How to launch a podcast with AI and no studio: from idea to first episode

TLDR: You can produce a complete podcast with no mic and no studio by chaining three AI tools: Kagi for research and scripting, ElevenLabs for the voices and Suno for the theme music. Final assembly happens in any free audio editor. From first script to published episode fits in one weekend.

This guide is for anyone who wants to launch a podcast but has no decent mic, no quiet room and no budget for voice actors.

By the end you will have a repeatable workflow: research, script, narrate, score and publish. AI covers the production, but the editorial judgment stays with you.

1. Research and script the episodes with Kagi

A podcast lives on what it says, not on how it sounds. Before touching any synthetic voice you need the topic mapped: what has already been said, what data exists and which angle is still open. Kagi is a paid search engine with no ads and no tracking, and that cleanliness matters when you research: primary sources surface before content-farm rewrites.

Two features fit this job. Lenses filter results by source type, handy for separating press from forums or technical docs. And you can boost or bury specific domains, so trusted sources always rank on top and copycat sites disappear. It starts at 5 dollars a month with a free trial, and its weak spot is exactly that: no permanent free plan.

With the research on the table comes the script, and one rule shapes everything downstream: write for the ear, not for the page. Short sentences, figures spelled out as words (“forty-two”, not “42”) and every proper noun double-checked, because a synthetic voice will read exactly what you type. The next step needs a full verbatim script, not an outline.

Write for the ear, not for the page: a synthetic voice reads exactly what you type.

2. Generate the voices with ElevenLabs

With the script locked, ElevenLabs turns it into narration: paste the text and it generates audio with intonation that is hard to tell apart from a human voice. It runs in the browser, freemium, starting at 5 dollars a month. Two paths: a voice from its catalog, ready in minutes, or cloning your own, the better option if the podcast carries your name, though cloning requires verification and a paid plan.

The decision that shapes the result most is consistency. Pick one voice per show (and one per character if you run a two-host format) and never change it between episodes: sonic identity is built by repetition, same as the theme music.

Two practical warnings. Free-plan characters run out fast, so the free tier is fine for auditioning voices but not for producing a full episode. And the commercial license comes with the paid plans: if you plan to monetize the podcast, you need one. Always listen to the complete narration before assembly. The slips cluster around proper nouns, figures and acronyms.

3. Compose the theme music with Suno

Every recognizable podcast has an intro, and until recently that meant buying library music or commissioning it. Suno generates complete songs, vocals and production included, from a text description: you write the genre, tempo and mood you want and within a minute you get a candidate. For a theme tune, ask for it to be instrumental right in the prompt, and if you want a sung tagline in the intro, it takes your own lyrics too.

The practical flow is generating several variants, keeping one and using it forever. The same intro on every episode, plus a trimmed version for transitions, is what makes a show sound like a show. Keep the tool’s known limit in mind: control over musical structure is limited, so do not try to direct it second by second. Generate, discard and select.

Suno offers free daily credits and Pro from 8 dollars a month billed annually, on web and iOS. The licensing rule matches the voices one: commercial use requires the piece to be generated on a paid plan, and a podcast with sponsorship or memberships is already commercial use. In our ranking of AI audio tools we compare both tools against the rest of the category.

4. Edit and publish

Assembly needs no studio and no paid software: any basic editor like Audacity covers what a podcast requires.

Before publishing, listen to the whole episode with headphones: it is your last chance to catch a mispronounced name or a harsh cut.

To publish you need a podcast hosting platform that generates the RSS feed, which is what Spotify, Apple Podcasts and every other app reads. On cadence, one criterion: an episode every two weeks that actually ships beats a weekly schedule abandoned in month three. If you later turn the same material into training, the workflow in our guide to launching an online course with AI starts from a similar base.

Common mistakes

The first one is publishing without listening to the full narration. Synthetic voices rarely fail, but they always fail in the same places: proper nouns, figures and acronyms. Ten minutes of listening saves a public correction.

The second is changing the voice or the theme music between episodes. Every change resets the listener’s sonic memory, and without sonic memory you do not have a show, you have loose audio files.

The third is legal: using free-plan voices or music in a monetized podcast. The commercial licenses of ElevenLabs and Suno come with paid plans, and skipping that is a problem that shows up late and ugly.

The fourth is a writing problem: scripting paragraphs meant for the page. What works on screen, long sentences with nested clauses, sounds tangled out loud. If you cannot say a sentence in one breath, cut it.

Frequently asked questions

How much does it cost to produce a podcast with this workflow?

The three bill separately, so their entry plans go one by one and with no total: Kagi from 5 dollars a month, ElevenLabs from 5 and Suno from 8 with annual billing. All three offer a free trial or starter credits to validate the workflow before paying.

Can I monetize a podcast with AI voices and music?

Yes, but you need the paid plans of ElevenLabs and Suno, which are the ones that include a commercial license. Free tiers are limited to personal use and do not cover a podcast with ads or memberships.

Will listeners notice the voice is synthetic?

With a good script, current quality is hard to tell apart from a human voice. What gives an AI podcast away is not the voice, it is a script written for the page: long sentences, unrounded figures and paragraphs with no spoken rhythm.

Do I need audio-editing skills?

No. Basic podcast assembly means aligning tracks, ducking the music under the voice and normalizing loudness: three operations any free editor does with a ten-minute tutorial. The hard part, researching and writing well, you already did in step 1.