SuperAPI AI Directory

humanizer

Writing & research

View repository

Updated 10-16 16:28

About this skill

Humanizer

Remove the "machine smell" from AI-written Korean text.

1. AI Text Detection Patterns

If there are 3 or more of the signals below, humanizing is essential:

Signal Example
Excessive structuring Every paragraph is topic sentence → explanation → conclusion
Repeated transition words "Also", "however", "meanwhile", "like this" at the start of every paragraph
Uniform sentence length Sentences of 15–25 characters repeat, no short/long sentences
Emotionless narration Only listing facts, no subjectivity, exclamations, or questions
Clichéd expressions "We cannot overlook the importance of ~", "in terms of ~"
Perfect grammar 0% grammar errors, 0% inversion, omission, or unconventional constructions
List parallelism "First, second, third" or "not only A but also B" repeated

2. Korean-Specific Issues

Translationese (the biggest enemy)

❌ Translationese ✅ Natural Korean
It is important to ~ You should ~ / ~ is the key
Let's talk about ~ Let me talk a bit about ~
We can see the fact that ~ It's that ~ / I found that ~
This is ~ This is ~ / You see, ~
Provides an opportunity to ~ You can ~
Plays a very important role in ~ It's really important in ~
It is an undeniable fact that ~ Right, ~

Excessive honorifics/polite forms

  • "Would you ~" → in blogs, "Shall we ~?" or "I'll try ~"
  • "Dear readers" → "you all" or omit
  • "~습니다" in every sentence → mix in "~예요/~거든요/~죠"

Unnatural ending patterns

  • Every sentence ends with ~합니다 → mix in haeyo-che/casual speech/noun endings
  • ~입니다 overused → "~거예요", "~할 거고", "~하려고요"
  • No interrogative sentences → insert self-questions like "But why is that?" in the middle

3. Fiction Humanizing

Differentiating character speech styles

❌ 모든 캐릭터가 동일한 경어체
"저는 그렇게 생각합니다." / "저도 동의합니다."

✅ 캐릭터별 개성
노인: "그게 말이야, 옛날엔 그런 게 없었어."
10대: "ㄹㅇ? 대박인데 그거."
직장인: "아, 그건 좀… 위에서 뭐라 하지 않을까요?"

Specific conversion rules

Rule Application
Incomplete sentences "So I—" (cut off), "Well, it's like that."
Repetition/speech tics Give each character a habitual word ("근데", "아니 근데", "솔직히")
Ungrammatical dialogue "I saw them yesterday and they were really funny, ah seriously"
Sensory description Directly stating emotions ❌ → through physical reactions ("손이 떨렸다", "목이 막혔다")
Varying sentence length Tense scenes = short sentences. Relaxed scenes = long, slow sentences.
Onomatopoeia/mimetic words "쾅", "후다닥", "찔끔" — actively leverage the strengths of Korean
Inserting colloquial speech Raw expressions like "아, 씨." "뭐야 이게." in inner monologue

Diversifying emotional expression

❌ "그녀는 슬펐다."
❌ "그의 눈에 눈물이 고였다." (클리셰)

✅ "입술을 깨물었다. 안 울 거다. 울면 지는 거다."
✅ "코끝이 시큰했다. 아, 됐다. 다 됐다고."

4. Blog Humanizing

Forcing insertion of a personal tone

  • One personal experience in the first paragraph ("저도 처음엔 몰랐는데요")
  • A failure story or doubt in the middle ("근데 이게 진짜 맞나? 싶었어요")
  • Perfect information delivery ❌ → exposing the trial-and-error process

Specific conversion rules

Rule Application
Rhythm variation Long sentence → short sentence → one word. Don't repeat.
Using parentheses "(Honestly, this is a bit annoying)"
Directly addressing the reader "Have you ever had an experience like this?"
Switching to informal speech No 3 consecutive "~입니다" sentences. Insert "~거든요", "~더라고요" in between
Numbers/specificity "A lot" ❌ → "3 times", "2 weeks", "about 70% from my experience"
Stating opinions "In my opinion", "to be honest", "personally"
Acknowledging uncertainty "I'm not sure about this, but", "I could be wrong, but"

Before/After: Blog

❌ BEFORE (AI 냄새)
Docker는 컨테이너 기반의 가상화 기술입니다. 이를 통해 개발자는
애플리케이션을 일관된 환경에서 실행할 수 있습니다. 또한 배포
과정을 간소화할 수 있다는 장점이 있습니다. 따라서 많은 기업에서
Docker를 채택하고 있습니다.

✅ AFTER (사람 냄새)
Docker 처음 접했을 때 솔직히 뭔 소린가 했어요. 컨테이너?
가상화? 그냥 서버에 올리면 되는 거 아닌가?

근데 한번 써보니까 아. 이래서 다들 쓰는구나. "내 컴에서는
되는데요?" 이 말을 더 이상 안 하게 되더라고요.

환경 세팅에 하루 날려본 적 있으시죠? (저는 3번쯤 날렸습니다)
그게 docker-compose up 한 줄이면 끝나요. 진짜로.

Before/After: Fiction

❌ BEFORE (AI 냄새)
민수는 카페에 들어갔다. 그는 커피를 주문했다. 잠시 후 그는
창밖을 바라보며 생각에 잠겼다. 그의 마음은 복잡했다. 어제
있었던 일이 계속 떠올랐기 때문이다.

✅ AFTER (사람 냄새)
카페 문을 밀었다. 에어컨 바람이 확 끼얹어졌다.

"아메리카노 하나요. 얼음 많이."

자리 잡고 앉았는데, 아까 그 표정이 또. 씨발. 자꾸 떠오른다.
민수는 빨대를 씹었다. 벌써 세 번째 씹는 거다.

창밖에 비가 올 것 같았다. 아니, 이미 오고 있나.

5. Pre-submission checklist

After completing the text, check all of the following items:

  • Does "또한/그러나/따라서" appear 3 or more times?
  • Does every sentence end with "~습니다/~입니다"?
  • Are there any translationese expressions? ("~에 있어서", "~하는 것이 중요")
  • Is there variation in sentence length? (At least 20% short sentences)
  • Is at least 1 personal opinion/experience/emotion included?
  • Does it sound natural when read aloud?
  • (Fiction) Is the dialogue distinguishable per character?
  • (Fiction) Are there no 2 or more consecutive sentences starting with "그는/그녀는"?
  • (Blog) Is there at least 1 sentence directly addressing the reader?
  • Is there at least 1 question sentence?

6. Banned words/banned patterns

Replace immediately when used:

Banned Replacement
importance of ~ the reason ~ is important is
we will look at ~ let's look at ~ / let's talk about ~
we can say that ~ it's ~ / it's ~
there are various ~ there are a few
we will find out about ~ let's dig into ~
in conclusion (delete and wrap up naturally)
to summarize (delete and wrap up naturally)
it is essential to ~ you have to ~ / it's a big deal if you don't do this