar is part of encountering a work from another culture. Translation should make the language legible. It should not make the culture disappear.

The Agnes and Helios examples also expose how quickly “authenticity” becomes temporally fragile. たまらん and Helios’s source-side gyaru language belong to the work. A target-side keyboard smash, meme formula, social-platform reference, or whatever happens to read as “Gen Z” is tied instead to the lifecycle of contemporary Anglophone internet speech. That is an inference from the localization strategy itself, but it follows directly from the fact that the chosen replacement expressions are drawn from current target-language communities rather than the Japanese wording.

The source has a stable historical identity. A meme localization puts an additional expiration date on top of it.

Consistency Does Not Turn Rewriting into Translation

Slides 33 through 38 shift from individual lines to production process. This is the part of the presentation most likely to sound reassuring in isolation. Character voices can drift when many translators work on a large live-service script. Shared glossaries, character-specific guidelines, task assignment, review, and institutional memory are sensible ways to reduce inconsistency.

None of those practices is controversial by itself.

The problem is what they are being used to keep consistent.

Slide 35 says translators were assigned characters they already liked or knew well and calls a character’s first translation a “passion project.” Subject familiarity can help. A translator who understands a character’s history, relationships, jokes, and recurring terminology is less likely to miss context.

But liking a character does not confer authorship over that character.

Passion becomes a liability when “I know what this character is like” turns into “therefore I know what she would say in English even when the Japanese did not say it.”

Slide 36 provides an unusually candid example. The presentation describes a casual Slack channel where team members can exchange experimental ideas, adding the line:

What entertains the team is likely to entertain the fanbase too.

The screenshot shows a discussion of Daitaku Helios’s nickname for Gold Ship, ゴルピッピ. Allison Cottrell asks whether anyone has encountered it and, if not, requests something “appropriately zoomer-y.” Trevor Gafa replies that the following is “Random garbage that my monkey brain just spat out,” then proposes variants including “Grizzy,” “Grizzle,” “G. Shizzle,” “G. Blingz,” and “Gold Spice.” “GOLD SPICE” receives the enthusiastic visible response.

As office banter, the exchange is harmless and funny.

As a showcase of translation methodology, it is devastating.

ゴルピッピ already is a nickname. The ピッピ form has attested cute, affectionate associations in modern Japanese nickname/slang usage; forms such as 彼ピッピ can be shortened to ピッピ, and related -ぴ formations have circulated in youth slang.

The translator’s first obligation is therefore to determine how this particular nickname functions in Umamusume: who uses it, how often, whether its sound matters, whether related -pippi forms appear, how strongly it is tied to Helios’s gyaru voice, and whether preserving the Japanese nickname is understandable from context.

A conservative English solution such as:

Golpippi

retains the invented sound and lets the surrounding characterization teach the reader what kind of nickname it is.

If the connection to Gold Ship needs to be visually clearer:

Gold-pippi

would at least preserve the nickname-building mechanism instead of replacing it wholesale.

“Gold Spice” does not.

It is a newly written English nickname that happens to retain “Gold.” Its selling point in the screenshot is not semantic fidelity, phonetic fidelity, relational fidelity, or preservation of the source-side word formation. The criterion displayed in the presentation is that the option sounds “appropriately zoomer-y” and entertains the localization team.

That line on slide 36, “What entertains the team is likely to entertain the fanbase too,” deserves more scrutiny than it receives.

Entertainment is a legitimate goal for a game. It is not an accuracy test.

A translator can make a room laugh by writing a better joke than the original, a more topical meme, a sharper insult, or a more flamboyant nickname. The success of the new joke does not tell us whether the old joke survived. The author’s line and the localization team’s line are competing for the same limited space, and only one reaches the player.

This is exactly why professional translation assessment separates source fidelity from target-language writing quality. ATA’s framework treats added meaning, omitted meaning, and “too free” creative recasting as transfer problems even if the resulting English is perfectly grammatical and lively. MQM likewise separates accuracy from fluency and provides categories for mistranslation, omission, and addition.

Slide 37 then describes character-specific guidelines containing a character overview, speech/dialect notes, and a glossary, with LQA checking readability and consistency. Slide 38 presents purposeful assignment, the Slack channel, and those character guidelines as a system that ensures consistent speech while enabling “more creative, higher-quality localization.”

Again, those are sensible tools.

But consistency is orthogonal to fidelity.

A team can preserve an invented mannerism perfectly across two million words. It remains invented.

A style guide can prevent five translators from giving the same character five different false accents. It can also ensure that all five use the same false accent.

That creates path dependence.

One translator makes an early interpretive choice.

