Automate YouTube Channel in 24 Mins [System Breakdown]

12 min readBy Johnathan Chen

Learn how to build production-grade AI workflows that automate video editing, generate thumbnails, and discover viral content ideas using Claude Code and agentic systems. Complete guide with real implementation examples.

Video production is time-consuming. Between editing, creating thumbnails, and researching content ideas, creators spend countless hours on repetitive tasks. But what if you could automate 90% of this workflow using AI agents?

In this guide, we'll explore three production-grade agentic workflows Nick built using Claude Code that have transformed video production processes:

Workflow #1

AI Video Editor

Automatically cuts, enhances, and uploads videos to YouTube

Workflow #2

AI Thumbnail Generator

Creates professional thumbnails using AI face-swapping

Workflow #3

AI Outlier Detector

Finds viral video ideas by analyzing high-performing content

The Directive-Orchestration-Execution Framework

Before diving into each workflow, it's important to understand the framework that makes these agentic systems reliable and maintainable. Nick uses what he calls the DOE Framework:

Three-Layer Architecture

Why This Matters

By separating concerns, you get the best of both worlds: AI flexibility for orchestration and deterministic reliability for execution. Language models are probabilistic, but your production workflows shouldn't be.

Workflow #1: End-to-End AI Video Editor

This workflow automatically edits talking head videos by removing silences, enhancing audio quality, applying color grading, and even adding intro animations. Here's how it works:

Step 1: Audio Extraction

Extract audio from the recorded video using FFmpeg. Works with any recording software (OBS, Riverside, etc.).

Step 2: Voice Activity Detection

Runs Silero VAD (Voice Activity Detection) to identify silence gaps between speech. Nick uses a 0.5 second threshold to maintain natural pacing.

Step 3: Smart Silence Removal

Automatically removes silence gaps while preserving intentional pauses. If Nick says specific trigger words, it even removes mistakes and cuts back to the previous section.

Step 4: Audio Enhancement

Applies audio normalization, EQ, and compression to ensure professional sound quality without manual mixing.

Step 5: Visual Polish

Applies color grading and adds a custom intro animation (the "swivel teaser" effect you see at the start).

Step 6: Hardware-Accelerated Export

Uses GPU acceleration to render the final video much faster than traditional editing software like Premiere Pro.

Building This Workflow

Nick built this entirely through Claude Code with zero programming knowledge required. The process took about 30 minutes:

  1. Described what he wanted: "Build a workflow that takes a video, identifies silences, cuts them, finds mistakes, and adds an intro"
  2. Claude Code researched existing solutions and presented 3 options
  3. Tested all three approaches in parallel
  4. Selected the VAD-first approach with Silero
  5. Iterated on details (thresholds, audio processing, intro animation)

Workflow #2: Cross-Niche Outlier Finder

This workflow helps discover viral content ideas by finding high-performing videos across different niches. It calculates an "outlier score" by comparing a video's views to the channel's average performance.

How Outlier Detection Works

1. Calculate Base Score

Outlier Score = Video Views ÷ Channel Average Views

2. Apply Recency Boost

Recent videos get multiplied by a higher factor since they haven't had as much time to accumulate views.

3. Cross-Niche Modifiers

  • +30% if title mentions money/income
  • +20% if title mentions time/speed
  • +15% if title mentions specific numbers

The workflow then fetches transcripts, summarizes them with AI, generates alternative titles tailored to your channel, and outputs everything to a Google Sheet for easy review.

Real Example Output

Video: "Launching a $100M Startup from Our Living Room"

Outlier Score: 8.4x

Why It Works: Combines aspirational outcome ($100M) with relatable setting (living room), creating intrigue

Adapted Title Example: "Building a Multi-Agent System from My Apartment"

Key Insight

Nick used the TubeLab API for this, which he discovered by asking Claude Code to research available YouTube scraping APIs. He had never heard of TubeLab before—the AI did the competitive analysis in seconds.

Workflow #3: AI Thumbnail Generator with Face Direction Matching

Creating professional thumbnails is an art form. This workflow uses AI face-swapping to recreate high-performing thumbnails with your face, but with a critical innovation: face direction matching.

✕ The Common Mistake

Most people try to swap faces without considering angle. If the source thumbnail has someone facing left but your reference photo is head-on, the AI has to guess what your face looks like from that angle. Result: uncanny valley weirdness.

The Solution: Face Direction Matching

Nick's workflow analyzes face direction (yaw and pitch) using MediaPipe, then selects the reference photo that most closely matches the source thumbnail's angle using Euclidean distance in pose space.

