This is a skill dedicated to lettering that handles speech bubbles, typefaces, and sound effects. The English snippets that appear here are not interpreted or translated, but used as-is and copy-pasted into AI image prompts — the name itself is a working token. It is a hub shared by three skills: episode (planning), art (prompt insertion), and qc (readability judgment).
1. What this skill does
- Speech bubble selection — scene emotion → choose a name → copy-paste the snippet. (
references/balloon-dictionary.md) - Typeface decision — one manuscript = one typeface. The skill decides it and does not ask the user. (
references/korean-lettering.md) - SFX style specification — Korean onomatopoeia → letter style × effect mapping. (
references/sfx-dictionary.md) - Placement verification — safely overlay SFX with a 2-pass overlay and control size and position. (
references/two-pass-overlay.md)
2. Core principles (short and always needed)
- The name forces the form — The form word in the name ("Spike") cannot be removed by the prompt. If you want to remove it, use a name that does not contain that word.
- No over-classification — Similar things converge into broad categories. Emotion labels (anger, shock, etc.) are not made into separate items but are mapped to forms.
- Tail rule — The tail points toward the speaker's mouth. Types without tails: thought (Aura, cloud), electronic/phone, narration box. Exception: Held-Back Spike has the spike itself as the tail (no separate tail allowed).
- One manuscript = one typeface — Depending on the manuscript tone, use upright Gothic or upright Myeongjo, decided by the skill and applied identically to all episodes. Handwriting fonts and distinctive typefaces are prohibited for body text (readability and consistency take priority).
- 2-pass = removal is dangerous, addition is safe — Leave the completed manuscript as is and overlay only the SFX. Avoid the risk of regeneration.
3. Three letter branches (manuscript type determination)
Lettering work starts by determining which branch the manuscript belongs to.
| Branch | 뜻 | This skill's involvement |
|---|---|---|
| Clean manuscript | No text (no speech bubbles or SFX) | No lettering. Dialogue is handed off to the editing tool as a references/dialogue-memo.md script |
| Dialogue included | Generate speech bubbles and dialogue in pass 1 | Apply speech bubble dictionary + typeface rules |
| Large text / SFX | Overlay SFX in pass 2 on top of the dialogue-included version | Apply SFX dictionary + 2-pass overlay |
The clean manuscript can be recovered by removing the speech bubbles from the finished version (maintaining the original proportions). It is a revision safety net linking dialogue-included ↔ clean.
Per-branch prompt tail
| Branch | Phrase to insert at the very end of the prompt |
|---|---|
| Clean manuscript | no text, no speech balloons, no SFX anywhere |
| Dialogue included | draw ONLY the specified 💬 balloons; no other text |
| Large text / SFX | (2-pass) keep the existing art & balloons 100% intact; add SFX only |
4. Speech bubble quick reference (frequently used cores)
About 40 types in total are references/balloon-dictionary.md. Only the most frequently used cores are placed inline.
| Emotion | Name | Snippet |
|---|---|---|
| Ordinary dialogue | Normal Speech | a normal speech balloon (smooth oval, small tail to mouth) |
| Thought (default) | Fine Spike Aura | a "Fine Spike Aura" balloon (oval with a fine thin spiky aura outline, NO tail) |
| Whisper | Whisper | a WHISPER balloon (dashed/dotted outline) |
| Shout·Scream | Loud Shout/Scream | a loud scream balloon (sharp spiky burst) |
| Surprise·Shock | Surprised/Shocked | a "Surprised/Shocked" balloon (jagged burst outline) |
| Suppressed tone | Held-Back Spike | a "Held-Back Spike" balloon (angular polygon body; the single SPIKE serves AS the tail) |
| Narration | Narration Box | a NARRATION BOX (rectangular caption box, no tail) |
| Over the phone | Telephone/Electronic | a TELEPHONE/ELECTRONIC balloon (rectangular zig-zag electronic outline) |
5. Speech bubble selection
Decide the scene emotion and from the dictionary NameAfter choosing, copy-paste that English snippet into the corresponding cut prompt. It is divided into core (dialogue/thought·shout/emotion·special) and extended (digital·media devices·environment·fantasy). Position behavior based on vertical scroll(Margin band intrusion rules, upper limit on the number of speech bubbles shown simultaneously on one screen) are placed in a separate section within the dictionary.
→ Details: references/balloon-dictionary.md
6. Font decision (one manuscript = one font)
- Between Jeongja Gothic (bright·everyday·comedy·action) / Jeongja Myeongjo (calm·emotional·drama·period piece), based on the manuscript tone, the skill decides.
