Blog to Audio: Claude + Make

A long-form essay blogger writes the kind of pieces people save for later and never read. A listen-to-this version would reach the commuters and the dish-washers, but recording herself reading each 2,000-word essay is not happening, and a raw text-to-speech of the full post, with its asides and footnotes, sounds terrible.

This build makes the audio version properly. When she publishes, Claude trims the post into a clean listening script, smoothing the asides a reader skims but a listener trips over, and a voice model reads that script into an audio file saved to Drive. The essay gets an audio twin without a microphone.

We wire it in Make, part of our video and voice set. It is a two-model build: Claude does the editing, the voice model does the speaking. Asking Claude to adapt the text for listening first is what makes the audio actually pleasant.

What does this workflow actually do?

In one line: a published post becomes an audio version. Make watches WordPress. When you publish, Claude rewrites the post into a listening script, a voice model turns that into audio, and Make saves the file to Drive. You add the audio player to the post yourself.

A few real cases, none of them the usual ones:

• The essay blogger above, every post getting a listenable twin.
• A recipe site adding a hands-free audio version for cooks at the stove.
• A local-news site making audio versions for the commute.
• A disability-advocacy group making its posts accessible by audio.

Writing the post is the work. Turning it into something pleasant to listen to, not a robotic reading of footnotes, is the part that stops people. That adaptation is what Claude is good at.

Why use two AI tools instead of one?

Because they do different halves. Claude is the strongest at reading your post and rewriting it into a script that flows when spoken, cutting the asides and parentheticals that read fine but sound clumsy aloud. The voice model is built to turn a script into natural speech. Ask either to do both and you get a worse result.

Make is the wiring that lets them hand off: post to Claude to the voice model to Drive, on publish. This is the multi-model idea again, the right tool for each step. I tried feeding the raw post straight to text-to-speech, and it read the footnotes and the captions aloud; Claude’s script pass in the middle is what fixed it.

What do you need before you start?

• A Make account. The free 1,000 operations a month covers a regular publishing pace.
• A Claude API key (for the script) and an OpenAI API key (for the voice). New to keys? See how to use Claude.
• A WordPress site Make can watch for new posts.
• A Google Drive folder to save the audio files into.

Two Make words. A scenario is the whole publish-to-audio automation. A module is one box. Four boxes, because two models each get a step.

How does the workflow work, step by step?

Four modules, left to right:

Module	App	What it does
1. Trigger	WordPress	Fires when you publish a post
2. Editor	Claude	Trims the post into a listening script
3. Voice	OpenAI	Reads the script into audio
4. Output	Google Drive	Saves the audio file

Step 1: Watch WordPress for new posts

Create a scenario and add WordPress, Watch posts, connect your site, and watch published posts. This trigger hands the next step the title and content.

Step 2: Let Claude write the listening script

Add a Claude module (Anthropic), Create a Prompt, paste your Claude key, pick Sonnet. Its job is to adapt, not summarize:

Rewrite this blog post as a script to be read aloud.
Keep the full ideas, but drop asides, parentheticals, footnotes, and
anything visual like "see the chart". Add a one-line spoken intro.
Post: {{title and content from Step 1}}

Adapting for the ear, not just reading the text, is what makes the audio pleasant instead of robotic.

Step 3: Let OpenAI voice the script

Add OpenAI, Generate speech from text, paste your OpenAI key, pick a voice, and map Claude’s script from Step 2. The model reads the cleaned script into an audio file.

Step 4: Save the audio to Drive

Add Google Drive, Upload a File, map the audio from Step 3, named from the post title. The file waits in Drive for you to add an audio player to the post, so you keep the final say.

How do you run it and check the result?

Click Run once on a recent post. When I tested this, the two-step hand-off was the difference: Claude’s script read smoothly, where feeding the raw post to the voice model read the image captions aloud. Tune Claude’s script prompt, then turn the scenario on so each post gets an audio version waiting in Drive.

After that, every essay quietly gets an audio twin, and the readers who would rather listen finally can, without you ever touching a microphone.

What does this cost to run?

Piece	Free tier	If you outgrow it
Make	1,000 operations/month free	Core plan from about $9/month
Claude API	Pay per use	A script on Sonnet costs a fraction of a cent
OpenAI voice	Pay per character	Voicing a post runs a few cents
WordPress + Drive	Your site, free Drive	No extra cost

Each post is a sliver of Claude plus a few cents of voice. The two keys are the only setup beyond Make. More on tiers in our Make guide.

What can go wrong, and how do you avoid it?

• The audio reads junk like captions. The Claude script step prevents that; keep its instruction to drop visual asides.
• It publishes the audio automatically. It does not; it saves to Drive so you place the player yourself.
• Two bills to watch. A little to Claude, a little to the voice model; both small, but two.
• Long posts make long audio. Very long posts cost more to voice; consider an audio version only for your best pieces.

The same multi-step hand-off powers many builds. See Make AI scenarios.

How do you build this in Zapier or n8n instead?

The four jobs are the same; only the names change.

Job	Make	Zapier	n8n
Catch a new post	Watch posts	New Post in WordPress	WordPress Trigger node
Write the script	Claude module	Claude (Anthropic) action	Anthropic node
Voice it	OpenAI speech	OpenAI TTS action	OpenAI node
Save the file	Upload a File	Google Drive Upload	Google Drive node

Make and Zapier are friendliest for a multi-step build. Zapier vs Make vs n8n compares all three.

Frequently asked questions

Why not feed the post straight to text-to-speech?

Because it reads everything, including captions and footnotes. Claude’s script step adapts the post for the ear first, which is what makes it sound good.

Can I use ElevenLabs for the voice instead?

Yes. Swap the OpenAI voice module for ElevenLabs. Claude still writes the script; the voice model is your choice.

Will it add the audio player to my post?

This build saves to Drive so you stay in control. You can add a WordPress media step later once you trust the output.

How long can a post be?

Longer posts cost more to voice and make longer audio. Many sites add audio only to their best or longest pieces.

Is the chat app the same as the API?

Same models, different door. Make needs API keys from the consoles, not the chat logins.

Sources and references

• Creating a scenario (Make Help)
• Claude API overview (Anthropic)
• OpenAI text-to-speech (docs)
• WordPress REST API
• Text-to-speech — Grokipedia

Last reviewed: May 2026. Make, Claude, OpenAI, and WordPress update their interfaces; check the official pages for exact button names.

You may also like

• AI Automation for Beginners (the hub)
• Transcribe a recording with Whisper + Make
• Draft WordPress posts from a spreadsheet with Claude
• Make.com Complete Beginner’s Guide
• How to use Claude AI
• Zapier vs Make vs n8n: which to start with
• The AI Tools Directory