How to Create a Podcast with AI (Script, Voices, and Editing)

Before writing a single word, decide the basic format. It guides everything else.

Key questions:

• What’s the topic? Choose something specific. “Technology” is too broad; “AI for freelancers” is a winner.
• Who listens? Define your audience: beginners, professionals, students?
• What format? Solo (one host), conversation (two AI voices), or mock interview (AI as interviewee).
• How long? To start: 5 to 15 minutes per episode. More manageable and easier to edit.

Tip: Google’s NotebookLM generates automatic conversational “Audio Overviews” from documents. Free. Use it to explore how your topic would sound before committing to a format.

1. Open Claude (https://claude.ai) or ChatGPT (https://chat.openai.com).
2. Use a prompt like this (adjust the details to your podcast):

Write a 10-minute podcast episode script about [topic].
Audience: [description of your audience].
Format: single host, [friendly / informative / entertaining] tone.
Include: intro with hook, 3 main points with concrete examples, and closing with a call to action.
Avoid unnecessary jargon. Write the text as it will be read out loud.

3. Review the script: add your perspective, verify facts, and personalize phrases. AI gives you a base; your voice (even if synthetic) should sound like you.
4. Split the script by sections if you’ll have more than one speaker or more than one segment.

Productivity tip: save a folder with all your scripts. Over time you’ll build a content library you can reuse and update.

1. Log in to https://elevenlabs.io.
2. Go to Text to Speech in the side menu.
3. Choose your voice:
   • If you already have a voice clone (see How to Clone Your Voice with AI): select it under My Voices.
   • If you don’t have a clone: pick a catalog voice that fits your podcast’s tone.
4. Paste your script text (or the section you want narrated).
5. Adjust stability and similarity sliders to your liking.
6. Click Generate and listen to the result.
7. Download the audio (download button, MP3 format).
8. Repeat for each section or speaker in the episode.

For podcasts with two voices:

• Create (or choose) two distinct voices in ElevenLabs.
• Generate each segment separately.
• Then combine them in Descript (next step).

ElevenLabs credits are consumed per character generated. A 10-minute episode uses approximately 7,000 to 10,000 characters. Confirm your monthly limit on the official page before scaling up.

Descript lets you edit audio like a text document: delete words from the audio simply by deleting them from the transcript.

1. Go to https://descript.com and create an account (there’s a free plan with transcription limits).
2. Create a new project and upload all your ElevenLabs audio files.
3. Descript automatically transcribes the audio.
4. Edit the transcript to adjust content: cut long pauses, remove repetitions, and join tracks from different speakers.
5. If you use the Creator plan ($12/mo), enable Remove Filler Words to automatically eliminate “uh”, “um”, and silences.
6. Reorder segments by dragging blocks in the timeline.
7. When done, export as an audio file: File → Export → Audio File.

Descript can also generate a video version if you want to publish the episode on YouTube with a static cover image or animated transcript.

Even though ElevenLabs audio is already clean, this step is useful if you recorded any part live or want to improve overall quality.

Option A: Adobe Podcast (free for basic use)

1. Go to https://podcast.adobe.com and log in with your Adobe account (or create one free).
2. Upload the exported audio file.
3. Use Enhance Speech to reduce background noise and improve voice clarity.
4. Download the enhanced audio.

Option B: Cleanvoice (30-minute free trial, no registration required)

1. Go to https://cleanvoice.ai.
2. Upload your audio file.
3. Cleanvoice removes filler words, reduces noise, and trims silences.
4. Download the clean result.

Confirm current prices on the official Adobe Podcast and Cleanvoice pages before subscribing to any plan.

With the final audio ready, all that’s left is uploading it to the platforms.

1. Choose a podcast distributor:
   • Spotify for Podcasters (formerly Anchor): free, direct distribution to Spotify and other platforms.
   • Buzzsprout: free limited plan (2 hours/month); paid plans from $12/mo.
   • Podbean: free plan with 5 hours of storage; paid plans from $9/mo.
2. Create your account on the chosen distributor.
3. Upload the audio file (MP3 or WAV).
4. Fill in metadata: episode title, description, category, podcast cover art.
5. Publish. The distributor generates an RSS feed and sends it to Spotify, Apple Podcasts, Google Podcasts, and more.

Podcast cover art: you need a square image of at least 1400x1400 px. You can create it with Canva (free) or with an AI image tool like the ones explained in How to Create Images with AI.

Problem	Solution
Generated audio sounds robotic	Adjust stability to the right in ElevenLabs; try another catalog voice
Descript doesn’t transcribe well	Normal with AI voices: edit the transcript manually where there are errors
Audio file is too large	Export as MP3 at 128 kbps: sufficient quality for podcasts and manageable size
Distributor rejects my audio	Check format (MP3/WAV) and minimum episode duration
Ran out of credits on ElevenLabs	Split the episode into parts and use the monthly credit renewal