SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsdouyin-viral-content

douyin-viral-content

Marketing & growth

View repository

Updated 10-16 09:45

About this skill

Douyin viral copy generator

Core concept

The user only needs to say: "generate new copy" and provide book excerpts or quote content.

All analysis, optimization, scoring, and estimation are fully completed automatically until a 5-star viral copy is output.

Fully automated workflow

When a copy generation request is received, automatically execute the following steps:

Step 1: Automatic reading and analysis of historical data

Automatically scan 已发布/ all historical copy in the folder:

  • Extract each copy's estimated play count, actual play count, and scores in each dimension
  • Identify which copy with high scores in viral elements actually performs better
  • Analyze whether the estimation model has systematic bias (always overestimating or underestimating)
  • Summarize the account's content characteristics (which type of emotion, style, and topic it is better at)

If historical data is insufficient (fewer than 3 items): use the general model, no need to adjust strategy

Step 2: Automatic strategy calibration

Based on historical analysis results, automatically adjust:

  • Element weights: increase the weights of elements that perform well (e.g., if copy with high scores in "emotional power reversal" has good play count, then strengthen that element)
  • Estimation parameters: calibrate the play count estimation model (e.g., if history is always overestimated by 20%, automatically lower the estimation coefficient)
  • Tag strategy: optimize tag selection (e.g., if certain tag combinations perform better, prioritize them)
  • Content direction: strengthen the content types the account is good at

Step 3: Apply the 9 major viral elements to generate copy

Reference references/viral-factors.md, ensure the copy contains key elements:

  1. Emotional power reversal
  2. Cognitive restructuring
  3. Authoritative endorsement
  4. Fatalistic philosophy
  5. Algorithm-friendly tags
  6. Quote spreadability
  7. Completion rate optimization
  8. Interaction hook design
  9. BGM adaptation suggestions

Incorporate historical learning: based on the strategy adjustments in step 2, strengthen successful elements.

Step 4: Multi-dimensional scoring evaluation

Reference references/scoring-system.md, automatically calculate:

  • Content quality score(100 points): viral element coverage, quote quality, emotional resonance
  • Algorithm adaptation score(100 points): tag strategy, completion rate expectation, interaction guidance
  • Innovation score(100 points): perspective innovation, differentiated expression
  • Comprehensive viral index(total score 300, converted to a 5-star scale)

Step 5: Quality control and automatic optimization iteration

Quality standard: must reach 5 stars (240-300 points) before output

If the first generation does not reach 5 stars:

  1. Identify the lowest-scoring dimension (e.g., insufficient innovation)
  2. Automatically optimize this dimension (e.g., change the angle, strengthen the key quote, optimize the opening)
  3. Re-score and evaluate
  4. Repeat iterations until the 5-star standard is reached

Automatically avoid risks: Automatically avoid potential risks during the generation process (such as copy being too long, the opening not being attractive enough, tags not being positive enough, etc.), without notifying the user.

Automatically apply optimization points: Directly optimize the copy during the iteration process, without telling the user what was optimized.

Step 6: Intelligent play count estimation

Reference references/estimation-model.md, estimate based on the following factors:

  • Content quality score (the scoring result from Step 4)
  • Account historical performance (average play count, follower count)
  • Track competitiveness (baseline traffic for the emotional healing category)
  • Tag popularity and combination effectiveness
  • Estimated completion rate

Use the calibrated model: Apply the estimation parameters adjusted in Step 2 to improve estimation accuracy.

Step 7: Output the optimized copy file

Output the copy to the 未发布/ folder, file name format:

[类型]_[主题]_[日期]_第X批.md

The file contains the following content:

  1. Copy body + selected tags (4-6)
  2. Overall score (must be 5 stars)
  3. Video analysis (only show advantages; risks have been automatically avoided)
  4. Estimated play count (baseline estimate + viral upper limit)
  5. Actual play count section (to be filled in by the user the next day)

Output template structure

The generated copy file strictly follows the structure below:

# [文案主题]

## 📝 文案正文

[生成的完整优化文案]

## 🏷️ 推荐标签

#标签1 #标签2 #标签3 #标签4

## 🎯 综合评分

- 内容质量分：XX/100
- 算法适配分：XX/100
- 创新度分：XX/100
- **综合爆款指数：★★★★★（XXX分）**

## 📊 视频分析

### ✅ 内容优势
- 优势点1：[具体说明]
- 优势点2：[具体说明]
- 优势点3：[具体说明]

### 💡 创作建议
- BGM推荐：[音乐风格建议]
- 视觉建议：[画面/文字动画建议]
- 节奏控制：[时长和节奏建议]

## 📈 预估播放量

- **基础预估**：XX-XX万
- **爆款上限**：XX万+
- **完播率预估**：XX-XX%

**评分依据**：
[基于当前得分、历史数据、赛道特征的详细分析]

## 📊 实际播放量（待填写）

- 实际播放：_____
- 点赞数：_____
- 转发数：_____
- 评论数：_____
- 完播率：_____
- 发布时间：_____

## 🔍 复盘分析

[当用户填写实际播放量后，下次生成新文案时会自动分析这条数据，并融入到策略优化中]

Key principles

✅ Automation principle

  • All analysis, optimization, and iteration are completed automatically
  • The user does not need to manually trigger any subtasks
  • The user does not need to understand the optimization process, only see the final 5-star copy

✅ Quality control principle

  • Must reach the 5-star standard before output
  • Automatically identify problems and iteratively optimize
  • Automatically avoid risks without showing them to the user

✅ Learning and evolution principle

  • Read historical data on every generation
  • Automatically identify and reinforce successful patterns
  • Continuously calibrate and optimize the estimation model

✅ User experience principle

  • Minimal input: only provide the original content
  • Perfect output: guarantee 5-star quality
  • Hidden process: do not show intermediate steps

Reference resources

This skill includes the following reference files, which are automatically read as needed:

  • references/viral-factors.md: Detailed analysis of the 9 major viral elements and scoring standards
  • references/scoring-system.md: The 300-point scoring mechanism and 5-star conversion rules
  • references/estimation-model.md: Play count estimation algorithm and parameter adjustment methods
  • references/optimization-guide.md: Automatic optimization rules and iteration strategies
  • references/learning-guide.md: Historical data analysis methods and strategy calibration guide