humanizer

name: humanizer description: Humanizer Skill

Humanizer Skill

Transform AI-generated video scripts, voiceovers, and social media copy into natural, engaging content that sounds human-written — not machine-produced.

When to Use

• Before rendering any AI-generated video script through Remotion
• When adapting voiceover text for Fish Speech / ElevenLabs TTS output
• Before publishing AI-written social media captions or descriptions
• When tone-shifting content between platforms (YouTube, TikTok, Instagram, etc.)
• After script generation by the Nexus Engine or Agent Zero brainstorming pipeline

Core Rules

1. Kill the AI Voice

AI content has telltale patterns. Strip them out.

2. Write for the Ear, Not the Page

Video scripts and voiceovers are spoken content. Read everything aloud before finalizing.

Short sentences win. Average 8-12 words per sentence for voiceover. Fragments are fine.

Contractions always. "We're" not "we are". "Don't" not "do not". "It's" not "it is".

Rhythm matters. Vary sentence length. Short. Then a slightly longer one that carries the viewer forward. Then short again.

Pause markers. Use ... or [beat] for natural breathing points in voiceover scripts.

3. Match the Platform Tone

4. Use Concrete Language

AI defaults to abstract. Push toward specific.

5. Hook in the First 3 Seconds

Every video script needs a pattern interrupt. Options:

• Contrarian: "Everything you know about X is wrong."
• Specific number: "I made $4,200 in one week with this."
• Question: "Why do 90% of creators fail at shorts?"
• Statement of fact: "This tool exists and nobody's talking about it."
• Visual hook: Describe an unexpected opening shot.

Never start with "Hey guys, welcome back to..." — the scroll-away rate is brutal.

6. Write CTAs That Don't Feel Like CTAs

Prompt Templates

Video Script Humanization

Rewrite this AI-generated video script to sound like a real person talking to camera. Rules:
- Contractions everywhere
- Sentences under 15 words average
- Kill filler phrases (important to note, let's dive in, without further ado)
- Add [beat] markers where natural pauses belong
- Keep the core information intact
- Hook must land in first 5 words

Platform: {platform}
Target length: {duration}
Script: {raw_ai_script}

Voiceover Text Cleanup

Clean this voiceover text for TTS synthesis. Rules:
- Remove anything that sounds robotic when spoken
- Replace complex words with simpler alternatives (utilize → use, commence → start)
- Break long sentences into 2-3 shorter ones
- Add natural rhythm: vary sentence length
- Remove parenthetical asides (TTS reads them awkwardly)
- Keep under {word_count} words

Text: {raw_voiceover_text}

Social Media Caption

Rewrite this AI caption to sound human. Rules:
- Platform: {platform}
- First line must hook (no "Excited to share...")
- Use contractions and casual tone
- {emoji_policy} (none / sparingly / OK)
- End with a question or soft CTA, not "link in bio"
- Keep hashtags to {hashtag_count} relevant ones, not 30 generic

Caption: {raw_ai_caption}

Multi-Platform Adaptation

Adapt this content for {target_platform}. Original was written for {source_platform}.

Adjust:
- Length (shorter for TikTok/Reels, longer for YouTube/LinkedIn)
- Tone (more casual for TikTok, more professional for LinkedIn)
- Structure (visual-first for Reels, narrative for YouTube)
- CTA style (platform-native)

Content: {original_content}

Anti-Patterns Checklist

Before finalizing any content, verify none of these slipped through:

• No "In today's video/article/post..."
• No "Let's dive in / jump in / get started"
• No "It's important to note that..."
• No "Without further ado"
• No "Game-changer / game-changing" without specifics
• No "Various / numerous / several" — name the count
• No "As we all know..." — if we all know, don't say it
• No "At the end of the day..."
• No "It goes without saying..." (then why are you saying it)
• No passive voice where active works ("The video was created by..." → "I made...")
• No hedging ("You might want to consider perhaps..." → "Do this")
• No listicles that promise "7 ways" but pad with filler

TTS-Specific Guidelines

When humanizing for Fish Speech or ElevenLabs synthesis:

• Avoid tongue-twisters: "She sells seashells" patterns trip up TTS
• Spell out numbers under 10: "three" not "3" (TTS reads "three" more naturally)
• Use phonetic spelling for unusual words: "synonym" → "SIN-uh-nim" if pronunciation matters
• Break on punctuation: Periods and commas create natural pauses. Don't put 3 sentences without any punctuation.
• Avoid slash-separated words: "and/or" sounds robotic. Pick one.
• No URLs in voiceover: TTS reads URLs character by character. Say "link in description" instead.

Style-Specific Notes

For the 25 Nexus Engine video styles, adjust tone accordingly: