How I Started Making Passive Income With AI‑Generated Shorts
I’ve always been the “side‑project‑junkie” type – building small tools, tinkering with APIs, and trying to squeeze a little extra cash out of the evenings after work. A few months ago I hit a wall: I wanted to keep producing content for YouTube and TikTok, but the manual editing and script‑writing took far more time than the views (and ad revenue) justified. That’s when I stumbled onto AI video automation and wondered if it could finally give me the passive income AI boost I’d been chasing.
Below is my honest journal of how I turned a vague idea into a functional, revenue‑generating pipeline using an n8n workflow. I’ll share the wins, the hiccups, and the exact steps I took so you can decide if it’s worth trying yourself.
Week 1 – The “What‑If” Moment and Early Research
I started the week feeling skeptical. My usual workflow was:
- Brainstorm a short‑form video idea (usually a trending keyword).
- Write a 60‑second script.
- Record a voiceover with my laptop mic.
- Pull royalty‑free images or short clips.
- Stitch everything together in Premiere Pro.
- Export, upload, add hashtags.
Even for a 30‑second short, this took at least 2 hours. Multiply that by 10 videos a week and I was looking at a 20‑hour commitment for less than $5 in ad revenue.
I Googled “automated video production” and immediately found a handful of AI tools promising to generate scripts, voiceovers, and even edit videos automatically. The buzzword that kept popping up? AI Shorts. I signed up for a few free trials, watched demo videos, and made a list of the features I needed:
- Script generation from a keyword or prompt.
- Image/video search that respects copyright.
- Text‑to‑speech with a natural voice.
- One‑click video rendering.
- Direct publishing to YouTube Shorts, TikTok, and Instagram Reels.
The biggest red flag: most of these services were either expensive SaaS subscriptions or required a lot of manual glue code. That’s when I remembered n8n, the open‑source workflow automation platform I’d used for syncing my GitHub issues with Slack. If I could stitch together the AI APIs inside an n8n workflow, I could keep the cost down and retain full control.
Week 2 – Building the First Prototype (and a Minor Setback)
I dove into the docs for OpenAI’s GPT‑4, ElevenLabs for voice, and Pexels for royalty‑free footage. My plan was simple:
- Trigger – A new row in a Google Sheet (the “idea bank”) starts the workflow.
- Script Node – Send the keyword to GPT‑4, ask for a concise, hook‑heavy script.
- Image Search Node – Query Pexels API for the top 5 relevant clips.
- Voiceover Node – Feed the script to ElevenLabs, get an MP3.
- Video Assemble Node – Use FFmpeg inside n8n to merge images, voiceover, and background music.
- Publish Node – Push the final MP4 to YouTube Shorts via Google API.
Setting up the Google Sheet trigger was painless, but the first FFmpeg step turned into a nightmare. I’m not a video engineer, and the command line options for scaling, transitions, and audio sync were cryptic. After three evenings of trial‑and‑error, I realized I needed a pre‑built n8n node that handled basic video stitching. A quick search turned up a community‑contributed “Video Composer” node that did exactly what I needed – albeit with a few quirks (it only accepted JPEGs, not MP4 clips). I compromised by pulling still frames from the Pexels videos using a tiny Python script, then fed those images into the composer.
The setback taught me two things:
- Expect a learning curve when mixing media APIs.
- Leverage community nodes; they can save you hours of debugging.
By the end of Week 2 I had a functional pipeline that produced a 30‑second video in under five minutes of total processing time.
Week 3 – First Batch of AI Shorts and Real‑World Results
I populated my Google Sheet with ten trending keywords (“AI video automation”, “remote work hacks”, “quick meal prep”) and hit “Run”. Within 30 minutes the workflow spat out ten polished shorts, each with a synthesized voice, royalty‑free visuals, and a clickable title.
I uploaded the first three to YouTube Shorts, TikTok, and Instagram Reels, using the same captions (adjusted for each platform’s character limit). The immediate reaction was modest – 50‑120 views each – but the engagement rate (likes + comments / views) was surprisingly high at ~12%. That’s a sweet spot for short‑form content and gave me confidence to keep going.
A minor hiccup appeared: TikTok rejected two videos because the background music was flagged as copyrighted. I quickly added a step in the workflow to pull royalty‑free music from the Free Music Archive API and replaced the track. Lesson learned – always double‑check licensing sources when automating content.
Week 4‑5 – Scaling Up, Fine‑Tuning, and Measuring Passive Income
With the workflow stable, I increased the batch size to 30 videos per week. The n8n workflow ran on a cheap DigitalOcean droplet ($5/mo) and processed the whole queue overnight. I set up a simple webhook to notify me via Slack when a video was successfully posted, so I could track performance without opening each platform manually.
After 30 days of consistent posting, the numbers looked promising:
| Platform | Total Views | Estimated Earnings* |
|---|---|---|
| YouTube Shorts | 12,300 | $4.80 |
| TikTok | 9,800 | $0 (no ad share yet) |
| Instagram Reels | 4,200 | $0 (no ad share) |
| Grand Total | 26,300 | ≈ $5 |
*Earnings are based on YouTube Shorts fund payouts (average $0.0004 per view). TikTok and Instagram don’t yet offer direct ad revenue for Shorts, but the follower growth is valuable for future sponsorships.
While $5 isn’t life‑changing, the time investment dropped from 20 hours/week (manual) to under 2 hours/week (monitoring). That’s the real win: a passive income AI stream that scales with minimal effort.
Week 6 – Adding Automation for Thumbnails and Analytics
I realized the workflow still required manual thumbnail creation. To close that loop, I added an extra node that generated a bold text overlay on a random frame from each video using ImageMagick. The thumbnail URLs were then pushed back to the Google Sheet for later reference.
Next, I hooked the YouTube Analytics API to pull daily view counts and revenue into the same sheet. Now I can see, at a glance, which keywords perform best and iterate on the script prompts accordingly. This content automation loop is the secret sauce – the more data you feed back, the smarter your AI‑generated scripts become.
My Takeaway and Recommendation
If you’re like me – juggling a day job, a side hustle, and a desire to experiment with AI – the combination of an n8n workflow and accessible AI APIs can turn a chaotic, time‑heavy content creation process into a sleek, semi‑autonomous machine. The key takeaways:
- Start small: a single keyword, a single video, and iterate.
- Embrace community nodes and open‑source tools; they cut the learning curve.
- Expect a couple of setbacks (FFmpeg quirks, copyright flags) and plan for quick fixes.
- Track performance metrics to refine prompts and keyword selection.
The entire system I built is packaged as a ready‑to‑go workflow called AI Shorts Factory. It’s an n8n workflow that costs a one‑time $20 and handles everything: AI script generation, image search, voiceover, video production, and auto‑posting to YouTube, TikTok, and Instagram.
The tool I'm using is called AI Shorts Factory (https://8622430312019.gumroad.com/l/gujqfy) — it's an n8n workflow that costs $20 one-time and handles everything: AI script generation, image search, voiceover, video production, and auto-posting to YouTube, TikTok, and Instagram.
Give it a spin, tweak the prompts to your niche, and you’ll see how quickly AI video automation can become a reliable side‑income engine. Happy automating!
Top comments (0)