Translation: It picks the photo of you that's facing the same direction as the person in the thumbnail, resulting in dramatically more realistic output.

Step 1: Face Direction Analysis

Uses MediaPipe to detect face landmarks and calculate yaw (left/right rotation) and pitch (up/down rotation) of the face in the source thumbnail.

Step 2: Reference Photo Selection

Analyzes your library of reference photos and calculates Euclidean distance in pose space to find the best match. Automatically selects the photo where you're facing the same direction.

Step 3: AI Face Swap

Sends the source thumbnail and matched reference photos to Replicate (using Replicate's face swap model) with optimized prompts. Generates three variations per run.

Step 4: Iterative Refinement

Optionally runs an edit pass to adjust text, colors, or background elements to match your brand style.

How to Build These Workflows Yourself

The beauty of modern AI agents is that you don't need to be a programmer. Here's Nick's exact process for building any agentic workflow:

1

Identify the Task

What do you do repeatedly in your workflow that feels mechanical? That's your target for automation.

2

Describe the Outcome

Tell Claude Code what you want in plain English. Use voice transcription to save time. Example: "Build a workflow that takes a video, removes silences, enhances audio, and exports."

3

Request Multiple Approaches

Ask for 3-5 different technical approaches. Let Claude Code research available APIs, libraries, and tools. Don't assume you know the best solution.

4

Test in Parallel

Spin up multiple Claude Code instances and test all approaches simultaneously. See which one works best for your use case.

5

Iterate and Refine

Once you have a working prototype, iterate on details. Adjusting thresholds, tweaking parameters, and improving output quality.

6

Document as Directives

Once it's working, ask Claude Code to convert the workflow into a directive file. This makes it reusable and maintainable.

Time Investment

Each of these workflows took Nick 30 minutes to 1 hour to build, including testing. Previously, building custom video editing software would have cost tens of thousands of dollars and months of development time.

API costs are minimal—Nick spent about $25 total building all three workflows, including testing iterations.

Results & Impact

Time Saved Per Video

2-3 hrs

From 3-4 hours of manual editing down to 30 minutes of review and QA

Weekly Time Savings

10+ hrs

Producing 4-5 videos per week, saving 2-3 hours each

Cost to Build

$25

Total API costs for building and testing all three workflows

Equivalent Custom Dev

$20k-50k

Estimated cost if hiring developers to build this custom software

The Real Breakthrough

These workflows don't just save time—they're better than manual work. The video editor is more consistent, the outlier detector finds ideas you'd never discover manually, and the thumbnail generator produces variations you couldn't create yourself. AI isn't just matching human capability; it's exceeding it for these specific tasks.

Key Takeaways

1. Separation of Concerns is Critical

Use the Directive-Orchestration-Execution framework to separate what you want (directives) from how it's done (execution). Let AI orchestrate, but rely on deterministic code for execution.

2. You Don't Need to Know the Tools

Nick had never heard of Silero VAD, TubeLab, or MediaPipe before building these workflows. Claude Code researched and recommended them. Your job is to define the problem, not know every solution.

3. Test Multiple Approaches in Parallel

Don't commit to the first solution. Ask for 3-5 approaches and test them simultaneously. The winning approach might surprise you.

4. Domain Details Matter

The face direction matching breakthrough came from understanding the domain problem (thumbnails), not just throwing AI at it. Combine your domain expertise with AI capabilities.

5. Software Moats Are Disappearing

You can now build custom software that would have cost $50k+ in less than an hour. The barrier isn't technical skill anymore— it's identifying what to build and having the agency to do it.

Conclusion

The question isn't "Can AI automate your job?" but rather "Can AI automate 90% of 100,000 people's jobs?" And the answer is yes—we're already there.

The only real barrier now is agency. Can you identify a problem in your workflow and decide to solve it? These workflows aren't magic—they're the result of systematically breaking down tasks, testing solutions, and iterating until they work.

Start by listing everything you do repetitively. Pick one task. Spend 30 minutes with Claude Code exploring solutions. You might be surprised at what becomes possible.

Want to Learn More?

Learn how to build AI agent systems and automation workflows that scale your business without hiring.

Get in Touch

Source & Attribution

This article synthesizes learnings from Nick Saraev's video on agentic workflows, combined with additional insights for educational purposes.

Original Content:
f*ck it. Here's how I automated my youtube channel in 24 mins (I show everything)

by Nick Saraev • Watch the full video walkthrough on YouTube