I Cloned My Voice and Built an AI Podcast (My Parents Were Shocked)
How I built a podcast that sounds exactly like me using AI voice cloning. 15 episodes published without recording a single word.
Key Takeaways
- Published 15 AI-generated podcast episodes without recording a single word - 90% of listeners can't tell it's AI
- The 5-step workflow uses Needle (content indexing + script generation) and ElevenLabs (voice cloning) for end-to-end automation
- Total cost: under $10/month. Time per episode: ~5 minutes (just uploading). Initial setup: ~2 hours
- 3 years ago this would have cost $50,000 and a team of engineers - now anyone can do it in an afternoon
Yesterday, I published an episode of my Spotify podcast. My voice. My tone. My style.
Except... I didn't record it.
When I played it for my parents, they were convinced it was me. "When did you record this?" my mom asked.
"I didn't," I told her. "AI did."
The look on their faces was priceless.
Here's the thing: I haven't actually recorded a podcast episode in weeks. Yet I've published 15 episodes. All of them sound like me. Most listeners have no idea they're AI-generated.
Why I Built This
Let's be honest: starting a podcast is exhausting.
You have to research topics, write scripts, record (and re-record when you mess up), edit audio, upload, publish... and then do it all over again next week.
I wanted the benefits of podcasting (building authority, reaching new audiences, consistent content) without burning 10 hours a week on production.
So I asked myself: What if AI could handle all of this?
Turns out, it can.
The 5-Step Workflow: How It Works
Here's the system I built using Needle.app (AI workflow automation) and ElevenLabs (voice cloning):
- Auto-Index Content - Needle's Gmail connector automatically pulls in the TLDR AI newsletter daily. You could also use industry newsletters, Reddit threads, blog posts, or any text source.
- AI Writes the Script - A Needle AI agent reads the content and generates a conversational podcast script (~900 characters / ~5 minutes of audio), matching my hosting style for "The Knowledge Pod."
- Clone My Voice - I uploaded a few videos to ElevenLabs, which analyzed my tone, cadence, inflection, and pauses to create a voice model. The result is uncanny - even I sometimes forget it's not me.
- Generate the Audio - The Needle workflow passes the script to ElevenLabs with my voice ID. Within seconds, it creates an MP3 file. No recording. No editing.
- Auto-Create the Episode - The workflow generates a title, creates episode metadata, and sends me a Slack notification with the MP3 link. I just drag and drop into Spotify for Podcasters.
The Results After 15 Episodes
90% of people can't tell it's AI. I've tested this with friends, colleagues, and random listeners. Most assume it's me. Some even compliment my "recording setup."
The quality is shockingly good. Early AI voices sounded robotic. ElevenLabs sounds human. Real emotion, natural pauses, actual personality.
I'm building authority without lifting a finger. People see me as a "daily podcast host" now. I get DMs asking about episodes. I'm building an audience. And I haven't recorded a single word.
AI Podcast: Cost and Time Breakdown
| Item | Traditional Podcast | AI Podcast (Needle + ElevenLabs) |
|---|---|---|
| Research time per episode | 2-3 hours | 0 (auto-indexed from newsletters) |
| Script writing | 1-2 hours | Automated (~seconds) |
| Recording + editing | 3-5 hours | Automated (~seconds) |
| Total time per episode | 8-10 hours | ~5 minutes (upload only) |
| Monthly cost | $100-500+ (equipment, editing tools) | Under $10 (ElevenLabs $5 + Needle free tier) |
| Initial setup | Equipment, acoustic treatment, software | ~2 hours (one-time) |
How to Get Started in 6 Steps
I've created a template workflow in Needle that you can clone and customize. Here's what you need:
- Clone your voice - Upload 5-10 minutes of clear audio to ElevenLabs
- Set up the Needle workflow - Use the podcast template
- Connect your content source - Gmail, RSS, or whatever you want
- Customize the AI prompt - Tell it your hosting style and tone
- Run it - Let the workflow generate your first episode
- Upload to Spotify - Create your podcast and publish
The Future of Content Creation
This isn't just about podcasts. This is about scalable authenticity.
You can now create daily content without daily work. Build expertise in topics you care about. Reach audiences while you sleep. Sound like yourself without being yourself.
Three years ago, this would've cost $50,000 and a team of engineers. Today, you can do it in an afternoon for less than $10/month.
The people who adopt this early will dominate their niches. The people who wait will be playing catch-up.
Summary
I built a fully automated podcast pipeline using Needle and ElevenLabs that published 15 episodes without me recording a single word. The 5-step workflow auto-indexes content from newsletters, generates scripts with AI, clones my voice, creates MP3 audio, and sends me a notification when each episode is ready. 90% of listeners can't tell it's AI-generated. Total ongoing cost is under $10/month, with ~5 minutes of work per episode (just uploading to Spotify). What once required $50,000 and a team of engineers is now an afternoon project for anyone.
P.S. If you're thinking "this is creepy," I get it. But remember: you're cloning YOUR voice, for YOUR content. It's just a more efficient version of you.
Jan Heimes is Co-founder at Needle. When he's not building AI workflows, he's listening to podcasts that may or may not be hosted by AI clones of himself.
