Create Your First AI Podcast With Zero Equipment: NotebookLM + ElevenLabs + Spotify Free Complete Guide

No microphone, no studio, no budget. These three excuses have kept countless podcast ideas trapped in people's heads forever. But in 2026, AI tools have turned "zero-equipment podcasting" from impossible into something you can finish on a Saturday afternoon. This article is my hands-on record of going from nothing to a published show: every step from script to voice generation to editing to publishing on Spotify, plus a quality checklist to make sure your first episode doesn't become yet another piece of AI slop buried by the algorithm.

TL;DR

• A completely free tool chain works: ChatGPT Free -> NotebookLM -> Audacity -> Spotify for Creators
• NotebookLM is the fastest route right now: 3 free generations per day, 80+ languages supported, default output is roughly 10 minutes of audio
• When you need more voice control, ElevenLabs Free (10,000 characters/month) is a solid supplement
• AI voice quality still has pacing and intonation issues; pre-publish quality checks are non-negotiable
• Legal and ethical risks are real (an NPR host sued Google in February 2026); transparent labeling beats hiding it

Why Now Is the Right Time to Start an AI Podcast

Let's look at the numbers. According to a Grand View Research report, the global podcast market hit $30.72 billion in 2024 and is projected to reach $131.13 billion by 2030 (CAGR of 27.0%). DemandSage estimates that global podcast listeners will reach 619 million by 2026. Meanwhile, the AI voice generation market is also exploding, with MarketsandMarkets projecting a CAGR of 30.7% from 2025 to 2031.

Three years ago, starting a podcast required at least $500 worth of microphones and audio interfaces. Now free AI tools handle 80% of the work. But there's a dark side: Podnews reports that some companies are already mass-producing 3,000 AI podcast episodes per week at roughly $1 per episode. This AI slop is eroding listener trust.

What does this mean? Lower barriers let more people in, but the quality divide is widening fast. Creators who take content quality seriously actually stand out more, precisely because of the contrast with all the low-effort content flooding the market.

Two Paths: NotebookLM Quick Route vs. Multi-Tool Full Route

Before you start, decide which route to take. Both are zero-cost; the difference is time investment and how much control you get.

Dimension	NotebookLM Quick Route	Multi-Tool Full Route
Production time	30-60 min/episode	2-4 hours/episode
Voice style	Dual-host conversation (fixed style)	Customizable voice, speed, tone
Control level	Low (can't pick voice characters)	High (adjust sentence by sentence)
Free quota	3 per day	ElevenLabs: 10,000 chars/month
Best for	Quick tests, knowledge sharing	Branded shows, consistent host persona

Route A: NotebookLM Quick Route

Google's NotebookLM is currently the fastest path from zero to audio. Upload your source materials (Google Docs, PDFs, web URLs) and it automatically generates a dual-host conversational audio. It supports 80+ languages.

The free tier allows 3 generations per day. Duration options include short (about 5 minutes), default (about 10 minutes), and long (about 20 minutes, currently English only). The Interactive Mode introduced in 2025 lets you join the conversation and ask questions to steer the content, though you can't interrupt the AI hosts mid-generation yet.

Key limitation: you can't customize the hosts' voice characteristics, and you can't regenerate a specific segment. If you're unhappy with a section, you have to regenerate the entire episode.

Route B: Multi-Tool Full Route

Tool stack: ChatGPT Free (script) -> ElevenLabs Free or TTSMaker (voice synthesis) -> Audacity (editing).

This route suits creators who want brand consistency, like a fixed host voice and specific pacing. The tradeoff is 3-4x more time investment.

How to choose?

• First episode, quick topic validation -> Route A (NotebookLM)
• Planning ongoing episodes, need a consistent host voice -> Route B (multi-tool)
• Want quality on a budget -> Hybrid: NotebookLM for the draft + Audacity post-production

Step-by-Step: From Script to Spotify

Step 1: Script Writing (ChatGPT Free / Gemini)

Regardless of which route you take, a good script is the foundation. When using ChatGPT Free or Gemini to generate scripts, this prompt framework has worked well in practice:

You are a podcast script writer. Write a two-person dialogue podcast script.

