Verified AI Recipe

Idea to Published Faceless Video Workflow

A complete step-by-step production blueprint to research, script, voice, animate, and publish high-retention faceless videos in under 90 minutes.

Primary Goal:Produce a high-retention YouTube Short or 5-minute video essay with zero camera appearance.

Required Inputs:

Target topic or news trend
Claude 3.7 or ChatGPT for scripting
ElevenLabs account for narration
Runway Gen-3 or Midjourney for visual assets
CapCut Desktop for final timeline assembly

Expected Outcome:A polished, export-ready 1080x1920 MP4 video with animated kinetic captions, atmospheric sound design, and 70%+ retention score.

Chained Tools:ClaudeElevenLabsRunway Gen-3MidjourneyCapCut AI

Workflow Execution Steps

1Step 1

Generate a 3-second hook and structured retention script with Claude 3.7 Sonnet.

You are a YouTube retention editor with 500M+ views. Write a 60-second viral Short script about [TOPIC]. Follow the formula: [0-3s Hook that defies expectations] -> [3-15s Context & Stakes] -> [15-45s Rapid 3 Key Insights] -> [45-55s Climax & Takeaway] -> [55-60s Seamless loop into beginning].

2Step 2

Generate voiceover audio with ElevenLabs using the Adam or George voice profile.

Voice Settings: Model: Eleven Multilingual v2 | Stability: 45% | Clarity + Similarity: 85% | Style Exaggeration: 10%

3Step 3

Generate cinematic 9:16 or 16:9 visual clips in Runway Gen-3 and Midjourney.

cinematic film still of ancient library glowing under moonlight, 35mm photograph, dramatic rim lighting, shallow depth of field --ar 9:16 --v 6.1 --style raw

4Step 4

Assemble footage, snap auto-captions, and add rhythmic SFX in CapCut.

Verification Method

Check audio levels at -14 LUFS, ensure captions never obstruct the bottom 20% on mobile, and verify the first 3 seconds contain no logos or silent gaps.

Limitations & Trade-offs

AI voices without varied pacing can sound repetitive in videos exceeding 15 minutes without human tone adjustments.