- Prompt pinning:
clean upright Korean Gothic sans-serif, even stroke weight, identical typeface on every page(if Myeongjo,Myeongjo serif). - When attaching the previous page, always specify "same font" — otherwise the weight and roundness waver from cut to cut.
- A font mapping table for each text element (speech bubble body/SFX/title/sign/digital screen) is also provided.
→ Details: references/korean-lettering.md
7. SFX style specification
SFX is combinatorial — Sound (text content) × Letter style (L1L16) × Effect processing (E1E10) × Placement. Sound is free input, only the style is chosen from the dictionary.
Frequently used mappings (full list in references/sfx-dictionary.md):
| Sound type | Style | Effect | Representative (Korean) |
|---|---|---|---|
| Impact·Strike | L1 Thick block / L2 Cracked | E1 Burst / E6 Crack | Kwang·Kwajik·Puk |
| Motion·Movement | L4 Slanted | E2 Speed lines | Shyuk·Tadadak |
| Emotion·Heart | L3 Round | E3 Pulse ring | Dugeun·Kongdak |
| Collapse·Crumbling | L11 Disintegration·Collapse | E6 Crack / E5 Dust | Warrr·Kugugung |
| Flame·Blaze | L15 Flame | E1 Burst / E5 Sparks | Whoosh·Blaze |
- Generation formula:
이름 토큰 + 형태 묘사 + 소리별 컬러. If you give it only a name, it gets muddled. - Readability tradeoff: Shatter·Lightning·Scream are harder to read the stronger they are → only at emotional peaks and short sound effects.
- L11은 Hangul Jamo separationincludes fragment texture levers premised on this (Gothic=stone chunk/Handwriting=ink drop).
→ Details: references/sfx-dictionary.md
8. 2-Pass Overlay Workflow
- 1st pass — Generate up to the clean draft + speech balloons (the finished version to preserve).
- 2nd pass — Attach the finished version and add only SFX. Lock in "100% preservation of existing art·speech balloons."
- On typos·failures, re-run only the 2nd pass from the 1st-pass original.
Placement rules to follow together:
- Size: Determined by sound intensity·distance. No oversizing relative to the screen·cut, focus on only 1~2 emotional peaks(do not attach to every sound).
- Same-texture trap: Water-on-water·fire-on-fire SFX get buried → thicken color + white outline, or move to background whitespace.
- Medium match: In painting styles such as oil·watercolor, there is a tendency to revert SFX to smooth vectors, so lock in "same brushstrokes·texture as the art" as a positive instruction.
- No encroaching on vertical whitespace bands: Ensure SFX do not cover the empty band between cuts or the art of the cut below.
→ Details: references/two-pass-overlay.md
9. Usage Pattern (How to Lock Into Prompts)
Under each cut's scene description, write 💬로 [predefined English snippet] + one line of position + Korean dialogue. Replace the clean warning at the very end of the prompt with **"draw only the specified 💬 speech balloons and no other text"**.
- 컷4 (전폭):
지우가 폰을 보고 흠칫 놀라는 상반신.
💬 a "Surprised/Shocked" balloon (jagged burst outline), tail to her mouth — 지우 머리 위, "헉, 엄마?!"
10. Related Skills
| Relation | Skill | Role |
|---|---|---|
| Before | story-webtoon-episode |
Pre-plan speech balloon·SFX placement in the cut script |
| Post-reference | story-webtoon-art |
Insert snippets into art prompts |
| Post-review | story-webtoon-art (qc branch) |
SFX readability judging |
| Sister hub | story-webtoon-spec |
Cross-reference manuscript specs·terminology glossary |
| Font license | moai-designer:design-system-library |
For Hangul typeface licenses, refer to this skill |
11. References
| File | Load condition |
|---|---|
references/balloon-dictionary.md |
When speech balloon type selection·tails·vertical placement is needed |
references/sfx-dictionary.md |
When Korean SFX style specification is needed |
references/korean-lettering.md |
When typeface unification·per-element typeface decisions are needed |
references/two-pass-overlay.md |
When overlaying SFX in 2 passes and controlling size·placement |
references/dialogue-memo.md |
When creating a dialogue script to hand off to editing tools from the clean draft branch |
Source
- Core asset: aitoon direction·planning system (internal asset port, adjusted for vertical scroll·Korean platform context).
- Original copywriter: Jo Nam-kyung (https://www.facebook.com/Bmisty)