Topic: [your topic]
Target audience: [describe your audience]
Target length: 10 minutes (approx. 1,500-2,000 words)
Tone: Relaxed but substantive, like two experienced friends chatting
Structure: Opening hook -> Core points (3) -> Real-world example -> Wrap-up

Guidelines:
- Avoid long monologue blocks; keep each turn to 3-4 sentences
- Include natural filler words and transitions ("Right, that's the key point," "Wait, you mean...")
- Don't write in bullet-point style; make it sound like a real conversation

Length estimate: English voice synthesis runs roughly 150 words per minute. For a 10-minute episode, aim for about 1,500 words.

Common mistake: Writing scripts that are too list-like. AI voice synthesis turns these into robotic point-by-point readings that sound terrible. The fix is to emphasize "real conversation" in your prompt and manually polish transitions after generation.

Step 2: Voice Generation

Route A: NotebookLM

1. Go to NotebookLM and create a new notebook
2. Upload your materials (Google Doc format works best for compatibility)
3. Click "Audio Overview" and select duration
4. To steer the content direction, enable Interactive Mode and type the key points you want emphasized before generating
5. Wait for generation (usually 2-5 minutes), then download the MP3

Hands-on takeaway: NotebookLM's generated dialogue genuinely sounds like two people chatting, with natural responses and follow-up questions. However, note that the Audio Overview AI hosts currently support English voices only. While NotebookLM handles text in 80+ languages, voice generation is still primarily English. After upgrading to the Gemini 3 architecture in mid-December 2025, its reasoning improved and it handles complex topics more coherently.

Route B: ElevenLabs Free / TTSMaker

ElevenLabs Free: 10,000 characters per month, roughly equivalent to 10 minutes of audio. The per-generation cap is 2,500 characters, so you'll need to generate in segments and merge them. The free tier is limited to non-commercial use.

Quota management strategy: A 10-minute episode needs about 1,500 words of script. For dual-voice dialogue with each role generated separately, total character count is around 3,000. You can produce about 3 episodes per month.

TTSMaker: 20,000 characters per week with commercial use rights included. Fewer voice options than ElevenLabs, but far more generous quotas.

Voice quality comparison: ElevenLabs' English voices are noticeably better than its non-English options. TTSMaker handles sentence breaks slightly better for some languages, but overall voice naturalness doesn't match ElevenLabs. For either tool, test with short paragraphs first to find the best voice before committing to a full recording.

Step 3: Audio Editing (Audacity / GarageBand)

No matter which route you take, the generated audio needs basic post-production. Audacity is free and open-source; macOS users can also use the built-in GarageBand.

Three must-do operations:

1. Noise Reduction: Select a silent segment -> Effect -> Noise Reduction -> Get Noise Profile -> Select entire track -> Apply again. AI-generated audio typically has very low background noise, but it's still worth running once.

2. Normalize: Effect -> Normalize -> Set to -1.0 dB. Spotify's recommended loudness standard is -14 LUFS; normalizing ensures consistent volume throughout.

3. Silence Trimming: Manually cut out overly long pauses. AI voices occasionally produce unnatural long gaps between sentences; trimming them dramatically improves the listening experience.

Adding intro music: Pixabay Music and Free Music Archive offer royalty-free music. Import into Audacity as a new track and lower the music volume so it doesn't overpower the voice (recommended: drop music to -15 to -20 dB).

Export settings: File -> Export as MP3, choose 128kbps (Spotify's minimum is 96kbps; 128kbps balances quality and file size).

Step 4: Publishing on Spotify + Apple Podcasts

Spotify for Creators (formerly Spotify for Podcasters) offers completely free podcast hosting with no storage limits or monthly fees.

Spotify publishing workflow:

1. Log in to Spotify for Creators with your Spotify account
2. Create a new show: fill in show name, category, and description
3. Upload cover art (specs: 3000x3000px, JPG or PNG. Free option: use Canva's podcast cover templates)
4. Upload your first episode audio, add episode title and description
5. Submit for review (typically 1-3 business days)

Cross-publishing to Apple Podcasts:

Spotify for Creators generates an RSS feed for your show (find it under Settings -> Availability). With the RSS feed in hand:

1. Go to Apple Podcasts Connect and sign in with your Apple ID
2. Click "Add a Show" and paste the RSS feed URL
3. Submit for review (typically 3-5 business days; subsequent episodes auto-sync within 24 hours)

Apple Podcasts is also completely free to publish on. Together, these two platforms cover the vast majority of podcast listeners worldwide.