The choice enters the character guide.

Later translators are expected to reproduce it.

Eventually readers encounter it so consistently that it feels canonical.

Consistency can make an invented target-language trait look like part of the character simply through repetition.

That does not prove the trait originated in the source.

If a target text introduces fanservice, deletes 真剣勝負, replaces a question with “I’m in the soup,” removes Turbo’s name, drops katsu, or inserts “let her cook,” perfect internal consistency does not repair those source-target mismatches. In the terminology used by professional quality frameworks, consistency and fluency do not cancel additions, omissions, or mistranslations.

There is another revealing asymmetry in the presentation. The deck spends substantial time explaining safeguards for recognizability, target-community authenticity, translator specialization, experimentation, readability, and character-voice consistency. It never demonstrates an equivalent checkpoint centered on the most basic bilingual question:

Does the English still say what the Japanese says?

That does not prove Cygames lacks such a check internally. The presentation is not a complete process manual, and it would be irresponsible to pretend otherwise.

It does show what this team chose to celebrate publicly.

The source is repeatedly treated as raw characterization data from which an English voice can be constructed. Once a localization-created speech pattern exists, the production machinery ensures that everyone reproduces it consistently.

That is localization as downstream authorship.

The more organized the process becomes, the more scalable the rewriting becomes with it.

What “Recognizability and Authenticity” Actually Mean Here

This section is about the presentation’s own final claim. The slides present “Recognizability and Authenticity” as the result of the localization method described throughout the talk. The first step, then, is to state what Cygames appears to mean by those terms before judging whether the examples actually support them.

Slide 39 gives the whole argument away in a single sentence. By analyzing “sonic similarities and word choice of various communities and dialects,” the team says it “created unique speech styles and quirks for each character.” The arrow below points to “Recognizability and Authenticity.” Created. That is not an incidental wording choice after a few imperfect examples. It accurately describes the method shown throughout the presentation.

What the Presentation Says It Created

The following is a summary of the presentation’s own method and examples, not an endorsement of them.

Tamamo’s English orthography is created from a phonetic experiment. Yukino’s period-American lexicon is created from an analogy between rural Iwate and an imagined American small town. Agnes Digital’s English fandom voice is created from platform research. Helios’s gyaru identity is converted into a Gen-Z parallel culture. Gold Ship’s nickname is recreated through a target-language brainstorming session. Character guides then preserve the results.

What the Presentation Means by “Recognizability”

In the presentation’s logic, recognizability means giving each character an English voice whose markers are immediately legible to the target audience and remain consistent across the script.

The word “recognizability” is doing a great deal of work. A stereotype is recognizable. A meme is recognizable. A stock dialect is recognizable. A current internet catchphrase is recognizable. Recognition tells us how quickly a target reader can categorize something. It tells us nothing about whether the category is the one the source supplied.

Why Recognizability Is Not the Same as Fidelity

That is where the criticism begins. A feature can be highly recognizable to an English-speaking audience while still being unrelated to the social or cultural information carried by the Japanese source.

“Authenticity” has the same problem. Agnes may sound authentic to an English-speaking online fandom community. Helios may sound authentic to someone’s idea of contemporary Gen-Z banter. Yukino’s expressions may sound plausibly old-timey to an American ear. Those are forms of target-culture authenticity. They are not authenticity to a Japanese otaku, Japanese gyaru, rural Iwate girl, or Kansai speaker. Cygames’ own Japanese material explicitly identifies Agnes as an Umamusume otaku, while contemporary Japanese gyaru material independently confirms that the phrase embedded in Helios’s source line was real gyaru/youth slang. The presentation repeatedly swaps the second kind of authenticity for the first and then uses the same word to make the exchange disappear.

What the Presentation Means by “Authenticity”

The presentation uses authenticity in a similarly target-facing sense: the English voice should resemble speech that feels plausible inside a recognizable English-speaking community or social type.

Slide 40 abruptly points to Umamusume winning Best Mobile Game at The Game Awards 2025. The award is real. Independent reporting on the 2025 awards likewise records Umamusume: Pretty Derby as the Best Mobile Game winner. The game’s success is not in question. It is also irrelevant to the specific claim being argued. A Best Mobile Game award is an evaluation of a game as a whole. It is not a controlled, bilingual source-target assessment of Tamamo’s omitted clause, Yukino’s rewritten question, Agnes Digital’s added “fanservice,” Helios’s meme substitution, or ゴルピッピ. Commercial success, audience affection, and awards can coexist with mistranslated lines. They cannot retroactively make an added joke present in the Japanese.

The Problem: Authentic to Which Culture?

