Emoji Pack Ideation

Goal

Turn a rough audience, emotion, character, meme trend, reference image, or topic into a practical WeChat sticker pack plan: positioning, style direction, character persona, concrete sticker contents, and image-generation prompts when needed. Prioritize stickers that users will save, reuse, send in chat, and potentially reward/tip.

Mode Selection

Use Topic Planning Mode when the user wants 表情包选题, 爆款方向, 打赏型策划, 人设, 风格建议, or content lists.

Use 16-Grid Rough Sticker Prompt Mode when the user provides or mentions reference images, a 16-grid sticker layout, 多行中文文案, 鼠标乱画, MS Paint, 低质量真实感, or asks to output a prompt/spec for generating a complete sticker collage.

Core Judgment

Optimize for chat utility before pure illustration quality. A strong pack usually has:

• Accurate high-frequency emotion: speechless, collapse, perfunctory, sarcastic, aggrieved, smug, guilty, social-anxious, cute, healing, or mildly unhinged.
• Frequent chat scenarios: work replies, friend roasting, group chat watching, intimate relationship comfort, greetings, thanks, apologies, refusal, and ending conversations.
• Short sharp text: usually 2-8 Chinese characters, readable in a tiny WeChat chat bubble.
• Clear small-size silhouette: exaggerated expression/action, low clutter, strong outline.
• Stable persona: users should feel they are adopting a chat personality, not sending random one-off images.
• Mild edge without real harm: playful, slightly rude, or teasing is fine; avoid cruelty, heavy insult, and overly niche references unless requested.

For tip/reward potential, weight companionship and emotional value more heavily than shock value. Reward-friendly packs often feel soft, repeatable, complete, and made with care.

Topic Planning Workflow

1. Clarify or infer the target user and relationship context: friends, coworkers, couples, family, fan community, students, creators, or general WeChat users.
2. Choose one primary emotional engine. Do not make the pack only generally cute or funny; anchor it in a repeatable tension such as soft but resentful, obedient but internally screaming, cute but poisonous, fragile but stubborn, calm outside and broken inside.
3. Define the character persona in one sentence: species/person/object, personality, speech habit, and social role.
4. Recommend visual style around chat readability: shape language, line weight, color palette, expression exaggeration, typography placement, and animation/static suggestion if relevant.
5. Build a complete content matrix, covering daily-use stickers before novelty stickers.
6. Score the idea and identify the weakest dimension before giving the final recommendation.

Scoring

Score each dimension from 1-5 and provide a one-line rationale:

• Emotion precision: does it express a real high-frequency feeling?
• Reuse frequency: will users send it every day or every week?
• Persona memorability: is the character recognizably itself across 16-40 stickers?
• Tip potential: does it create companionship, cuteness, empathy, or gratitude?
• Production scalability: can the same visual system produce many stickers without becoming repetitive?
• Safety/social comfort: can users send it without awkwardness in common relationships?

If the average is below 3.5, revise the topic before drafting a full pack.

Style Recommendations

Always give concrete style advice, not generic adjectives. Include:

• Character form: animal, dumpling-like creature, office object, tiny deity, human chibi, food, plant, blob, or other form suited to the premise.
• Shape and expression: round/soft for reward-friendly companionship; angular/squashed for sarcasm or collapse; big eyes and tiny limbs for weak-but-cute energy; deadpan face for cold humor.
• Color system: 2-3 main colors plus text color. Favor clean, readable contrast over complex gradients.
• Text treatment: short Chinese phrases, thick readable font, text outside the most expressive face area, no tiny multiline paragraphs.
• Action grammar: repeatable poses such as bowing, rolling away, lying flat, holding sign, peeking, crying, hugging, clapping, saluting, hiding, melting, exploding, and presenting a blank face.
• WeChat readability: large face/action, minimal background, strong contour, avoid tiny props unless they carry the joke.

Content Matrix

When asked for content suggestions, output 16-24 stickers by default unless the user requests another number. Cover these buckets:

• Basic replies: 收到, 好的, 嗯嗯, 在呢, 不在, 明白了
• Emotion reactions: 笑死, 无语, 崩溃, 委屈, 害羞, 得意
• Social moves: 谢谢, 对不起, 辛苦了, 抱抱, 加油, 晚安
• Refusal and boundary: 不要, 别催, 我撤了, 你先, 下次一定
• Relationship/intimacy: 想你, 贴贴, 哄我, 原谅你了, 亲亲, 求抱
• Work/study pain: 在忙, 已读乱回, 假装懂了, 不想上班, 马上改
• Group-chat utility: 吃瓜, 围观, 路过, 鼓掌, 震惊, 退下了
• Signature meme pieces: 2-4 phrases that only this persona would say.

For each sticker, include: text, emotion/use case, visual action, and why it is useful.

16-Grid Rough Sticker Prompt Mode

Use this mode to produce a single copy-ready generation spec or prompt for a 16-sticker collage based on reference images and Chinese copy.

Inputs To Expect

• Image 1: character reference. It may contain one or more people, pets, or combinations.
• Image 2: layout reference. Use it only to understand a 4x4 16-grid structure; do not copy its style unless requested.
• User copy: multiline Chinese text, fewer or more than 16 lines.

