No need to learn editing! Use AI to easily recreate the talking-head video style of “Xiao Lin Says”

AI allows us to create text effortlessly, and by connecting to image generation APIs, we can also create graphics with ease. Can we go a step further and use AI to effortlessly create videos? Fortunately, CodeX’s HyperFrames plugin greatly lowers the barrier to creating educational videos.

But what if I want to create a talking-head video like “Xiao Lin Says”, where the visuals are real images/videos relevant to the script, and when presenting data, the corresponding MG (Motion Graphics) animations are needed? How should I do that?

A common technical route would be:

• Use TTS to generate audio, and use remotion to splice clips into a video.
• Use frontend techniques for animation effects.

But this has problems:

• You need to prepare a large number of image assets and also need an image retrieval system to find images relevant to the script.
• MG animations require a large number of presets.

To solve the challenge of making talking-head videos in the style of “Xiao Lin Says”, today I introduce an AI tool — a video production Agent called “Peanut 🥜” from Bilibili.

What is Peanut 🥜?

“Peanut 🥜” is a video production agent specially designed for long-form video creators. It can turn text and audio into professional videos, allowing creators to focus on content creativity without learning complex editing.

Advantages of Peanut 🥜 over traditional editing tools:

• Zero-barrier generation: Input subtitle text or upload a voiceover audio, and AI automatically analyzes the content, creates a plan, and generates a complete video.
• Conversational editing: Modify video content in real time using natural language.
• High-definition asset library: Tens of millions of HD images/video assets covering all categories.
• Native MG animations: AI automatically recognizes data, concepts, and logic, generating dynamic charts and progressive animations.
• AI voice cloning: Generate a custom cloned voice with 10 seconds of recording.

Making a Slide-style Video with Peanut

Step 1: Prepare the Voiceover Script

First, you need to convert your article-type content into a video voiceover script. Here is a prompt:

You are a professional video content creator, skilled at rewriting articles into natural-sounding voiceover word-for-word scripts suitable for video. Please rewrite the article I provide according to the following rules:

【Core Goal】 The rewritten content should “sound natural when read aloud, and be comfortable to listen to”. There should be no expressions that cause the speaker to pause or hesitate.

【Rewriting Rules】

1. Written to Spoken

   • Replace all formal written expressions with colloquial ones.
   • Example: “此外” → “另外”/“还有”; “然而” → “但是”/“不过”; “已然” → “已经”; “鉴于此” → “所以”

2. Replace Awkward Words

   • Identify words that are hard to pronounce or unsuitable for speech, and replace them with spoken equivalents.
   • Example: “发心” → “初衷”/“出发点”; “审计” → “财务检查(审计)”; “极大地减小” → “大幅降低”; “赋能” → “帮助”/“让…能够”

3. Break Long Sentences into Short Ones

   • If a sentence exceeds 25 characters or contains multiple clauses, split it into several short sentences.
   • After splitting, each sentence should have a complete meaning. Connect sentences with spoken conjunctions.

4. Expand Numbers and Symbols

   • All numbers must be written in the way they are spoken, to avoid the speaker having to decide the pronunciation.
   • Year: 2024年 → 二〇二四年; Percentage: 50% → 百分之五十; Multiple: 3倍 → 三倍; Currency: $100 → 一百美元
   • Replace special symbols (/, &, @, #, etc.) with their Chinese equivalents or delete them.

5. Handle Technical Terms

   • On the first occurrence, supplement with a spoken explanation in parentheses, e.g., “ROI (也就是投资回报率)”
   • For subsequent occurrences, use the already established term without repeating the explanation.

6. Natural Transitions and Connections

   • Add colloquial transition words between paragraphs to make the content flow smoothly.
   • Common transitions: “好，”, “那么，”, “说到这里，”, “接下来，”, “你可能会问，”

7. Increase Interactivity

   • Use expressions like “你”, “咱们”, “你有没有发现”, “你可能也遇到过” appropriately.
   • Add a hook sentence at the beginning to spark curiosity; add a call to action or a concluding gem at the end.

8. Remove Redundant Written Expressions

   • Delete words that are readable visually but redundant in speech: “上文提到的”, “如前所述”, “综上所述”, “由此可见”
   • Replace with: “刚才说的”, “前面提到”, “所以”, “你看”

9. Convert Lists to Spoken Enumeration

   • Change bulleted lists (• - 1. etc.) into spoken enumerations: “第一……第二……最后……”
   • If there are more than 4 items, merge similar ones or simplify to the most important 3.