Tip: If you're based in Taiwan, consider SoundOn as a hosting platform. It's a local service, free with no upload limits, and supports automatic distribution to Spotify, Apple Podcasts, and other platforms, saving you the manual RSS submission steps.

Note: Spotify for Creators' ad revenue sharing uses a 50/50 split between creators and Spotify. But for creators just starting out, free hosting and discoverability matter far more than revenue share terms.

Quality Checklist: How to Avoid Becoming AI Slop

AI slop isn't a technology problem; it's an effort problem. When companies are already mass-producing 3,000 AI podcast episodes per week, the 30 minutes you spend on quality control is exactly what separates your work from theirs.

10-Point Pre-Publish Checklist:

Content

• Listen through the entire audio at least once, marking all unnatural breaks or pauses
• Every data point and factual claim has a verifiable source
• Content includes genuine opinions or firsthand experience, not just repackaged information
• All AI-generated names, organizations, and numbers have been fact-checked (AI states incorrect information with full confidence)

Technical

• Audio has been normalized
• No obvious background noise or AI voice artifacts (robotic repetition, unnatural pauses)
• Clear topic introduction within the first 30 seconds
• Episode length is reasonable (8-15 minutes recommended for first episodes)

Compliance

• Show description includes a note like "partially produced with AI assistance"
• All music assets have confirmed licensing terms (CC0 or royalty-free commercial use)

Risk Disclosure

Legal Risks

In February 2026, former NPR host David Greene filed a lawsuit against Google, alleging that NotebookLM Audio Overview's male AI voice replicated his vocal characteristics. AI forensic testing showed a 53-60% match confidence score. Google responded that the voice was based on a "paid professional actor." The case is still in early litigation, but it has already drawn a clear warning line for AI voice copyright issues.

What this means for individual creators: don't use AI tools to imitate any real person's voice. Sticking with a platform's default voices is the safest play.

Music licensing deserves attention too. Even "free" platform assets sometimes have commercial use restrictions. Check each license before downloading (look for CC0, CC-BY, or non-commercial-only terms).

Ethical Risks

Transparency is the baseline. If your podcast uses AI-generated voices or scripts, say so in the show description. Not disclosing AI involvement is essentially hiding something from your listeners, and it will erode trust over time.

The damage AI slop does to the podcast ecosystem is real. Every carefully made AI-assisted podcast demonstrates that AI is a tool, not a shortcut, and that's good for the entire creator community.

Quality Risks

Known shortcomings of AI voices: occasional mispronunciation of specialized terms, and unnatural pause rhythms between sentences. There's no perfect fix yet; the only approach is post-production editing and repeated testing.

The biggest hidden risk is hallucination. AI can insert inaccurate numbers or fabricated quotes into scripts, and it does so with complete confidence. All factual content generated by AI must be manually verified before publishing.

Listener Trust

One of podcasting's core appeals is the sense of personal connection between listeners and hosts. Purely AI-generated content is inherently weak on this front. The long-term strategy is to use AI as an efficiency tool, while your personal perspective, real experiences, and unique viewpoint are what keep listeners coming back.

Conclusion

The free AI tool chain has turned "zero-equipment podcasting" from a slogan into an executable workflow. NotebookLM's 3 free daily generations are enough for you to finish your first audio draft today.

But tools are just the starting point. In a market flooded with AI slop, quality control is the real moat. Spending time verifying facts, trimming audio glitches, and proactively labeling AI involvement -- these "slow" practices are exactly what mass-produced content can't replicate.

Open NotebookLM right now, drop in a topic you've been wanting to share, and generate your first audio clip. It doesn't have to be perfect. Just hear what "your own podcast" sounds like. The first episode is always the hardest, but with AI tools, it no longer has to be the most expensive.