If images are not actually attached, state that the prompt is a template and leave bracketed placeholders such as [角色参考图] and [16格版式参考图].

Character Recognition Rules

Identify every usable character in the reference image. Any human, animal, or pair/group can become a sticker protagonist. Different stickers may use different characters. Multi-character interaction is allowed when the chat emotion benefits from it: sticking together, fighting over space, staring, grabbing, pushing away, comforting, or ignoring each other.

Preserve only the minimum recognizability traits: hairstyle outline, rough color blocks, iconic clothing/accessory, body size relationship, pet ears/tail/markings, or another obvious marker. Do not preserve realism.

Error Consistency Rule

The collection must use error consistency, not realistic consistency.

The same character does not need to look identical across stickers. It must look like the same unskilled person drew it badly 16 times. Keep consistent:

• the same clumsy drawing method
• the same simplification choices, such as symbolic face, stick body, blob body, or lopsided head
• the same mistakes, such as shaky lines, bad proportions, misplaced features, and missing details
• the same level of visual chaos

Allow facial drift, changing proportions, broken anatomy, missing details, and structural mistakes. Consistency means being wrong in the same way.

Layout Rules

Generate exactly 16 stickers in a clear 4x4 grid. Each cell must read as an independent chat sticker. A single sticker may be messy, but the overall grid must remain legible and separated.

Never use visible numbering in the final visual. Do not make the stickers feel sequential. Treat the 16 cells as 16 independent, random, temporary expressions.

Copy Handling

If the user provides fewer than 16 lines, complete the rest in the same voice. If they provide more than 16 lines, choose the 16 with strongest emotion, spread, and visual potential.

Prioritize:

• roasting, self-talk, emotional explosion, perfunctory replies, impatience, absurdity
• internet-native Chinese phrasing
• uneven length: some very short, some crowded, some almost blank
• strong readability and instant emotional recognition

Avoid polite standard replies, stiff customer-service phrasing, repeated sentence structures, and common template phrases such as 你好, 哈哈, 在吗 unless the user explicitly wants them.

Handwritten Text Rules

Text must feel drawn, not typeset. Specify that Chinese words are mouse-handwritten: crooked, shaky, uneven in size, irregular spacing, squeezed, tilted, partly overlapping, sometimes attached to the character, sometimes on the edge.

Allow repeated characters, stretched sounds, chaotic punctuation, and ugly handwritten forms. Preserve readability; the audience must still understand the text.

Visual DNA

The style source is extremely bad hand drawing + MS Paint mouse scribble + failed imitation of the reference image. Enforce:

• shaky jagged mouse lines
• rough pixel edges
• failed proportions
• stick figures, symbolic faces, blob bodies, or very simplified shapes
• almost no polished detail
• low-quality, real, funny, direct, absurd emotion
• blurred resemblance to the reference image, like a failed copy drawn from memory

When realism conflicts with style, prioritize style. If the output starts looking beautiful, polished, correctly proportioned, clean, or designed, it has failed. Push it back toward rough, messy, low-skill, and funny.

Image-Text Relationship

The image and text must feel created in one action: someone draws badly with a mouse and writes whatever they think at that second. Do not describe it as designed artwork with later captions added. Each cell should feel accidental, temporary, half-bothered, and unexpectedly funny.

The text may intensify the emotion, roast the character, become self-talk, or slightly mismatch the drawing to create absurd humor. Precision is less important than funniness.

Output Format For This Mode

Default to Chinese. When the user asks for a prompt, output a ready-to-copy prompt in this structure:

• 任务目标
• 输入图像说明
• 角色识别与使用规则
• 16格结构
• 文案处理
• 错误一致性
• 文字绘制规则
• 视觉风格DNA
• 每格生成要求
• 负面约束
• 最终效果

When useful, add a compact 16格内容分配 table before the final prompt. The table may be numbered for the user to review, but the generated visual itself must not contain numbers.

Topic Planning Output Format

Use Chinese by default. Keep the answer practical and production-oriented. Preferred structure:

1. 选题定位: one concise recommendation.
2. 适合人群/聊天关系: who will use it and where.
3. 人设: stable character description and speech style.
4. 风格建议: visual form, palette, typography, action grammar.
5. 爆款/打赏判断: score table or compact score bullets with improvement notes.
6. 表情包内容建议: a table with text, scene, visual action, and priority.
7. 可扩展方向: how to make series 2 or seasonal variants.

Practical Defaults

• If the user gives no topic, propose 3 directions and recommend the strongest one.
• If the user wants a reward/tip-friendly pack, bias toward soft cute companionship plus high-frequency daily phrases.
• If the user wants viral spread, allow sharper sarcasm and stronger contrast, but still keep the pack usable in normal WeChat chats.
• If the user wants commercial/IP potential, emphasize persona consistency, series naming, color ownership, and scalable content buckets.
• If the user gives an existing character, preserve its core traits and translate them into chat scenarios instead of replacing the character.
• If the user asks for the rough 16-grid style, do not polish. The aesthetic goal is: a person who cannot draw used a mouse to scribble 16 times and accidentally made something funny.