Viral Sticker Pack

Turn “looks good” into “this says exactly what I mean.” Optimize for repeated use first, then memorability and forwarding.

Core equation

Use this working model:

viral potential = high-frequency moment × emotional precision × character recognition × forwarding impulse × updateability

Treat it as a design heuristic, not a guarantee. Read references/viral-framework.md when developing positioning, copy, a hero sticker, or a test plan.

Workflow

1. Lock the audience and send moments

Identify the primary audience, relationship context, character type, tone, and target platform. Ask only when the missing choice would materially change the pack; otherwise state a reasonable assumption and proceed.

Describe every sticker as a concrete chat moment: “when would someone send this, and to whom?” Reject abstract emotions without a send moment.

2. Define one sentence of character IP

Write the persona as:

identity + contradiction + signature habit

Also define one visual hook and one verbal hook. Prefer tension such as “polite outside, furious inside” over generic virtues such as cute, kind, or perfect. Keep the visual hook original and clear at 120 px; avoid resemblance to existing IP.

3. Build the content matrix before generating art

Create a plan with these columns:

index | caption | category | audience | send moment | emotion | pose/action | prop/effect | silhouette | visual hook | tier | scores

For a general 16-sticker pack, start from:

• 5 daily replies
• 4 target-work or target-context survival lines
• 3 social attitudes
• 2 intimate reactions
• 2 character-only signature jokes

Adapt the categories to the target audience, but preserve the jobs: frequent use makes the pack useful; signature jokes make it memorable.

Use the longevity mix:

• 70% evergreen, high-frequency moments
• 20% audience-specific language
• 10% surprising, strange, or highly screenshotable images

Read references/plan-schema.md before writing a machine-checkable plan.

4. Make emotion arrive before text

Require one dominant emotion and one key action per sticker. Make the subject occupy roughly two-thirds or more of the usable frame. Use face, eyes, hands, posture, and silhouette to communicate before the caption becomes legible.

Prefer close-up faces, kneeling, collapsing, freezing, peeking, turning away, forced smiles, clasped hands, or exaggerated pointing when they match the send moment. Remove decoration that does not contribute to meaning.

Keep captions conversational and usually 2–8 Chinese characters. Write language a user can naturally send, not slogans or ad copy. Use copy to perform a social action: agree, refuse, apologize, self-deprecate, tease, comfort, or set a boundary.

5. Choose one hero sticker

Pick the image that can explain the character and emotion without context. It should work as a comment reply, screenshot, or single-image share.

Design 3–7 nearby derivatives around the same character tension and chat context. Do not market every sticker equally; use the hero as the entry point to the full pack.

6. Score and gate the plan

Score every concept from 0–2 on:

• send-fit
• emotion clarity
• visual recognition
• forwarding impulse
• character continuity

Revise concepts below 8/10. Require the hero sticker to score at least 9/10.

Run:

python scripts/check_content_matrix.py <plan.json>

Treat warnings as revision prompts even when the script exits successfully.

7. Pilot before scaling when validation is the goal

For a new character or uncertain audience, produce 6–8 pilot stickers first. Test actual usage, not stated preference:

• first-use or save rate
• repeat use within a week
• forwarding into other groups
• spontaneous recaptioning or remixing

Seed with active users who participate in multiple chats. If the user explicitly requests a complete production run, execute the full set but still review the first wave at 120–240 px before finishing the remainder.

8. Hand off to production

For image generation, transparency, animation, supporting assets, compression, validation, packaging, or upload, also use $sticker-pack-production.

Preserve these design decisions during production:

• Keep the same character identity, voice, visual hook, and proportions.
• Vary at least three high-salience features between adjacent stickers: pose, framing, expression, prop, effect, or silhouette.
• Add exact Chinese text locally after art generation when possible.
• Build an actual-size contact sheet and re-check emotion, copy, distinction, and hero dominance.
• Build the 50×50 chat-panel icon as a dedicated transparent, head-only asset. Do not shrink or loosely crop a decorated sticker; remove text, body, clothing, props, stars, sparkles, petals, clouds, frames, and every detached ornament, then inspect the result at actual size.
• Save a reversible platform draft only; never submit for review without explicit approval.

Failure checks

Reject or revise a pack when it has only cuteness, captions without actions, long copy, repeated poses, private jokes outsiders cannot decode, all-negative emotions, expiring weekly slang, obvious IP similarity, or unlicensed photos/fonts/film material.

Do not claim that a score, study, or visual pattern guarantees virality.