The problem is not that those English voices can never sound authentic. The problem is what they are authentic to. The examples repeatedly measure authenticity against an Anglophone analogue rather than against the Japanese cultural identity encoded in the source.

The final slide says the localization will let trainers experience the world and characters of Umamusume “as intended, with excitement and emotion.” That promise is precisely what the examples fail to establish. “As intended” is a source-side claim. It requires showing that intentions encoded in the Japanese survive the move into English. Excitement and emotion are not substitutes for that demonstration. A rewritten line can be more exciting than the original and less faithful at the same time.

How the Presentation Supports Its Claim of Success

After presenting “Recognizability and Authenticity” as the outcome of its method, the deck points to the game’s reception and awards as evidence that the localization succeeded.

The most defensible principle for translating character-driven Japanese media is much less glamorous than this presentation’s theory, but it produces better work: preserve what the source gives you, find natural English for it, and intervene only as far as the language barrier actually requires.

Why Success Does Not Prove Translation Fidelity

That evidence can support a claim about reception. It cannot, by itself, support a claim about source-target fidelity.

Keep a joke funny when possible, but do not write a different joke because it plays better in Slack. Make dialect audible in prose, but do not assign the speaker a new ethnicity, region, or century. Let an otaku sound like an otaku when the Japanese makes her sound like one, but do not add “fanservice” because an English otaku might say it. Let a gyaru use energetic English, but do not erase gyaru culture and replace it with a meme bundle because “Gen Z” is easier to recognize.

The Presentation’s Final Promise: “As Intended”

The strongest claim comes at the end, when the localization is described as allowing players to experience the characters and world “as intended.” Unlike popularity or recognizability, that is directly a claim about preserving the source.

Good localization removes friction caused by language. It should not remove foreignness caused by the work being foreign.

Why the Examples Do Not Establish That

That is the broader problem exposed by this deck. Aggressive localization often begins from a flattering story about respect for the character. The source is said to contain an “essence” too delicate for ordinary translation, so the localizer must become more creative, more culturally adaptive, more willing to rewrite. Once that premise is accepted, every source detail becomes negotiable.

What a Better Translation Principle Looks Like

Once the presentation’s claims and the criticism are separated, the alternative is straightforward: characterful English does not require stripping away the cultural location of the Japanese line.

A regional dialect can be replaced by invented spelling. A rural identity can be redressed in American nostalgia. A Japanese subculture can be mapped onto a Western one. A simple emotional outburst can acquire fandom terminology. A nickname can become a brainstormed gag. The process calls the result more authentic because the target audience recognizes the replacement more quickly.

Calling that cultural vandalism is severe language, but this presentation comes remarkably close to making the case for it itself. The damage is not that English sentences are allowed to sound natural. They should. The damage occurs when the Japanese cultural and authorial marks are scraped off so that more familiar Anglophone marks can be painted over them, while the result is still presented as the original character “as intended.”

The plain translations shown in the deck are frequently clumsy and sometimes incomplete. Several need correction. Yet they are usually the better starting point because they leave the translator something to improve rather than something to undo.

What Happens When Target-Culture Recognizability Becomes the Priority

The broader criticism follows from the same distinction. Once recognizability to the target audience is allowed to outrank fidelity to the source, replacement starts to look like preservation.

Tamamo needs her omitted “serious race” clause restored, not dis. Yukino needs her actual frightened question translated, not “in the soup.” Agnes needs the intensity of たまらん, not an invented fanservice reference. Helios needs her direct praise, Turbo’s name, and rainbow katsu curry, not “let her cook” and “Umasta-worthy.” ゴルピッピ needs its nickname mechanism respected, not replaced by “Gold Spice.”

Those conclusions line up with the most basic distinction made by professional translation-quality frameworks: a translation must be idiomatic in the target language and faithful to what the source communicates. Additions and omissions do not become invisible because the resulting English has stronger “flavor.”

What the Actual Lines Needed Instead

The earlier examples make the distinction concrete. None of them required the English to be lifeless; they required the translator to preserve information and characterization that were already present before adding new target-culture material.

Natural English is not the enemy of fidelity. Restraint is not the enemy of personality.

Natural English and Fidelity Are Not Opposites

A translator does not have to choose between a word-for-word-Japanese-sentence-structure crib and a new character written for the target market. The real skill lies in the space the presentation keeps skipping over: faithful English that trusts the original writer enough to let the original character survive.

The English-speaking player came to Umamusume to meet Umamusume. They did not need the localization room to stand between them and the work, replacing pieces of Japanese culture with a collage of fake accents, vintage American idioms, fandom jargon, and current memes. The best translation would make that room disappear.