10. Rhythm and Pause Markers

    • Add “——” or “……” where pauses or breaths are needed.
    • Enclose words to be emphasized with 【】, e.g., “这才是【真正的关键】”
    • Do not use bold, italics, or other typographic symbols (they are meaningless in spoken content).

【Disallowed Behaviors】

• Do not change the core viewpoints or information of the original text.
• Do not arbitrarily add facts or data not in the original.
• Do not delete important arguments or cases from the original.
• Do not confuse multiple person/brand names.

【Output Format】

• Output the word-for-word script body directly, without any explanations or notes.
• Leave a blank line between paragraphs for the speaker to breathe.
• If the original has clear sections, preserve the sense of sections, but section titles must be spoken, e.g., “好，我们先来说第一点——”

Take the introduction of European countries’ happiness index as an example.

Written version:

According to the latest Global Happiness Report, European countries continue to rank high in terms of national happiness. Data shows that Finland has topped the list for the seventh consecutive year, with a national happiness index score of 7.74. Close behind are Denmark and Iceland, with scores of 7.58 and 7.53 respectively, ranking second and third. Notably, among the top ten, European countries occupy eight seats, accounting for a total of 80%. Studies indicate that the high happiness of these countries is mainly due to their comprehensive social welfare systems, per capita GDP exceeding $60,000, and a social trust level as high as 95%. In addition, low corruption rates and high freedom are also key factors. However, in recent years, the happiness of young people in some Western European countries has seen a slight decline of about 5%, which has attracted widespread attention from academia. In view of this, researchers point out that continuously paying attention to mental health and economic resilience will be a long-term challenge for maintaining high happiness.

Voiceover script: (The 【】 will be kept in subtitles; remove them if you don’t need them)

好，你有没有发现，每年的全球幸福报告一出来，排在前面的【总是那几个欧洲国家】? 刚才最新的数据显示啊，芬兰已经连续第七年拿到第一名了，他们的国民幸福指数得分高达七点七四分。—— 紧跟在后面的呢，是丹麦和冰岛，分别以七点五八分和七点五三位居第二和第三。你可能不知道，在排名前十的国家里面，欧洲国家就占了八个席位，整体比例高达【百分之八十】! 那么，说到这里你可能会问了，为什么他们能感觉这么幸福呢? 其实，这背后的原因主要有三点: 第一，是他们有非常完善的社会福利体系; 第二，是大家的口袋都挺鼓的，人均国内生产总值，也就是人均吉迪匹，超过了六万美元; 最后一点，也是非常重要的一点，就是他们人与人之间的信任度高达百分之九十五。另外，社会环境非常廉洁，大家的生活自由度也很高。不过，这里面也出现了一个新情况。最近几年，部分西欧国家的年轻人，他们的幸福感其实出现了大约百分之五的微幅下滑。—— 这个现象也引发了专家们的广泛关注。所以你看，研究人员就说了，以后想要长期维持这么高的幸福感，除了看经济好不好，能不能持续关注大家的心理健康，【这才是真正的关键】。

Step 2: Create the Video Script

Using the Input Script mode as an example, paste the voiceover script, select a voice, and start creating.

After AI analysis, it finds the content involves number comparisons, ranking displays, etc., and recommends using the B - Asset Mixed with MG Animation mode to better present the MG animations. But for now, let’s first choose the A - Asset Editing to Video mode, and later compare the differences between the two modes.

After AI analysis, it provides detailed video production requirements. The circled items in the image are the requirements for visual assets.

For example, I ask it to use materials from 2025-2026 as much as possible, and describe the modification with natural language.

After modification, AI provides a Production Plan, or Video Script, for your confirmation. If you have revisions, describe them in natural language in the dialog on the right and send.

Step 3: Start Generation

After confirming the production plan, AI starts making the video. The dialog on the right shows the current progress.

After a short wait, generation is complete. You can play the video to preview it. The first play might be a bit choppy; pause when it lags, then continue playing.

Step 4: Interactive Editing

Modification 1: Manually Replace Clip Assets

Click on the clip you want to replace, and select a new asset on the right. If no suitable asset is available, you can expand more options or ask the AI to re-match visuals.

In the pop-up window, you can preview the asset or capture a specific segment.

After replacement, the clip displays the new asset.

Note that modifications consume points (the “Peanut seeds” shown in the top right).

Modification 2: Replace Clip Assets Using Natural Language

Creators can not only manually replace clip assets but also describe their requirements in natural language in the right dialog, and the system will replace the asset with a suitable one. For example, I ask that the asset for this clip should not include flags of other countries.

The new clip after replacement indeed does not contain flags of other countries.

Modification 3: Split or Merge Clips

Take splitting a clip as an example. Notice that the last clip corresponds to a longer word-for-word script:

以后想要长期维持这么高的幸福感，除了看经济好不好，能不能持续关注大家的心理健康，【这才是真正的关键】。

Use natural language to ask the AI to split the last clip into 2 clips.

Final Video Showcase

Adding MG Animations with Peanut

Step 1: Start Generation

Similarly, paste the voiceover script.

Here I use my preset “Modern Digital Girl Vibe”, which mainly influences the style of MG animations.

Prompt:

MG Animation Style Description

I. Overall Style Positioning

Core style: Modern Digital Girl Vibe × Flat MG Animation

Combine the softness and cuteness of anime with the simplicity and efficiency of modern infographics, creating an animation style that is both visually impactful and friendly. Suitable for knowledge sharing, marketing promotions, data presentations, etc.

---

II. Color Scheme

Main Colors

• Pink series: #FFB6D9 (Soft Pink), #FF85C0 (Vibrant Pink), #FF4D94 (Deep Pink Accent)
• White series: #FFFFFF (Pure White), #F8F8F8 (Soft White Background)
• Gold series: #FFD700 (Gold Particles), #FFC700 (Gold Highlight)

Auxiliary Colors

• Gradients: Soft gradient from pink to white (for titles, key text)
• Accent colors: Light purple #E6D5FF, Light blue #D5E8FF (for info panels, icons)

Color Usage Principles

• Backgrounds use soft gradients, avoid solid flat fills.
• Key information is highlighted with high-saturation pink.
• Gold particles and light effects are used sparingly as accents.
• White info panels maintain a semi-transparent feel for depth.

---

III. Visual Elements

Icons and Symbols

• Geometric icons: Heart ❤️, Star ⭐, X ✖️, Speech bubble 💬, Checkbox ☑️
• Data icons: Line chart 📈, Bar chart 📊, Brain icon 🧠, Photo stack 📷
• Decorative elements: Glowing particles, halos, star sparkles, bow 🎀

Character Design (Optional)

• Chibi anime character:
  • Big head, small body (head-to-body ratio ~1:1.5)
  • Long dark hair, pink hair accessories (X-shaped, bow)
  • Expressive face (blinking, smiling, excited)
  • Rich gestures (pointing, heart hand, waving)
  • Outfit: pink-white uniform skirt and jacket with girly details

Info Panels

• Style: Semi-transparent white card, rounded rectangle
• Content: Title + icon + list/chart
• Layout: Left-aligned, clear hierarchy
• Shadow: Soft shadow for depth

---

IV. Motion Effects Features

Entry Animations

• Fly-in: Icons/text fly in from screen edges with elastic easing.
• Fade-in: Info panels fade from transparent to opaque with slight scaling.
• Particle burst: Gold particles burst outward from center forming a halo.

Emphasis Animations

• Blink: Key numbers/titles blink 1-2 times to grab attention.
• Pulse: Heart/star icons pulse periodically.
• Light sweep: A gold light beam sweeps left to right across text for tech feel.

Transition Animations

• Erase: Current scene is erased by particles revealing the next.
• Rotate: Info panel rotates/flips to new content.
• Zoom: Entire scene zooms into a point, then new scene zooms out.

Loop Animations

• Floating: Particles/icons float slowly across the screen.
• Breathing: Halo luminosity cycles.
• Rotation: Star/X icons slowly rotate.

---

V. Fonts and Typography

Font Selection

• Title font: Bold, rounded, powerful (e.g., Source Han Sans Bold, Alibaba PuHuiTi Heavy)
• Body font: Medium weight, readable (e.g., Source Han Sans Regular, PingFang SC Regular)
• Number font: Monospace, modern feel (e.g., DIN, Roboto Mono)

Typography Principles

• Titles: Split into lines, main title large, subtitle small, with gradient colors.
• Body: Left-aligned, 1.5 line spacing, clear paragraph spacing.
• Numbers: Extra large display, emphasized with pink or gradient.
• Annotations: Small font size, light color, not overpowering.

Text Animations

• Word-by-word appearance: Title words pop in one by one with slight bounce.
• Typewriter effect: Body text appears character by character like a typewriter.
• Number scrolling: Numbers scroll quickly from 0 to target value.

---

VI. Scene Atmosphere

Background Design

• Gradient background: Soft pink-white gradient from top-left to bottom-right.
• Particle background: Gold/white particles float slowly in the background.
• Halo effect: Soft halos around screen edges for a dreamy feel.

Spatial Layers

• Foreground: Character, main title, key icons.
• Midground: Info panels, charts, lists.
• Background: Gradient, particles, decorative elements.

Lighting Effects

• Glow: Key elements (titles, numbers, icons) have a self-glow effect.
• Shadow: Info panels and characters have soft shadows for depth.
• Highlight: Gold highlights at key positions.

---

VII. Technical Specifications

Aspect Ratio

• Horizontal video: 16:9 (suitable for YouTube, Bilibili landscape)
• Vertical video: 9:16 (suitable for Douyin, Kuaishou, Xiaohongshu)
• Cover image: 5:2 or 16:9

Resolution

• Standard: 1280×720 (720p)
• HD: 1920×1080 (1080p, recommended)
• UHD: 3840×2160 (4K, for high-quality display)

Frame Rate

• Standard: 30fps (smooth, small file)
• High: 60fps (smoother, for fast-paced animation)

Animation Duration

• Single effect: 0.3–0.8 seconds (quick, not dragging)
• Transition: 0.5–1 second
• Full clip: 2–5 seconds

---

VIII. Style Reference Keywords

Use the following combination of keywords for video generation:

MG animation style, 2D flat, soft pink and white tones, gold particles and halos, chibi anime character, modern digital feel, girl vibe, infographic, white semi-transparent cards, heart and star icons, glow effects, gradient text, elastic animation, clean and simple, high saturation, rounded design, 16:9 landscape

---

IX. Mood and Rhythm

Mood Tone

• Positive and uplifting: Bright colors, light animations.
• Friendly and cute: Chibi characters, soft lines.
• Professional and credible: Info panels, data charts.

Rhythm Control

• Opening: Quickly grab attention (0–2 sec).
• Development: Information appears gradually (2–8 sec).
• Climax: Key data emphasis (8–10 sec).
• Conclusion: Summary or call to action (10–12 sec).

---

X. Notes

1. Avoid over-decoration: Use particles and glow effects in moderation; do not obscure main information.
2. Maintain readability: Ensure text size and color contrast are legible.
3. Unified style: All elements in the same video should maintain a consistent style.
4. Platform adaptation: Adjust aspect ratio and duration according to the publishing platform.
5. Test effects: Test viewing on target devices after generation.

This time we choose the B - Asset Mixed with MG Animation mode.

The generated video script document now includes MG animation-related sections.

Some fonts are used here; to avoid copyright issues, I wanted to change them to the free-for-commercial-use Source Han Sans font.

But it didn’t replace successfully; upon inquiry, it only supports a few fonts.

After font selection, start generation. Some MG animation-related components appear.

Generating MG animations takes some time; after a short wait, generation is complete.

Since MG animations need to be shown dynamically, here we first capture some static effects; the final video showcase will display dynamic effects.

Effect 1: Bar chart displaying values

Effect 2: Donut chart showing proportions

Effect 3: Animation effects enriching the scene

Effect 4: Line chart showing trend changes (Note: AI automatically searched for data from previous years)

Note: If a clip has MG animations, the MG animation may continue playing when the video is paused. So after pausing, it’s best to go back to a clip before the MG animation before resuming play.

Step 2: Interactive Editing

Modification 1: Adjust MG Animation Colors

I want to change the color scheme of the last MG animation to Bilibili’s brand colors (I don’t know the exact colors, but the AI can search for them).

The result after modification looks like Bilibili’s brand colors are pink + blue + white.

Modification 2: Add New MG Animation

Clip 05 originally had no MG animation; add a decorative MG animation to it.

A triangle was successfully generated, but without a photo of Musk. Describe the modification in natural language.

After a few attempts, a photo was indeed added (though not of Musk).

Final Video Showcase

How to Experience

Peanut 🥜 also has other features like voice cloning.

I’ve put the link to Peanut 🥜 in the comments below; go try it out!

You can either subscribe as a member or pay per usage, which is fairly flexible. Point consumption is not fast, and it offers good value for money.

Oh, one more thing: Do not use a VPN magic or browser extensions while using it, otherwise online video playback will be very laggy, especially when MG animations are present. If it’s too laggy to play, you can also download it locally to watch, but you’ll need to wait for the video to be exported.